Analysis of direct shots at goal from free kicks in elite women's football

Southern Cross University ePublications@SCU Theses 2010 Analysis of direct shots at goal from free kicks in elite women's football Alison Marie Alc...
Author: Clarissa Holt
0 downloads 0 Views 9MB Size
Southern Cross University

ePublications@SCU Theses

2010

Analysis of direct shots at goal from free kicks in elite women's football Alison Marie Alcock Southern Cross University

Publication details Alcock, AM 2010, 'Analysis of direct shots at goal from free kicks in elite women's football', PhD thesis, Southern Cross University, Lismore, NSW. Copyright AM Alcock 2010

ePublications@SCU is an electronic repository administered by Southern Cross University Library. Its goal is to capture and preserve the intellectual output of Southern Cross University authors and researchers, and to increase visibility and impact through open access to researchers around the world. For further information please contact [email protected].

Analysis of Direct Shots at Goal from Free Kicks in Elite Women's Football

by

Alison Marie Alcock BSc (Hons) - Sports Science (Biomechanics) Liverpool John Moores University

A thesis submitted for the degree Doctor of Philosophy November 2010

School of Health and Human Sciences Southern Cross University, Lismore, Australia

Declaration I certify that the work presented in this thesis is, to the best of my knowledge and belief, original, except as acknowledged in the text, and that the material has not been submitted, either in whole or in part, for a degree at this or any other university.

I acknowledge that

I have read

and

understood the

University's rules,

requirements, procedures and policy relating to my higher degree research award and to my thesis. I certify that I have complied with the rules, requirements, procedures and policy of the University (as they may be from time to time).

Miss Alison Marie Alcock

Signature: Date:

23 rd November 2010

ii

Supervisor's Declaration I certify that this thesis entitled "Analysis of Direct Shots at Goal from Free Kicks in Elite Women's Football," submitted by Alison Marie Alcock in fulfilment of the degree of Doctor of Philosophy is ready for examination.

Associate Professor Wendy Gilleard f ) \A

Signature: Date:

Q·J!.iO~·;/ . .J--...

(,

)

23 rd November 2010

iii

Abstract In elite football, approximately one third of all goals originate from set plays, and of those direct free kicks are the most effective for scoring goals. The purpose of this study was to determine the key attributes of successful direct free kicks in elite women's football and the mechanisms involved in expert performance of the skill. A method was developed that could reliably locate a football on a pitch within 0.24 m using television coverage and the official pitch markings, and then applied to all direct shots at goal from free kicks in the 2007 women's World Cup. All seven shots that resulted in a goal were taken from within 27 m of the goal, entered the goal within approximately 1 m of the goalpost, and had a significantly faster flight time than those that were unsuccessful. All shots directed towards the bottom and centre of the goal resulted in straightforward saves for the goalkeeper. This information can facilitate decisions on where direct shots from free kicks should be practised from in training and when a direct shot should or should not be attempted in competition. The attributes of a successful direct free kick were used to set up a replica laboratory-based free kick that would likely score in elite women's competition. Fifteen international female footballers performed simulated free kicks (curve kicks) and instep kicks at goal from the same location. Ball flight characteristics and full-body three-dimensional kinematics were analysed. Curve kicks had significantly greater lateral and vertical launch angles, increased sidespin and spin about the antero-posterior axis, and more spin about the vertical axis compared with instep kicks. Regression models demonstrated how carefully controlled the flight characteristics must be to hit the target with launch angles constrained to within 3°. To achieve the curved ball trajectory, players should take a wide approach angle to the ball, point the support foot to the right of the intended target, swing the kicking limb across the face of the target, and impact the ball with the foot moving upwards and in an abducted position. In both kick types, peak knee angular velocity and ankle linear velocity occurred at ball impact providing biomechanical evidence to support the common coaching recommendation of kicking through the ball. These findings could assist coaches by focusing their attention to the fundamental coaching points necessary to achieve a curved trajectory of the ball when observing and correcting technique. iv

List of Publications I warrant that I have obtained, where necessary, permission from the copyright owners to use any third-party copyright material reproduced in the thesis (e.g. questionnaires, artwork, unpublished letters), or to use any of my own published work (e.g. journal articles) in which the copyright is held by another party (e.g. publisher, co-author).

Journal Publications Alcock, A, Hunter, A & Brown, N. (2009). Determination of football pitch locations from video footage and official pitch markings. Sports Biomechanics, 8, 129-140.

Alcock, A. (2010). Analysis of shots at goal direct from free kicks in the women's football World Cup 2007. European Journal of Sport Science, 10,279-284.

Alcock, A., Gilleard, W., Baker, J., Brown., N. A T., & Hunter, A, (in press). Initial ball flight characteristics of curve and instep kicks in elite women's football. Journal of Applied Biomechanics.

Alcock, A, Gilleard, W., Hunter, A, Baker, J.,&

Brown, N. A T., (submitted).

Curve and instep kick kinematics in elite female footballers. Manuscript submitted

for publication.

List of Conference Presentations Alcock, A, Hunter, A & Brown, N. A T. (2008). Analysis of the direct free kicks in

the 2007 Women's World Cup. Presented at The First World Conference on Science and Soccer. Liverpool, UK. May 2008.

Alcock, A. (2009). Analysis of free kicks taken directly at goal in elite women's

football. Presented at the Sports Medicine Australia ACT Branch Conference, Batemans Bay, Australia. January 2009.

Alcock, A, Brown, N., Baker, J., Gilleard, W. & Hunter, A (2009). Comparison of

straight and curve kick impact kinematics in elite female football players. v

Presented at the

7th

Australasian Biomechanics Conference, Gold Coast,

Australia. November 2009.

Alcock, A., Gilleard, W., Baker, J. & Brown, N. A. T. (2010). Flight characteristics of successful direct free kicks in elite women's football. Presented at the 2 nd World Conference on Science and Soccer, Port Elizabeth, South Africa. June 2010.

vi

Statement of contribution of others Chapter 3 - Determination of Football Pitch Locations from Video Footage and Official Pitch Markings.

Adam Hunter assisted with data collection and analysed the location of 18 balls on the pitch to determine the inter-reliability of the curve-fitting method which was developed. It was Nicholas Brown's idea to perform a sensitivity analysis on the data to determine how potential inaccuracies in pitch markings affected the reconstructed ball coordinates and he also provided advice on the writing phase of this chapter.

Chapter 6 - Initial ball flight characteristics of curve and instep kicks in elite women's football, and Chapter 7 - Comparison of curve and instep kick kinematics in elite female footballers

Wendy Gilleard, Nicholas Brown and John Baker provided advice on the research design and writing phase of these studies. The customised software used to calculate the ball centre was written by Adam Hunter.

vii

Acknowledgements I wish to acknowledge the Southern Cross University Division of Research, Southern Cross University School of Health and Human Sciences and the Australian Institute of Sport for their financial support in funding the projects undertaken to form this thesis, and for attendance at international conferences. In addition I would like to acknowledge the use of the University of Western Australia BodyBuilder model in the analysis of data.

To my supervisors, Associate Professor Wendy Gilleard, Professor Keith Lyons, Mr. John Baker, Dr. Nick Brown, and Professor John Hammond, thanks for the expert advice, words of wisdom and valuable time you have spent guiding me through this journey. You have both challenged me and encouraged me along the way, you have influenced the scientist in me in many ways and for that I will always be grateful.

Special thanks to Tom Sermanni and Robbie Hooker for their expert opinions on what contributes to performance of the different kick types in elite women's football, and to all the athletes who willingly gave up their time to participate. This thesis would not exist without your help and cooperation.

To the staff from the Australian Institute of Sport who selflessly gave up their time to assist with equipment set-up and data collection, and for sharing your knowledge. Thank you all for your time, whether it was a lot or a little, I couldn't have done it without you.

Finally, thanks to my family and friends for the love and support you have given me in everything I have ever decided to do.

viii

Table of Contents ANALYSIS OF DIRECT SHOTS AT GOAL FROM FREE KICKS IN ELITE WOMEN'S FOOTBALL. ............................ I DECLARATION .................................................................................................................................................... 11 SUPERVISOR'S DECLARATION ............................................................................................................................... III ABSTRACT ........................................................................................................................................................ IV liST OF PUBLICATIONS .........................................................................................................................................V STATEMENT OF CONTRIBUTION OF OTHERS ............................................................................................................ VII ACKNOWLEDGEMENTS ...................................................................................................................................... VIII TABLE OF CONTENTS .......................................................................................................................................... IX liST OF FIGURES ............................................................................................................................................... XII liST OF TABLES ................................................................................................................................................. XV ABBREVIATIONS AND NOMENCLATURE ................................................................................................................. XVI

1.

2.

INTRODUCTION ................................................................................................................................. 1

1.1.

SCOPEOFTHESTUDY ............................................................................................................................ 7

1.2.

SIGNIFICANCE OF THE STUDY .................................................................................................................. 8

1.3.

RESEARCH QUESTIONS ..........................................................................................................................

1.4.

RESEARCH AIMS ................................................................................................................................

9

10

LITERATURE REVIEW ........................................................................................................................ 13

2.1.

PERFORMANCE ANALYSIS ....................................................................................................................

13

2.1.1.

Development of performance analysis use in football ............................................................. 14

2.1.2.

Performance analysis research in football ............................................................................... 18

2.2.

ANALYSIS OF THE KICKING TECHNIQUE ...................................................................................................

23

Kinematics of kicking ............................................................................................................... 24

2.2.1. 2.2.1.1.

Kinematic variables relating to performance ...................................................................... 29

2.2.1.2.

Different types of kicks ........................................................................................................ 30

2.2.1.3.

Differences between male and female players ................................................................... 32 Mechanics of the foot to ball impact ....................................................................................... 35

2.2.2. 2.2.2.1.

Mechanics of the foot to ball impact in curve kicks ............................................................ 38

2.2.2.2.

Ball flight of curve kicks ...................................................................................................... 40

2.3.

METHODS IN BIOMECHANICAL STUDIES ..................................................................................................

43

2.3.1.

Data collection techniques ................................. ...................................................................... 43

2.3.2.

Shoes and surface .................................................................................................................... 44

2.3.3.

Consideration of entire body ................................................................................................... 45

2.3.4.

Task constraints ....................................................................................................................... 46

2.3.5.

Data processing and analysis .................................................................................................. 48

ix

2.4. 3.

CONCLUDING REMARKS ......................................................................................................................

51

DETERMINATION OF FOOTBALL PITCH LOCATIONS FROM VIDEO FOOTAGE AND OFFICIAL PITCH

MARKINGS ................................................................................................................................................ 53

4.

5.

3.1.

ABSTRACT ........................................................................................................................................

53

3.2.

INTRODUCTION .................................................................................................................................

54

3.3.

METHODS ........................................................................................................................................

56

3.3.1.

2D-DLTmethod ........................................................................................................................ 58

3.3.2.

Curve-fitting method ............................................................................................................... 58

3.3.3.

Reliability analysis ................................................................................................................... 61

3.3.4.

Sensitivity analysis ................................................................................................................... 62

3.4.

RESULTS ..........................................................................................................................................

62

3.5.

DISCUSSION .....................................................................................................................................

66

3.6.

CONCLUSION ....................................................................................................................................

70

ANALYSIS OF DIRECT FREE KICKS IN THE WOMEN'S FOOTBALL WORLD CUP 2007 ........................... 71

4.1.

ABSTRACT ........................................................................................................................................

71

4.2.

INTRODUCTION .................................................................................................................................

72

4.3.

METHODS ........................................................................................................................................

73

4.4.

RESULTS ..........................................................................................................................................

76

4.5.

DISCUSSION .....................................................................................................................................

79

4.6.

CONCLUSION ....................................................................................................................................

82

METHODOLOGY FOR BALL FLIGHT CHARACTERISTICS AND KINEMATIC ANALySiS ........................... 83

5.l.

INTRODUCTION .................................................................................................................................

83

5.2.

LABORATORY AND EQUIPMENT SET-UP ..................................................................................................

84

5.3.

PARTICIPANT INFORMATION ................................................................................................................

88

5.4.

OVERVIEW OF THE MOTION ANALYSIS MODEL..........................................................................................

89

5.4.1.

Motion analysis model marker set .......................................................................................... 91

5.5.

BALL MARKER SET .............................................................................................................................. 97

5.6.

DATA COLLECTION - KICKING TRIALS ......................................................................................................

97

5.7.

DATA ANALySiS .................................................................................................................................

98

5.7.1.

Filtering data through impacts ................................................................................................ 99

5.7.2.

Choice of filter .......................................................................................................................... 99

5.7.3.

Ball data treatment ............................................................................................................... 104

5.7.4.

Calculation of ball flight variables ......................................................................................... 106

5.7.5.

Shot accuracy ...................... ................................................................................................... 108

x

6.

INITIAL BALL FLIGHT CHARACTERISTICS OF CURVE AND INSTEP KICKS IN ELITE WOMEN'S FOOTBALL

109

7.

8.

6.1.

ABSTRACT ...................................................................................................................................... 109

6.2.

INTRODUCTION ............................................................................................................................... 110

6.3.

METHODS ...................................................................................................................................... 111

6.3.1.

Kicking trials ........................................................................................................................... 111

6.3.2.

Data analysis ......................................................................................................................... 113

6.4.

RESULTS ........................................................................................................................................ 116

6.5.

DISCUSSION ................................................................................................................................... 119

6.6.

CONCLUSION .................................................................................................................................. 123

CURVE AND INSTEP KICK KINEMATICS IN ELITE FEMALE FOOTBALLERS ......................................... 125 7.1.

ABSTRACT ...................................................................................................................................... 125

7.2.

INTRODUCTION ............................................................................................................................... 126

7.3.

METHODS ...................................................................................................................................... 127

7.4.

RESULTS ........................................................................................................................................ 130

7.5.

DISCUSSION ................................................................................................................................... 133

7.6.

CONCLUSION .................................................................................................................................. 136

CONCLUSIONS AND RECOMMENDATIONS ..................................................................................... 138 8.1.

CONCLUSiONS ................................................................................................................................. 143

8.2.

PRACTICAL IMPLICATIONS FOR THE GAME ............................................................................................. 144

8.3.

liMITATIONS OFTHE STUDY ............................................................................................................... 147

8.4.

RECOMMENDATIONS FOR FURTHER RESEARCH ...................................................................................... 149

REFERENCES .............................................................................................................................................. 151 APPENDIX A - VISUAL BASIC FOR APPLICATIONS CODE FOR CURVE-FiniNG METHOD ............................ 171 APPENDIX B - COMPARISON OF CURVE-FiniNG METHOD ACCURACY WITH 2-D DLT .............................. 177 APPENDIX C - PARTICIPANT INFORMATION SHEET ................................................................................... 182 APPENDIX 0 -INFORMED CONSENT FORMS (ADULT AND MINOR) ........................................................... 185 APPENDIX E - VISUAL BASIC FOR APPLICATIONS CODE FOR CALCULATING THE BALL CENTRE .................. 188

xi

List of Figures Figure 1-1: An example of a defensive wall (the four players standing at the edge of the penalty area) used to defend a direct free kick in the 2007 women's football World Cup tournament held in China .......... 2 Figure 2-1. An example of the camera placement for an 8-camera ProZone3" system, with the coloured lines representing the camera rays. The camera placements shown are those used at the stadiums of Manchester United and Bolton Wanderers. Figure from Oi Salvo, et aI., (2006) p111. ........................ 17 Figure 2-2. The relationship between the horizontal offset distance, spin ratio and ball velocity. Figure from Asai et al. (2002), p188 .......................................................................................................................... 39 Figure 3-1: Markers with a diameter of 0.21 m were placed on the pitch at each of the 99 known locations. Oistances shown are the official dimensions of the penalty area stipulated by FIFA and were confirmed on the test pitch used in this study. Two video cameras were located at approximately x = 23 m, y

=52.5 m, and 7 m above the pitch. The dashed line shows the large field of view in which 69

markers were visible and the black circles represent the 11 control points used for 20-0LT analysis. 57 Figure 3-2: The method for calculating the camera view using known pitch markings as control points. In this example, the two longest lines used to calculate the intersection point coordinates are line 1 (AA': goal line) and line 2 (00': front of penalty area). Pixel coordinates of other known pitch markings, B (front right corner of goal box) and C (penalty mark), are used to calculate the gradient from these features to the intersection point (m3 and m4)' The slope and magnitude of the line from these control points to the intersection point are entered into a least squares regression analysis to calibrate the field of view ............................................................................................................................................ 59 Figure 3-3: Error in metres of reconstructed pitch width coordinates from simulated television coverage. Larger circles reflect larger digitising errors. These measurements represent 27 different still frames from the simulated television coverage ................................................................................................ 64 Figure 3-4: Error in metres of reconstructed pitch length coordinates from simulated television coverage. Larger circles reflect larger digitising errors. These measurements represent 27 different still frames from the simulated television coverage. All markers were calculated within 0.35 m of their known location .................................................................................................................................................. 65 Figure 4-1: The position of the ball relative to the goal as it crossed (or would have crossed) the goal line for all direct free kicks that resulted in a goal being scored (or were saved) were classified into one of these six areas ofthe goal, as viewed from the penalty mark .............................................................. 75 Figure 4-2: Locations and outcomes of all free kicks taken directly at goal in the 2007 women's football World Cup. Those contained within a circle represent estimated distance of the kick from the touchline as the field of view had insufficient information to calculate the location ........................... 77 Figure 4-3: Schematic representation of the goal as viewed from the penalty mark to illustrate where shots crossed the goal line or were saved, estimated from video footage. All free kicks that resulted in a goal being scored or forced a difficult save were placed within approximately 1 m of the goalposts or

xii

crossbar. All easy saves were directed towards the bottom and centre of the goal, with the exception of one placed at the right ofthe goal which had a flight time of 2.32 s................................................ 78 Figure 5-1: A schematic of the laboratory set-up for the data collection procedures for a right-footed player. Not to scale ............................................................................................................................................ 86 Figure 5-2: View of the laboratory set-up in the frontal plane as viewed from the goal to illustrate the height and angle of the 17 camera positions .................................................................................................... 87 Figure 5-3: The laboratory set-up for data collection of a curve kick trial. The set-up was the same for instep kicks with the exception of the defensive wall. The red (1 m

2

)

and yellow squares show the target

area, and the four mannequins represent a defensive wall .................................................................. 87 Figure 5-4: The four mannequins and the position of the defensive wall used for the curve kicks. Note the position of the wall was such that 1.5 mannequins were outside the goalpost when viewed from the ball ......................................................................................................................................................... 88 Figure 5-5: Anterior view of the static calibration trial. .................................................................................. 94 Figure 5-6: Posterior view of the static calibration trial. ................................................................................. 95 Figure 5-7: A cluster was affixed to the lateral aspect of the football boot to allow for the first metatarsal marker to be considered as a virtual marker during kicking trials. This ensured that the contact between the boot and the ball was not affected by any external materials ......................................... 96 Figure 5-8: One of the FIFA approved balls used during the test procedures with two markers made from reflective tape visible. Markers were approximately 2 cm

2

97

..................................................................

Figure 5-9: Example of the residual analysis to determine the most appropriate MSE value for kicking hip flexion/extenion linear velocity in an instep kick ................................................................................ 100 Figure 5-10: Example graph of the kicking leg ankle resultant linear velocity for a curve kick. The point of ball impact is at frame 20 ........................................................................................................................... 101 Figure 5-11: An example of the effect of different cut-off frequencies on the resultant ankle linear velocity ofthe kicking leg. The point of ball impact is at frame 20 ................................................................... 102 Figure 5-12: Resultant hip joint centre linear velocity for a curve kick. Impact is at frame 50 ..................... 103 Figure 5-13: Fast Fourier transformation on the resultant linear toe velocity of a curve kick ...................... 104 Figure 5-14: Raw linear velocity data for the ball centre of an instep kick ................................................... 106 Figure 5-15: The elevation angle (8) was defined as the angle between the spin axis and the horizontal. .. 107 Figure 5-16: Illustration ofthe curve and instep kick accuracy ofthe shots used for analysis in this study. The 2

origin is the target centre and the solid black line represents the 1 m target ................................... 108 Figure 6-1: The vertical (y) and lateral (4)) launch angles were calculated from the ball centre linear velocity vectors. The ball elevation angle (8) is the angle between the spin axis and the horizontal, and the alpha angle (a) is the angle between the spin axis and the ball linear velocity .................................. 115 Figure 6-2: The initial resultant linear velocity, resultant angular velocity and elevation and alpha angles for a typical curve and instep kick for the first 15 frames of flight. Only the first ten frames for each trial were used for analysis as that was the number of frames available for the shortest trial and it was

xiii

important to use the same number of frames for each trial for accurate comparisons. A deceleration in linear ball velocity was evident even in the first ten frames of flight .............................................. 117 Figure 7-1: Comparison of kicking limb joint resultant linear and angular velocities for all curve and instep kicks normalised from support foot contact to ball impact. Curve kicks had a significantly greater knee angular velocity, and instep kicks had a significantly greater linear velocity of the hip and knee joint at ball impact ........................................................................................................................................... 132 Figure 7-2: Trajectories (thin black lines) of the kicking foot and the angle of the kicking leg plane relative to the horizontal target line for an elite female performing a) a curve kick, and b) an instep kick. The body orientation shows the impact position ....................................................................................... 133

xiv

List of Tables Table 1-1: Frequency of goals scored directly from free kicks in all women's football World Cup tournaments. Data from FIFA (2008) ....................................................................................................... 3 Table 3-1: Mean ± SO reconstruction error in metres resulting from the different methods and different views (maximum absolute errors are reported in parentheses) ........................................................... 63 Table 3-2: Mean ± SO reconstruction error in metres for reliability analyses (maximum absolute errors are reported in parentheses) ....................................................................................................................... 66 Table 4-1: Intra- and inter-observer reliability following repeat analysis of all kicks ...................................... 76 Table 5-1: Marker positions for three-dimensional segment definition during the kicking trials. For positioning of clusters refer to Figure 5-5 and Figure 5-6 ..................................................................... 92 Table 5-2: The additional markers attached to participants for the static calibration and the corresponding segment technical coordinate system in which these positions were stored. These markers were removed for the kicking trials ................................................................................................................ 93 Table 5-3: Values representing the appropriate cut-off frequency (average of six trials) of the Butterworth filter as determined by residual analysis ............................................................................................. 102 Table 6-1: The average range of the first ten data frames provided an indication of the data quality and the ball oscillation over the initial part of ball flight. Results are presented as mean ± standard error. ... 116 Table 6-2: Initial ball launch conditions (mean ± standard error) of the kicks that hit the target (curve: n = 39; instep: n

=33) ................................................................................................................................ 119

Table 7-1: Kinematic descriptors of curve and instep kicks performed by elite female right-footed football players. Results provided as mean ± SO .............................................................................................. 131 Table 8-1: Description of the instep kick technique performed by elite female football players and the coaching points required to modify the technique to achieve a curved ball trajectory ...................... 147

xv

Abbreviations and Nomenclature ANOVA

analysis of variance

CAST

Calibrated Anatomical Systems Technique

CM

centre of mass

cm

centimetres

DLT

direct linear transformation

FIFA

Federation Internationale de Football Association

GPS

global positioning system

Hz

Hertz

ISB

International Society of Biomechanics

kg

kilograms

m

metres

mm

millimetres

ms

milliseconds

m.s- 1

metres per second

MSE

mean square error

s

seconds

SD

standard deviation

rad.s- 1

radians per second

UWA

University of Western Australia

xvi

CHAPTER ONE 1. Introduction Football, otherwise known as soccer and often referred to as 'The World Game', is reportedly the most popular sport in the world (FIFA, 2010a; Lees & Nolan, 1998). There are 208 football associations affiliated to the world governing body (FIFA, 201 Ob), more than 200 million active players around the world (Price, Neilson, Harland, & Jones, 2004), and a cumulative audience of more than 26 billion viewers who watched the 64 matches played in the 2010 World Cup (FIFA, 2010a). High profile male professional football players are amongst the highest paid sportsmen in the world with David Beckham reportedly earning £29.7 million in 2008 comprising of £4.49 million in basic salary, £1.37 million in bonuses and a further £23.84 million in sponsorship deals (Blake, 2009). Whilst the women's game is still behind its male counterpart with regards to salaries, television rights and commercial deals, it has been reported to be the fastest growing women's sport in the world (Scott, 1999). There are now 105 national women's teams around the globe with an official FIFA ranking (FIFA, 2010c).

Football at the elite level is typically a low scoring game and thus how goals are scored and the identification of critical features of successful attacking strategies are important research issues. Previous research has revealed that approximately one third of all goals at the elite level are scored either directly or indirectly from a set play, irrespective of the tournament or player gender (Ensum, Williams, & Grant, 2000; FIFA, 2006, 2007; Grant & Williams, 1999; Grant, Williams, Reilly, & Borrie, 1998; Horn, Williams, & Grant, 2000; Yiannakos & Armatas, 2006). According to Carling and colleagues (Carling, Williams, & Reilly, 2005), there has been a recent increase in set play efficiency in men's domestic and international football (defined as more goals scored from fewer set plays), and successful teams are more efficient than their opponents at scoring from set plays. Therefore, preparation and planning of set plays from offensive and defensive points of view are important for winning games.

1

Of all the set plays, free kicks are consistently the most effective for scoring goals (Carling, et aI., 2005) and analysis of the men's European Championships in 2000 showed that direct shots at goal from free kicks in central areas of the pitch were more effective than a short pass followed by a shot at goal (Ensum, et aI., 2000). A direct free kick is typically defended by a wall of players, set up to preclude a straight shot at goal (Figure 1-1). However, for players who can strike a ball well with spin, this provides an opportunity to curve the ball around the wall and directly into the corner of the goal. A small number of male professional football players such as David Beckham, Roberto Carlos and Christiano Ronaldo are renowned for their expert performance of this skill and have changed the outcomes of many games based on one kick of the ball. Using a mathematical model of the ball's flight, Bray and Kerwin (2003) advocated that a well executed direct shot at goal from a free kick gives a goalkeeper little chance of saving a goal. The potential however for a direct free kick to be successful is largely dependent on the location from where it is taken, as this influences the distance the player must kick the ball, the positioning of the defensive wall of players and the angle to the goal.

Figure 1-1: An example of a defensive wall (the four players standing at the edge of the penalty area) used to defend a direct free kick in the 2007 women's football World Cup tournament held in China.

2

In the English Premier League, the locations on the pitch of free kicks used for direct shots at goal in the 1991-92 season encompassed areas extending to the halfway line and touch lines whereas in the 1997-98 season, teams crossed or passed the ball from wide areas, and attempts at goal were confined to a central zone just outside the penalty area (Williams, Lee, & Reilly, 1999). In the 1997-98 season, twice as many free kicks resulted in a goal being scored and it was suggested this could be due to an increased number of free kick specialists in the game (Williams, et aI., 1999). In women's football, statistics from previous women's World Cup tournaments show a progressive increase in the number of goals scored directly from free kicks (Table 1-1). However, characteristics of direct free kicks in the women's game remain neglected in the literature. Knowledge of the attributes of successful free kicks and the locations on the pitch where elite females are capable of scoring from would provide information on areas with the most goal scoring potential. Additionally it could facilitate decision making on where free kicks should be practised from in training and when a direct shot at goal should or should not be attempted in competition. For this information to be useful for coaches and players, it is important for it to be quantitative and objective to give an unbiased view of events on which informed tactical decisions can be made (Carling, et aI., 2005). Research has shown that the more quantitative and objective the feedback, the greater effect it has on performance (Franks, 1997). Table 1-1: Frequency of goals scored directly from free kicks in all women's football World Cup tournaments. Data from FIFA (2008). Women's

Number

Number

Number of goals

Number of goals direct

World Cup

of games

of goals

direct from a free

from a free kick as a

kick

percentage of all goals

Tournament 1991

26

99

1

1.01%

1995

26

99

Data not available

Data not available

1999

32

123

5

4.07%

2003

32

107

5

4.67%

2007

32

111

7

6.31%

Video analysis has been used extensively during sporting competition as a means of recording objective observations to collate statistical details on performance

3

(Reilly, 2001), particularly for the analysis of movements and the evaluation of technical and tactical aspects of play (Hughes, Hughes, & Behan, 2007). It provides an unobtrusive way of collecting competition specific information and providing objective feedback on performance. This information can then be used by coaches to provide athletes with a practise environment conducive to effective and efficient learning with the aim of improving future competition performance (Hughes & Bartlett, 2002). However, qualitative analysis of video footage is not always adequate for collecting specific performance data, and more intrusive analysis methods can provide an improved understanding of the mechanisms involved in performing a specific skill.

For example, although qualitative

information on the technique used to perform a free kick can be gathered from video, laboratory-based kinematic analyses are necessary to identify and quantify movement patterns related

to skilled

performance.

Understanding

kicking

biomechanics could assist coaches by focusing their attention on the key variables when correcting players' techniques on the field. Additionally, in a free kick it is not possible, from competition footage, to accurately determine initial velocity or rates and directions of ball spin, all of which are key components of the ball's trajectory in order to avoid the defensive wall and goalkeeper. However, with the sophisticated motion analysis systems currently available, laboratory-based tests can readily quantify these variables.

For scientific laboratory-based studies to provide useful information to coaches and athletes they must replicate the competition environment as closely as possible (George, Batterham, & Sullivan, 2003). Of the different skills used in football, kicking is the most extensively studied in the biomechanics literature (Lees & Nolan, 1998). However, with the open nature of the skills required in a game, along with limitations in data collection procedures, laboratory space and equipment, researchers have inevitably made assumptions and simplified the complexity of kicking skills. For example, the angle of the approach run has been constrained and the length has been limited to a specific number of steps (Barfield, Kirkendall, & Yu, 2002; Brown, Wilson, Mason, & Baker, 1993; Bull Andersen, Dorge, & Thomsen, 1999; Dorge, Bull Andersen, S0rensen, & Simonsen, 2002; Isokawa & Lees, 1988; Kawamoto, Miyagi, & Fukashiro, 2007;

4

Kellis, Katis, & Gissis, 2004; Kellis, Katis, & Vrabas, 2006; Lees & Davies, 1987; Manolopoulos, Papadopoulos, & Kellis, 2006; Rodano & Tavana, 1993; Teixeira, 1999). In addition, participants have not worn football specific boots or performed kicking tasks on standard surfaces (Kellis, et aI., 2004; Shan & Westerhoff, 2005) which could affect the kinematics of the kicking motion compared to playing on grass or artificial turf, both of which are currently used in elite competition (Andersson, Ekblom, & Krustrup, 2008; FIFA, 2009a). Such constraints also make it difficult to compare results across studies. Further research is required on kicking

techniques

with

minimal

constraints

to

develop

knowledge

and

understanding of realistic match scenarios.

Early kinematic studies on kicking used two-dimensional methods to analyse sagittal plane motions of the kicking limb, largely due to technological limitations as sophisticated motion analysis systems that enable rapid three-dimensional data to be collected have only become commonplace in the last decade (Davids, Lees, & Burwitz, 2000). Human motion is three-dimensional in nature and more recent

studies have shown the importance of rotations about both the medio-Iateral and longitudinal axes of the body as well as the contribution of the upper body in skilled participants (Lees & Nolan, 2002; Shan & Westerhoff, 2005). Therefore, it is acknowledged that for a thorough understanding of the mechanisms involved in expert kicking techniques, the entire body should be considered in threedimensions (Davids, et aI., 2000).

There are many different types of kick used in football, depending on ball speed, ball position and the intent of the kick, but the maximal instep kick of a stationary ball is by far the most widely reported in the literature (Lees & Nolan, 1998). An instep kick, where the ball is hit with the medial-superior portion of the boot (Levanon & Dapena, 1998), is kicked with a straight trajectory and is commonly used for generating a fast ball speed (Nunome, Asai, Ikegami, & Sakurai, 2002). It has been suggested that the maximal velocity instep kick corresponds to a penalty kick (Lees & Nolan, 1998), however research has shown that only 18% of penalties in elite male competition were taken at 100% effort and of these, approximately one third missed the goal completely (Hughes & Wells, 2002). A 5

trade-off between speed and accuracy is evident in many sporting actions (Engel horn, 1997) and kicking is no exception (Lees & Nolan, 2002; Teixeira, 1999). Thus for scientific studies to be task representative, the trade-off between accuracy and velocity must be considered.

Curve kicks are often used for a direct shot at goal from a free kick to swerve the ball around a defensive wall. The ball velocity and accuracy are vital components of a well executed direct free kick so that it reaches its intended target and gives the goalkeeper minimal time to react and save the ball. According to Newton's first law of motion, the flight characteristics of the ball are determined by the magnitude and direction of the forces applied to the ball by the kicking foot at impact, which in turn result from the kicking technique prior to impact. Thus, to achieve different trajectories, different impact characteristics between the foot and the ball are required (Asai, 2000). To achieve these different impact mechanics, the kicking technique prior to impact must differ, yet to the author's knowledge only one study has investigated the technique used to create a curved trajectory of the ball. Asai (2000) studied the kicking limb of amateur male players in curve and instep kicks and found differences in the position of the knee joint of the kicking leg and the approach angle to achieve the different impact points with the ball. A full-body kinematic analysis is required to further understand the mechanisms involved in applying spin to the ball.

The majority of biomechanical research in kicking has used male participants. However, the few comparative studies to explore gender differences have indicated that differences exist in kicking techniques between males and females (Orloff et aI., 2008; Tant, Browder, & Wilkerson, 1991) and females are generally not capable of achieving ball velocities as high as their male counterparts (Barfield, et aI., 2002; Orloff, et aI., 2008; Shan, 2009; Tant, et aI., 1991). Physiological studies have shown male footballers are taller, heavier, faster, stronger, and more powerful than female players (Bunc & Psotta, 2004; Davis & Brewer, 1992, 1993; Helgerud, Hoff, & Wisl0ff, 2002; Mujika, Santisteban, Impellizzeri, & Castagna, 2009; Reilly, Bangsbo, & Franks, 2000; Tumilty, 1993, 2000; Wisl0ff, Helgerud, & Hoff, 1998). Additionally, males have longer and

6

heavier limbs than females (De Leva, 1996), while females have a wider pelvis (Marieb, 2001). Given these inherent anatomical and physiological differences between genders, one would expect kinematic differences in the way males and females kick a ball. Therefore, findings from studies using male participants may not be applicable to females and more detailed research specific to the female population is required. Research on elite level females would allow for the development of a female specific model of technique that developing players could aspire to. The research in this thesis focused on understanding the technique used by elite female footballers to perform two different kick types (curve and instep) with a view to use that information, in conjunction with the coaches and athletes, to improve performance in these players. The differences in genders are not experimentally investigated in this thesis but the findings are compared with relevant results from male participants reported in the literature.

1.1. Scope of the study The research investigated the use and outcomes of direct free kicks in elite women's football during competition and the mechanisms involved in expert performance of the skill. First, a novel method was developed to objectively quantify specific locations on a football pitch using television footage and the marked lines which constitute the official pitch dimensions. The method was then applied to the video footage of all shots at goal taken directly from free kicks in the 2007 women's football World Cup tournament in order to determine the areas of the pitch with the most goal scoring potential and how flight time and the placement of the ball relative to the goal affect the outcome of the free kick. This information is beneficial for decision making on where direct shots from free kicks should be practised from in training and when a direct shot at goal should or should not be attempted in elite women's competition.

The competition specific data from the World Cup tournament were used in the set up of a laboratory-based task so that it replicated as closely as possible a realistic direct shot from a free kick with regards to the distance kicked, the position and size of the target to aim at and a defensive wall to curve the ball around. This 7

allowed for an investigation into the mechanisms that experts use to perform a direct shot at goal from a free kick that would likely score in elite women's competition. A sample of international female footballers from the 14th ranked team in the world performed simulated free kicks in the laboratory, and ball flight characteristics and full-body three-dimensional kinematics were analysed. The players also performed instep kicks at goal from the same location. As the final location of the ball relative to the goal mouth is determined by the ball flight, the magnitude of change in ball flight characteristics required to avoid the defensive wall and score a goal were investigated. Finally, the kinematic analysis identified the technique changes employed by elite females in order to achieve the different ball trajectories of the curve and the instep kicks. This information is important as it can reveal important coaching points, which in turn benefits coaches by focusing their attention to those key variables that relate to expert performance when observing and correcting technique.

1.2. Significance of the study The study investigated the

kinematics of international female footballers

performing curve kicks, and through a comparison with the kinematics of an instep kick, it enabled identification of differences in the mature kicking technique to achieve the different ball trajectories. This information will assist coaches when training players to perform a curve kick, and this is important because a direct shot at goal from a free kick, that typically uses a curve kick, provides a goal scoring opportunity that is more easily replicated in training compared with the more complex areas of open play. Analysis of the ball trajectory identified the initial flight characteristics utilised by elite females to both avoid a defensive wall and score a goal from a direct free kick in the elite women's game, and how these compare with the flight characteristics of a straight shot at goal from the same location. A number of methodological issues of previous laboratory-based research due to limitations in available data collection techniques, laboratory space and equipment were addressed. These include a full-body three-dimensional analysis, players wore their own football boots and performed the tasks on artificial turf, no restrictions were placed on the approach angle, a longer run-up was allowed

8

compared with previous research, and players aimed at a target with a level of accuracy indicative of that required in a match. Placing minimal constraints on the laboratory task is important in obtaining as true a representation as possible of how the athletes would perform the skill in a game.

A major benefit of this series of studies is that the laboratory-based task emulated a realistic match scenario that was grounded in scientific research. Observations from competition video footage were used to set up the replica free kick in the laboratory, enhancing the external validity of the laboratory task. The development of a method that could accurately and reliably locate the ball on a football pitch from

television

coverage

provided

a

cost-effective

method

of

collecting

quantitative, competition-specific information. Football is one of the most televised sports in the world and therefore information on elite teams is readily available. The application of the method to television coverage of an elite women's competition allowed for collation of previously unknown data on direct free kick locations that were successful in creating goals. By using the competition specific data to simulate a realistic task within the laboratory environment, and placing minimal constraints on the players performing the kicks, the results of the study provide a more accurate representation of what actually happens in a match. This is important if coaches are to use the information to train players to improve performance of the skill.

1.3.

Research questions

The aim of this thesis was to determine the attributes of a successful free kick in elite women's football and the mechanisms involved with successful performance of the skill. This information is potentially useful for designing training programs grounded in scientific research and identifying the fundamental coaching points for achieving a technique that produces the necessary ball flight characteristics from a direct free kick that would likely score in elite women's competition.

The specific objectives of this thesis were threefold. Firstly, the attributes of successful shots at goal direct from a free kick in an elite women's football

9

tournament were investigated with regards to their location, outcome and flight time. Secondly, the initial flight characteristics of successful direct free kicks were quantified and compared with instep kick flight characteristics to determine the ball launch conditions necessary to avoid the defensive wall and score a goal. Thirdly, full-body three-dimensional kinematics of elite female footballers performing direct free kicks were compared with instep kicks to determine the technical adaptations employed to achieve the different ball trajectories.

The research reported in this thesis focused on the following specific research questions: •

What are the attributes of a successful direct free kick in elite women's football with regards to the location on the pitch from which it is taken, its placement in the goal and its flight time?



What are the initial ball flight characteristics utilised by elite female football players in order to produce a successful direct free kick (curve kick), and how do these differ from a straight kick at goal from the same location?



How do the kicking techniques of elite female footballers differ when performing a kick resulting in a curved ball trajectory compared with a straight trajectory?

1.4.

Research aims

To address the aforementioned research questions, a combination of performance analysis and biomechanical studies was utilised. In Chapter 3 the methods for the performance analysis study are presented. Here, a novel method was developed to locate the position of a football on a football pitch and the validity and reliability associated with the method were quantified. In Chapter 4 the developed method was applied to elite competition video footage to determine the attributes of successful direct free kicks at the highest level of women's international football. The findings from the performance analysis study were used in the development of the methods used for the laboratory-based biomechanical studies described in Chapter 5. The differences in initial ball flight characteristics and player kinematics

10

of elite females performing curve and instep kicks are presented in Chapters 6 and 7 respectively.

The thesis is comprised of four studies, each with the following specific aims: Study one (Chapter 3):

Determination of football pitch locations from video footage and official pitch markings

to develop a method of calculating a specific location on a football pitch from digitised video simulating television coverage, especially areas outside the penalty area, and to compare the error of the developed method with the two-dimensional direct linear transformation method.

Study two (Chapter 4):

Analysis of direct free kicks in the women's football World Cup 2007

to quantify all pitch locations where players attempted to score directly from a free kick in the 2007 Women's World Cup, to determine the outcomes (scored, saved, hit the wall or off-target) of those free kicks, and to analyse the characteristics of ball flight (flight time and placement relative to the goal) for all those free kicks that resulted in a goal being scored or were saved.

Study three (Chapter 6):

Initial ball flight characteristics of curve and instep kicks in elite women's football

to compare the initial ball flight characteristics (linear velocity, angular velocity, launch angle, orientation of spin axis) of a direct free kick (curve kick) that would likely score in an elite women's football match with those of an instep kick at goal from the same location, and to determine the initial ball flight characteristics that best predicted the final placement of the ball relative to a target centre.

11

Study four (Chapter 7):

Curve and instep kick kinematics in elite female footballers

to compare full-body, three-dimensional kinematics of elite female football players performing an instep kick at goal and a curve kick that simulated a direct free kick in competition. Specifically, to compare the following variables: -

the approach run to the ball (angle, velocity and length of last stride), angular and linear velocities of the kicking limb joints during the kicking motion,

-

the trajectory of the kicking limb prior to ball impact, and

-

the orientation of the trunk, pelvis, support foot and kicking limb at ball impact.

12

CHAPTER TWO 2. Literature review Football is reputedly the most popular sport in the world, and often referred to as 'The World Game.' The game has been subject to extensive research. This chapter reviews the available literature in two main sections. The first investigates the use and development of performance analysis in football. Here, the term 'performance analysis' describes unobtrusive field-based methods of analysing player performance within their usual training and competition environment. The second section reviews laboratory-based technique analyses that collect more detailed data than those available through basic video analysis.

2.1.

Performance analysis

Feedback provided to athletes about their performance is one of the most important variables affecting learning and subsequent performance (Franks, 2004). The provision of feedback promotes efficient learning, ensures correct development of the skill and influences the athlete's motivation to persist with practice (Williams & Hodges, 2005). The coach must first observe and then evaluate performance to provide meaningful and timely feedback. Traditional coaching interventions have involved subjective observations and conclusions from the coach, however, a number of studies have shown that subjective observations are potentially unreliable and inaccurate (Maslovat & Franks, 2008). In addition, research has shown that the more quantitative and objective the feedback, the greater effect it has on performance (Franks, 1997). Advances in technology have allowed for improved feedback available to athletes during training and competition (Liebermann et aI., 2002).

Video analysis provides an unobtrusive method of collecting competition specific information and providing objective feedback on performance. The information gathered from video can be used to provide athletes with a practise environment conducive to effective and efficient learning with the aim of improving future 13

competition performance (Hughes & Bartlett, 2002). For example, training can be designed that is highly specific to the physical demands of competition and tactics and playing styles for future games determined based on what was previously successful or unsuccessful. In addition, the tactics and playing styles of upcoming opponents can be analysed.

2.1.1.

Development of performance analysis use in football

In elite football, there has been a growing acceptance of sports science as coaches and club managers realise its performance-enhancing role as they continually strive for a competitive edge against rival teams (Carling, Bloomfield, Nelsen, & Reilly, 2008). For example, a full time job as a performance analyst in football did not exist 15 years ago. Today, Manchester City, an English Premier League team, employs seven full time performance analysts working across all levels of the club from the first team to the academy, and each with a different focus including post match analysis, pre-match/opposition analysis, player recruitment and player development (Wilson, 2010).

The process of analysing competition has evolved with advances in technology, moving from early hand notation systems to include less subjective and more quantitative measures of performance (Hughes, et aI., 2007). Notation analysis evolved as a means of objectively recording specific instances in competition and collating accurate statistical information on performance (Hughes, 1988). The main objectives of any notation system are to collate information on player movements, and evaluate the technical and tactical aspects of play (Hughes, 1988).

One of the pioneering performance analysis papers in football estimated the distance covered by players by counting the number of strides taken for discrete activities (walking, jogging and sprinting) and converting this to distance based on the average stride length for each movement type, (Reilly & Thomas, 1976). The method was modified to calculate mean speeds from the time taken for players to pass pre-markers in the grass, the centre circle and other known distances as estimated visually from video (Krustrup & Bangsbo, 2001). Other systems have

14

simply counted the number of occurrences of particular events in real-time to collate statistics on performance (Worsfold & Macbeth, 2009).

Notation analysis methods are inexpensive and in general accurate (Hughes, 1991), but they do have a number of disadvantages. The systems involve considerable time to learn how to use them (Hughes, et aI., 2007), analysis is limited to one player at a time, and the methods are extremely labour-intensive in terms of data capture and analysis (Carling, et aI., 2008). Carling et aI., (2008) suggested that data collected from complex notation systems are generally restricted to academic research projects because the competitive schedule of elite teams require data to be available within 24-36 hours of the match completion. The reliability of notation analysis data can be compromised because of data entry errors

(Bradley,

O'Donoghue,

Wooster,

&

Tordoff,

2007),

and

different

interpretations of the same event from different observers (Carling, et aI., 2008). Post-match analysis has been found to be more reliable than real-time data entry due to having control over the video footage (Williams, Hughes, O'Donoghue, & Davies, 2007). In a study assessing the reliability of match statistics provided by four different television companies and an independent post match analysis, Worsfold & Macbeth (2009) found mean percentage errors between data sets up to 59% for shots on and off target, and mean percentage errors up to 9% for the less subjective notation of corners. Hughes and colleagues (Hughes, Cooper, & Nevill, 2002) found 70% of 72 reviewed notation analysis research papers did not report on the system reliability. However, if coaches are to use information from video analysis to provide detailed feedback to players and to make important tactical and technical decisions about how they will prepare for competition, it is critical to know the amount of error associated with the system and its reliability. For good reliability, definitions should be clear and unambiguous, the operator should be trained in observing appropriate visual information and using the system (Hughes & Franks, 1997), and the system should be user-friendly and intuitive.

The time-consuming nature of notation analysis led to the use of global positioning systems (GPS) to determine movement patterns and workloads. The GPS technology requires a receiver to be worn by each player, usually placed on the 15

upper back, in a pouch attached to a harness around the shoulders or in a sewn in pocket (Randers et aI., 2010). The receiver uses signals from Earth orbiting satellites to determine positional information and calculate movement speeds and distances (Randers, et aI., 2010). The main advantage of the GPS devices is that no manual analysis is required after the activity and all players can wear one at the same time, providing information on the whole team. However, the validity of these devices has been questioned, especially for categorising high speed movements such as sprinting (Coutts & Duffield, 2010; Duffield, Reid, Baker, & Spratford, 2010), and current football regulations prohibit the use of players carrying any electronic device during official competitive matches (FIFA, 2009b). Therefore the use of GPS information to monitor player movements and velocities in football is currently limited to training sessions and friendly games only if both teams agree to them being used.

Multiple player analysis in competitive games is now available through the development of computerised video systems. Many professional football teams use sophisticated analysis systems such as ProZone3® (ProZone Sports Ltd., Leeds, UK) or AmiscoPro® (Amisco UK, Birmingham, UK), which use 8-10 stationary cameras, permanently installed around the top of the stadium so that all areas of the pitch are seen by at least two cameras (Di Salvo et aI., 2007; O'Donoghue & Robinson, 2009) (Figure 2-1). Video images from the cameras are combined to produce a two-dimensional representation of the movements of all players, officials and the ball. Despite being largely computer automated, trained human operators are required to verify the tracking of players where the automatic tracking process may fail, for example when players move within close proximity of each other (O'Donoghue & Robinson, 2009). Typically the manual process is necessary during 42% of the match time (Di Salvo, Gregson, Atkinson, Tordoff, & Drust, 2009), and while a lengthy process, clubs receive their fully analysed data within 24 hours of the match completion (O'Donoghue & Robinson, 2009). The automatic tracking systems have been shown to be valid and reliable for tracking player movements (Di Salvo, Collins, McNeill, & Cardinale, 2006; O'Donoghue & Robinson, 2009). However, the permanent installation of the system into the stadium means that the information is not available for training sessions, reserve 16

or academy teams, or away games (unless the home team also has the system). Additionally, data are confidential to the teams involved meaning it is not useful as a scouting tool for upcoming opponents, and the cost of the systems (£100,000 for ProZone3® camera installation plus further costs for the analysis of each game, (Setterwall, 2003)) means they are available only to the richest football clubs.

-----------------------------,

~ ,'?

c""'c

",J'"

--

\

, .....- - - - 1 1)

""

\.......

,....-("

\

, \

~~~~--~~~------------~------------------------~~

Figure 2-1. An example of the camera placement for an 8-camera ProZone3® system, with the coloured lines representing the camera rays. The camera placements shown are those used at the stadiums of Manchester United and Bolton Wanderers. Figure from Oi Salvo, et aI., (2006) p111.

Given the above mentioned advantages and disadvantages of notation analysis, GPS devices and automatic tracking systems, even those professional clubs who do have access to the most sophisticated systems will use a combination of all these performance analysis methods to cover their analysis needs. Automatic tracking systems are limited to only the richest clubs, and the stadiums where the systems are installed. Therefore, post-match video analysis and methods based on visual estimation are still widely used for match analysis (Barros et aI., 2007). 17

The actual performance analysis system used is dependent on the information required and the systems available. Football is one of the most televised sports in the world, making video on elite teams readily available. Information that can be extracted straight from television coverage provides a cost-effective method of analysing elite teams and scouting opposition teams. A large number of scientific research studies have used manual video analysis of television coverage to collate information on elite teams (Armatas, Yiannakos, Galazoulas, & Hatzimanouil, 2007; Armatas, Yiannakos, Papadopoulou, & Galazoulas, 2007; Armatas, Yiannakos, & Sileloglou, 2007; Carey et aI., 2001; Grant, Reilly, Williams, & Borrie, 1998; Grant & Williams, 1998a, 1998b, 1999; Grant, Williams, et aI., 1998; Horn, et aI., 2000; Hughes & Wells, 2002; Jones, James, & Mellalieu, 2004; Luhtanen, Belinskij, Hayrinen, & Vanttinen, 2001; O'Donoghue, 2002; Scoulding, James, & Taylor, 2004; Yiannakos & Armatas, 2006). Women's football does not have the commercial, spectator or financial support that the professional male clubs benefit from, and therefore few women's games are played in the large stadiums that have automatic tracking systems installed. In their review paper, Carling and colleagues (Carling, et aI., 2008) reported that no data on the women's game were available from automatic tracking systems. Hewitt and colleagues (Hewitt, Withers,

& Lyons, 2007) reported data on distances travelled by the Australian women's team using GPS devices, however the prohibition of using the technology in competition means manual video is often the only analysis method available for the women's game.

In summary, performance analysis is now accepted as a way of providing valuable feedback that can be used by football coaches and athletes to enhance performance both technically and tactically. The actual method of analysis used depends on the specific information required, any competition regulations, the cost of any analysis systems, and the availability of televised coverage of the game.

2.1.2.

Performance analysis research in football

Football is a dynamic game with 22 players continuously interacting, moving at different speeds, performing different actions in different areas of the pitch and 18

continually creating and transforming spaces. It is therefore difficult to represent the complexity of the game even with a detailed analysis system and researchers have simplified the game accordingly. Seabra & Dantas (2006) stated that while notating certain events from the complex, dynamic and situational whole that is a football game, and grouping them into categories contributes to the understanding of tactical, technical and strategic aspects of the game, the reduction in the complexity of the data gathering process also leads to some loss in the situational significance of the actions.

In an attempt to identify key aspects of successful performance, analyses of successful teams have been performed and in some cases compared to unsuccessful teams in both men's (Grant & Williams, 1999; Grant, Williams, et aI., 1998; Horn, et aI., 2000) and women's football (Konstadinidou & Tsigilis, 2005). Results from these studies indicate that successful strategies for one team are not necessarily successful for another.

It has been suggested that to derive the most benefit from the analysis of football, it is necessary to move beyond the mere description of behaviours and progress towards prediction of performance with a view to identifying common patterns of behaviour (Grehaigne, Mahut, & Fernandez, 2001). However, in a longitudinal study of an English Premier League team over an entire season, playing patterns of both the teams and individual players were found to differ between games (James, Mellalieu, & Hollely, 2002). This could be due to the complexity of the analysis not reflecting the complexity of the game or because elite players have a number of different responses to certain situations which they alternate to confuse the opposition (James, 2006). The playing style of a football team is not only dependent on the interaction of the players on the pitch at any given time, but also weather conditions, standard of the opposition, tactical strategies of the manager, whether the team is winning or losing at the time, whether the game is a league or knock-out game, and which 11 squad members make up the team at anyone time since the starting line-up regularly changes between games and substitutions are made within games. This makes the already dynamic game of football even more

19

complex and the role of using performance analysis to predict performance even harder.

Due to the complexity of the game, many researchers have developed simplified systems that consider only parts of the game. As the object of the game is to score goals and elite football is characterised by a low frequency of scoring, most research has focused on methods of goal scoring, including the build up to the goal, from which areas on the pitch most goals are scored from, and what kind of shots are most successful (Grant, Reilly, et aI., 1998; Grant & Williams, 1999; Grant, Williams, et aI., 1998; Yiannakos & Armatas, 2006). Results of these studies indicate that set plays (free kicks, corners, penalties and throw-ins) consistently play an important role in the creation of goals at the highest level of competition.

In elite football, approximately one third of all goals are scored either directly or indirectly from a set play, irrespective of the tournament or player gender (Ensum, et aI., 2000; FIFA, 2006, 2007; Grant & Williams, 1999; Grant, Williams, et aI., 1998; Horn, et aI., 2000; Yiannakos & Armatas, 2006). Ensum and colleagues (Ensum, et aI., 2000) suggested that whilst set plays in the attacking third occur infrequently in international football, they account for a significant proportion of goals scored, and the most important factors to consider when executing set plays are the quality of execution, the organisation and positioning of players, variety and surprise. In men's international and domestic football, successful teams are more efficient than their opponents at scoring from set plays; that is, successful teams score more goals from fewer set plays (Carling, et aI., 2005). Therefore, preparation and planning of set plays from both an offensive and defensive point of view is important for winning games. Given the large number of goals scored from set plays, several researchers have focused only on key indicators of successful set plays (Ensum, et aI., 2000; Grant, Reilly, et aI., 1998; Williams, et al.,1999).

Of all set plays, free kicks are consistently the most effective for scoring goals (Carling, et aI., 2005). In the men's European Championships in 2000, an average 20

of ten free kicks were awarded in the attacking third of the pitch per match, and a goal was scored either directly or indirectly from free kicks every other game (Ensum, et aI., 2000). Given the importance of free kicks in creating goal scoring opportunities, it is not uncommon for teams to have at least one free kick specialist, and Bray & Kerwin (2003) suggested that spectators' expectations of success from a direct free kick is approaching that of a penalty kick. Williams and colleagues (Williams, et aI., 1999) reported that free kicks taken in the English Premier League in the 1997-98 season were more likely to result in a shot on goal and twice as many resulted in goal scoring compared with the 1991-92 season, possibly due to an increased number of free kick specialists in the game.

The potential for a direct free kick to be successful is largely dependent on the location from where it is taken, as this influences the distance the player must kick the ball, the positioning of any defensive wall of players and the angle to the goal. As the free kick location approaches the touch line, the wider angle to the goal reduces the margin for error because the area of the goal that can be seen from the ball reduces, which in turn increases the demand for accuracy. Williams and colleagues (Williams, et aI., 1999) found the locations on the pitch of free kicks used for direct shots at goal in the 1991-92 English Premier League season encompassed areas extending to the halfway line and touchlines whereas in the 1997-98 season, teams crossed or passed the ball from wide areas and attempts at goal were confined to a central zone just outside the penalty area. Therefore, the increased efficiency observed in the 1997-98 season may be because players became aware of the ineffectiveness of areas far away from goal and wide areas in scoring goals and, through knowledge of their individual skill limits, only took a direct shot at goal when they perceived the probability of scoring a goal to be high. For wide areas, it is possible that players perceived their chance of scoring to be lower, and crossed or passed the ball to build an attack rather than providing the opposition with an easy turnover of possession.

Analysis of the men's European Championships in 2000 found that direct shots from central areas were more effective than a short pass followed by a shot at goal (Ensum, et aI., 2000). Using a mathematical model of ball flight, Bray & Kerwin 21

(2003) advocated that a well-executed direct shot from a free kick gives a goalkeeper little chance of saving a goal. However currently no research has analysed the characteristics of ball flight in competition, namely the placement of the shot relative to the goal and its flight time, and how these relate to the outcome of the free kick. Previous research on penalty kicks in male football has shown the placement of the ball relative to the goal is related to the shot outcome, with those placed at the top of the goal and within one yard (0.91 m) of the goalpost more likely to score than those in the bottom and centre of the goal (Hughes & Wells, 2002; L6pez-Botella & Palao, 2007; Morya, Bigatao, Lees, & Ranvaud, 2004). The pace of the shot is also important as it determines how long the goalkeeper has to react and save the ball (Kerwin & Bray, 2006), but if the penalty taker kicks it maximally there is more chance of missing the goal completely (Hughes & Wells, 2002).

Despite women's football being one of the fastest growing sports in the world (Scott, 1999), to the best of the author's knowledge, there has been no published research on the use and efficacy of free kicks in the women's game. The notion of an increased number of free kick specialists in football may also hold true in the women's game with statistics from previous Women's World Cup tournaments showing a progressive increase in the number of goals scored directly from free kicks (Table 1-1). In the first Women's World Cup competition in 1991 only one goal was scored directly from a free kick in the 26 games played (FIFA, 1991), while in 2007 there were seven goals scored directly from free kicks in the 32 games played (1 per 4.6 games) (FIFA, 2007).

The statistics from major men's and women's football tournaments highlight a direct shot at goal from a free kick as a skill that is not only efficient in creating goals, but also a relatively closed skill that is easy to simulate in training compared to open play. Further research is required on the flight characteristics and the placement of the ball relative to the goal for direct free kicks that score a goal. In addition, an investigation into the locations on the pitch used for direct free kicks by elite females in competition is required because the locations that females are capable of scoring from may differ from the male game given inherent anatomical 22

differences (see Section 2.2.1.3). Such an analysis would be beneficial in defining areas with the most goal scoring potential and determining attributes of successful free kicks. This information could then facilitate tactical decisions on when a direct shot at goal from a free kick should or should not be attempted in competition and on the specific areas where free kicks should be practised from in training.

2.2. Analysis of the kicking technique Technique analysis has been widely used in the sporting domain to improve technique and performance, and to determine how movements are made, the most effective way movements are made, and the effect of the movements on performance (Lees, 2002). Kicking techniques of skilled athletes have been used (Barfield, et aI., 2002), and sometimes compared to novice athletes (Shan & Westerhoff, 2005), to develop a model of optimal technique and identify key factors related to expert performance. While it is acknowledged that a better performance does not necessarily indicate a better technique, generally a better technique will lead to improved performance (Lees, 2002). Therefore analysis of athletes considered to be the best in the world at performing a certain skill can provide information conducive to effective performance of a skill, which in turn can be used by lower level performers to improve their technique, and ultimately their performance.

Following qualitative analyses of kicking skills, a description of the main movement characteristics common to the mature form of the kicking technique was developed (Wickstrom, 1983). The mature kicking technique is characterised by a step on to the support leg to rotate the pelvis backward on the kicking side and at the same time the kicking limb extends at the hip and flexes at the knee. Forward motion is initiated by the pelvis rotating about the support limb and the kicking limb swinging forwards with simultaneous flexion of the hip and knee. Next, the thigh begins to decelerate and the shank is vigorously extended just before the foot contacts the ball. Finally the arm on the non-kicking side swings forward in reaction to the vigorous action of the kicking limb (Wickstrom, 1983). The description of the mature kicking technique is applicable across kicking sports and 23

across kick types. For different types of kick, task specific modifications are necessary to achieve the required difference in ball trajectory, but the basic movement pattern remains. Technique analysis has been used to identify underlying principles of movement. Lees (2007) identified five principles of movement associated with kicking: range of motion, stretch-shorten cycle, end-point speed, action and reaction, and proximal-to-distal sequencing. The advantage of representing a kick using these qualitative terms is that it allows coaches and practitioners to evaluate performance qualitatively on the field without having to refer to detailed biomechanical data (Lees, Andersen, Nunome, & Sterzing, 2010).

Qualitative analysis can be used to identify the fundamental aspects of the kicking technique, such as an angled approach run and the tilt of the trunk and pelvis, however a quantitative analysis can provide a more detailed understanding of these mechanisms and information on the actual angle of the run and the amount of trunk and pelvis tilt (Davids, et aI., 2000). Such a quantitative analysis requires full-body

three

dimensional

analysis

using

sophisticated

biomechanical

instrumentation (Davids, et aI., 2000). The remainder of this section (2.2) and section 2.3 review the biomechanical analyses of football kicking relevant to the research undertaken in this thesis.

2.2.1.

Kinematics of kicking

This section (2.2.1) describes the kinematics of kicking in general. The modifications to the general kicking kinematics to achieve different types of kicks are reviewed in Section 2.2.1.2.

The approach to the ball is important in creating horizontal momentum that can be used during the kick, as well as positioning the body to allow for optimal execution of a successful kicking motion (Wang & Wiese-Bjornstal, 1994). The length, angle, and velocity of the approach run to the ball all influence the shot outcome. Opavsky (1988) found that greater ball speeds were achieved (30.8 m.s- 1) when a six to eight stride run-up was used compared to a stationary approach (23.5 m.s- 1). 24

In a study of elite men's football competition, players took an average of five steps in the approach for penalty kicks (Hughes & Wells, 2002). With regards to approach velocity, Bo Kristensen & Bull Andersen (2009) found maximal ball velocities were achieved at self-selected approach velocities, which equated to 73.53 ± 4.7% of their maximal running velocity. An approach velocity above or below their self-selected value resulted in a lower ball velocity (Bo Kristensen & Bull Andersen, 2009). It appears there is an optimal approach velocity for achieving a high ball velocity.

In a study investigating the approach angle on foot and ball velocities, six male players used a one-step approach to perform a maximal instep kick at approach angles of 0, 15, 30, 45, 60 and 90° (the direction of the kick was 0°) (Isokawa & Lees, 1988). Peak ankle and ball velocities were achieved with approach angles of 30° and 45° respectively, leading the authors to suggest the optimum angle for an instep kick with one step is around 45°, a figure which is in agreement with what players choose to do (Egan, Verheul, & Savelsbergh, 2007). Lees and colleagues (Lees, Steward, Rahnama, & Barton, 2009) stated that angled approaches are used for both maximal and submaximal kicks and therefore any benefits may not be directly related to ball speed. Rather, the angled approach enables the body to tilt away from the ball, which raises the hip of the kicking limb allowing the foot to clear the ground and contact the ball in an optimal position, and the knee to extend through impact (Lees, et aI., 2009).

The length of the last stride is important in kicking (Lees, et aI., 2010). Lees & Nolan (2002) reported a larger last stride length for a maximal kick (0.77 m) compared with a submaximal kick (0.54 m). These values appear surprisingly small, however the method of calculating the stride length was not documented. Stoner & Ben-Sira (1981) reported last stride lengths of 1.69 m for a long-range kick and 1.5 m for a medium range kick. In Australian rules football, last stride lengths of 1.74 m were reported for players kicking for distance (Ball, 2008). The length of the last stride is larger than the preceding strides, and in conjunction with the angled approach run, this orientates the body in such a way that the pelvis can rotate through a greater range of motion to ball contact (Lees & Davids, 2002), 25

which in turn increases the acceleration path of the foot towards the ball, (Wang & Wiese-Bjornstal, 1994). In addition, the retraction of the pelvis on the kicking limb side enables the hip-shoulder separation angle to increase (the angle between a line connecting the two shoulder joint centres and a line connecting the two hip joint centres), and allows the musculature in the upper body to contribute to the kick by rotating the pelvis, (Lees & Nolan, 2002).

The medio-Iateral placement of the support foot is important because if it is too far from the ball, the direction of the kick and the kicker's balance will be compromised (Barfield, 1998). An optimal position of 5-10 cm to the left side of the ball has been suggested for a right-footed player (Hay, 1993). The anteriorposterior positioning of the support foot depends on the objective of the kick and whether it is intended to have a low or high trajectory (Barfield, 1998). McLean & Tumilty (1993) reported foot placements in junior players to be 37 cm to the side of the ball and 8 cm behind the ball for a drive kick, while Orloff and colleagues (Orloff, et aI., 2008) found male and female collegiate players placed the support foot heel 31 and 32 cm to the side of the ball respectively. The placement of the support foot in elite players has received little scientific interest in the research literature (Lees, et aI., 2010) and it is currently unknown if the foot placement changes with different kick types.

Wang & Wiese-Bjornstal (1994) stated that for the optimum application of force and control at ball contact, the non-kicking leg should be placed alongside the ball with the knee slightly flexed. The support leg knee remains flexed throughout the kicking motion to provide dynamic stability and balance which are important because the body weight is dependent on this one foot throughout the kicking motion (Lees, et aI., 2009). Additionally, the relatively large flexion observed (42 0 at ball impact) allows for a small extension of the support leg knee just before ball impact, indicating that it contributes to performance by lifting the body and contributing to the vertical velocity of the foot at impact (Lees, et aI., 2009).

The role of the upper body is less documented than the lower body yet research has shown it is a major contributor to the kicking motion in skilled athletes.

26

Following the large final stride and the placement of the non-kicking foot, Shan & Westerhoff (2005) identified a 'tension arc' formed by trunk rotation towards the non-kick side, abduction of the arm on the non-kicking side, and hyperextension of the kicking limb hip. Levanon & Dapena (1998) reported hip extension values of up to 29° in this backswing phase, with 0° corresponding with the anatomical position. The formation of the tension arc stretches the quadriceps, abdominal and pectoral muscles and enables a stretch-shorten cycle to take place across the entire body. The tension arc is released with upper body flexion and trunk rotation towards the ball, and simultaneous movement of the kick limb towards the ball (Shan & Westerhoff, 2005).

The motion of the kicking limb is characterised by segmental and joint rotations in multiple planes (Kellis & Katis, 2007). Following the formation of the tension arc, forward motion is initiated by the pelvis rotating about the support limb and the kicking limb swinging forwards. The pelvis and kicking hip rotate almost simultaneously (Browder, Tant, & Wilkerson, 1991; Tant, et aI., 1991) suggesting little transfer of energy from the pelvis to the thigh. Rather, the pelvis rotation acts to increase the axis of rotation of the foot and the range of motion through which the lower limb travels, which provides the hip muscles with more time to generate force, thus resulting in increased foot speed at ball impact (DeWitt, 2002). Mean ranges of motion of the pelvis in the transverse plane for skilled male players performing instep kicks are reported between 30.4° and 36° (Lees & Nolan, 2002; Levanon & Dapena, 1998). As the thigh moves forward, the shank moves backwards until maximum knee flexion is achieved (Kellis & Katis, 2007). The flexion of the knee as the thigh is brought forward serves two important purposes (Lees, et aI., 2009). First, it stretches the quadriceps muscles before contracting, thus producing a stretch-shorten cycle that allows for a more forceful contraction. Second, the knee flexion reduces the moment of inertia of the leg allowing for an increased hip angular velocity (Lees, et aI., 2009).

As the kicking limb continues to move towards the ball, both the thigh and shank move forward (Kellis & Katis, 2007). The angular velocity of the thigh continues to increase, reaching its peak value just before the knee starts to extend, at which 27

point the thigh and shank angular velocities are equal and thus, knee joint angular velocity is zero (Kellis & Katis, 2007). As the shank accelerates forward just before impact with the ball, the thigh slows down due to a momentum shift and exchange of energy between the two body segments (Lees & Nolan, 1998). This proximal-todistal sequential motion whereby the angular velocity of the proximal segment decreases and the distal segment increases is important for proficient execution of kicking skills (Putnam, 1991). At ball impact, the angular velocity of the thigh is almost zero, while the shank reaches its peak angular velocity, resulting in a high foot speed (Lees & Nolan, 1998). A large foot velocity is important for maximal kicks because it is strongly related to resultant ball velocity (Asami & Nolte, 1983; Kawamoto, et aI., 2007; Levanon & Dapena, 1998; Nunome, Lake, Georgakis, & Stergioulas, 2006; Rodano & Tavana, 1993).

For instep kicks, the hip is slightly flexed (22°), abducted and externally rotated (Levanon & Dapena, 1998) at ball impact. The knee of the kicking leg is widely reported to be flexed at impact, which has been attributed to a protective mechanism as the joint is prevented from reaching full extension at maximum velocity (Lees, et aI., 2009). Lees and colleagues (Lees, et aI., 2009) found the knee to be flexed

5r at ball impact and to extend a further 12° during the impact,

and suggested that this appeared to be a large margin of safety and there may be other reasons for the large knee flexion. First, because the support leg knee is flexed to 42° at impact, and the kicking foot is plantarflexed at ball impact, knee flexion of the kicking leg is required for the foot to clear the ground. Second, the knee has a greater range of internal/external rotation when it is flexed which could be useful in orientating the foot in the optimal position for impact (Lees, et aI., 2009). During impact, the ankle rapidly plantarflexes to its extreme range of motion, a movement which is assumed to be completely passive as a result of the contact with the ball, (Nunome, Lake, et aI., 2006). Additionally, the foot has been found to abduct (15.3°) and evert (2.9°) during the contact with the ball (Shinkai, Nunome, Ikegami, & Isokawa, 2009).

Skilled players use a large follow-through after impact (Shan, 2009) which keeps the foot in contact with the ball for as long as possible so that greater momentum 28

can be imparted on it (Baliield, 1998). In addition, the follow-through acts to ensure the forces necessary to decelerate the leg do not inteliere with the projection of the ball into the air whilst also providing a protective mechanism to the body as the forces generated by the musculature during loading and the forward swing to the ball are dissipated during the follow-through, (Hay, 1993).

2.2.1.1. Kinematic variables relating to performance In a maximal kick, ball speed is a measure of success. Ball velocities up to 34.6 m.s- 1 have been reported for maximal instep kicks peliormed by male participants (Nunome, lkegami, Kozakai, Apriantono, & Sano, 2006) and 21.5 m.s- 1 for female participants (Baliield, et aI., 2002). Several researchers have investigated the relationship between kinematic variables and resultant ball speed through comparisons of skilled and novice players (Shan, 2009; Shan & Westerhoff, 2005), dominant and non-dominant kicking limbs (D6rge, et aI., 2002), different kick types (Browder, et aI., 1991) and across a speed-accuracy paradigm (Lees & Nolan, 2002).

D6rge and colleagues (D6rge, et aI., 2002) found that players achieved a better quality impact between the foot and the ball with their dominant limb compared with their non-dominant limb which resulted in a higher ball speed, as more energy was transferred to the ball. At impact, increased linear and angular velocities of the kicking hip, knee, ankle and foot are associated with increased ball velocities (Asami & Nolte, 1983; DeWitt & Hinrichs, 2002; Kawamoto, et aI., 2007; Levanon

& Dapena, 1998; Nunome, Lake, et aI., 2006; Rodano & Tavana, 1993). An increased foot speed, and therefore ball speed, can be achieved using a greater range of motion at the joints. Greater ball velocities have been associated with increased ranges of motion at the shoulder, trunk, pelvis, hip, knee, and ankle joints and an increased hip-shoulder separation angle (Browder, et aI., 1991; Lees & Nolan, 2002; Shan, 2009; Shan & Westerhoff, 2005). Shan & Westerhoff (2005)

found the distance between the non-kick side shoulder and the kick side hip during kicking was highly related to the effectiveness of the kick, and the authors 29

suggested this variable can be used to evaluate a player's skill level since the range of this distance was very small for novices. Essentially this is a measure of the quality of the 'tension arc' formed, as discussed in section 2.2.1. The importance of joint range of motion in generating a fast ball speed indicates that flexibility could playa role in enhancing kicking performance.

Caution should be taken when interpreting the findings of studies that relate kinematic variables to ball velocity because speed may not be the best indicator of success. Accuracy is also an important factor with regards to a successful kick reaching the desired target or player. Comparisons between skilled and novice athletes may be a better way of identifying parameters that contribute to expert performance and for understanding the learning of the skill.

2.2.1.2. Different types of kicks In football, different types of kick are used depending on the required distance, speed, and intent of the kick (Lees & Nolan, 1998). The flight characteristics of a football are determined by the magnitude and direction of the forces applied to the ball by the kicking foot at impact, which in turn are dependent on technique variables such as the foot velocity and orientation at ball impact, and the foot trajectory prior to ball impact. Therefore, it follows that different types of kicks require different kicking techniques.

The maximal instep kick is by far the most widely reported in the biomechanics literature (Lees & Nolan, 1998). The kinematics of other kick types have been investigated, but to a much lesser extent. Browder and colleagues (Browder, et aI., 1991) found a reduction in pelvis rotation for high drive kicks compared with low drive and maximum distance kicks. Increased hip flexion and extension and knee flexion and extension were also observed during the high drive kick (Browder, et aI., 1991). Prassas and colleagues (Prassas, Terauds, & Nathan, 1990) attributed the different ball launch angles of high and low trajectory kicks to the difference in ball impact, which in turn was attributed to a significantly elevated hip of the kicking limb and a more perpendicular shank segment of the support leg.

30

In football many shots and passes are played with sidespin to produce a curved trajectory (Neilson, Jones, Kerr, & Sumpter, 2004). For example, this is useful for playing the ball around an opposition player, or swerving a shot away from the goalkeeper. In a direct free kick, a defensive wall is set up to prevent a straight shot at goal but for players able to strike a ball well with spin, this provides an opportunity to swerve the ball over or around the wall and into the goal. A number of male free kick specialists such as David Beckham, Roberto Carlos and Cristiano Ronaldo are renowned for scoring spectacular goals from their expert performance of this skill. Despite the prevalence of direct free kicks in creating spectacular goals, and a direct shot at goal from a free kick being a relatively closed skill and therefore easier to analyse compared with the more complex aspects of open play, to the author's knowledge only one study has investigated the technique used to produce a curved trajectory of the ball. Asai (2000) compared curve and instep kick kinematics and found that in order to achieve the different impact points on the ball, amateur male players used a wider approach angle and a more dorsiflexed ankle in the curve kick. In addition, the impact point for the curve kick caused the position of the knee joint in the sagittal plane to be slightly behind that of the instep kicks and the upper body to lean back (Asai, 2000).

In analyses of the pass kick, a kick similar to a curve kick with regards to the medial aspect of the foot and ankle providing the impact surface with the ball, the ball velocity has been found to be slower than that of an instep kick (Levanon & Dapena, 1998; Nunome, et aI., 2002). The reduced ball velocity was attributed to the slower foot speed at impact, which in turn was attributed to the pelvis and kicking limb being more externally rotated, representing a leg configuration whereby the shank-foot segment was almost perpendicular to the direction the ball was kicked, (Levanon & Dapena, 1998; Nunome, et aI., 2002). Lees & Nolan (1998) suggested that the external rotation of the leg is necessary to facilitate ball impact with the medial aspect of the foot, but prevents the knee from moving in the same way it does for an instep kick, which in turn inhibits the speed the foot can attain at ball impact.

31

Curve kicks are characterised by increased lateral and vertical release angles compared with instep kicks (Carre, Asai, Akatsuka, & Haake, 2002), resulting from differences in the direction of the forces applied to the ball by the kicking foot (Asai, Carre, Akatsuka, & Haake, 2002). This indicates that the trajectory of the kicking foot prior to ball impact must differ between kick types. The plane of the kicking limb has not been documented in the literature, however the concept of a 'swing plane' has been reported in other sports such as golf (Coleman & Anderson, 2007), and field hockey (Willmott & Dapena, 2008) to describe the trajectory of the club or stick prior to impact with the ball. Coleman & Anderson (2007) found different swing planes were used for different types of golf clubs, revealing important coaching implications because different clubs may require different swing coaching points. In football, it is possible that the kicking limb may use different swing planes for different kick types such as straight and curve kicks; however, no scientific research has investigated this.

Given the inherent anatomical and physiological differences between genders, generally males can kick a football faster than females (Barfield, et aI., 2002) and technique differences between genders for kicking have been observed (Shan, 2009; Tant, et aI., 1991). Although there has been a recent increase in the number of goals scored direct from free kicks in the women's game, scientific research on females performing curve kicks remains neglected in the literature.

2.2.1.3. Differences between male and female players Studies on the physical characteristics of football players have shown that male athletes are taller, heavier, faster, stronger, more powerful and have a larger aerobic capacity compared with females (Bunc & Psotta, 2004; Davis & Brewer, 1992, 1993; Helgerud, et aI., 2002; Mohr, Ellingsgaard, Andersson, Bangsbo, & Krustrup, 2004; Mujika, et aI., 2009; Reilly, et aI., 2000; Tumilty, 1993, 2000; Wisl0ff, et aI., 1998). These physiological differences likely affect the intensity at which the game is played and the execution of individual skills requiring strength and power, such as kicking. 32

Gender comparisons have shown that generally females produce lower ball velocities than their male counterparts in maximal instep kicks (Barfield, et aI., 2002; Orloff, et aI., 2008; Shan, 2009; Tant, et aI., 1991). Barfield and colleagues (Barfield, et aI., 2002) found that elite males demonstrated greater maximum toe velocity, ball contact toe velocity, mean toe velocity, mean toe acceleration and ankle velocity at contact than elite females. Females had a higher angular velocity of the knee at impact, and it was suggested this may be indicative of a protective mechanism by the males in order to prevent knee hyperextension following a powerful kick. Shan (2009) found that skilled male players followed through with a jump to dissipate leg momentum after a powerful kick, whereas skilled females counteract the movement with flexion of the upper-body. In addition, skilled male players displayed greater pre-lengthening of the pectoral and abdominal muscles, thus creating a better quality tension arc than the females, and allowing greater muscle power to be generated in the male kick, resulting in the observed higher ball velocities (Shan, 2009).

A high linear velocity of the kicking foot at impact is advantageous because it is strongly related to the resultant ball velocity (Asami & Nolte, 1983; Levanon & Dapena, 1998; Nunome, Ikegami, et aI., 2006). The kicking motion is essentially a series of rotational movements and consequently, through segmental interactions, the length of the limb segments will influence the end-point velocity of the foot at impact (Luhtanen, 1994). The limb segments of males are longer and heavier than females (De Leva, 1996), which could be beneficial for male players in obtaining greater foot and ball velocities and applying greater force to the ball at impact (Luhtanen, 1994).

Anatomically, females have a wider pelvis than males (Marieb, 2001). Browder and colleagues (Browder, et aI., 1991) suggested that females took advantage of their increased pelvic width to produce maximum ball velocities during kicks. Pelvis transverse rotation has been identified as an important aspect of the segmental interactions involved in the kicking motion (Browder, et aI., 1991; Lees

& Nolan, 2002; Levanon & Dapena, 1998; Tant, et aI., 1991). Therefore a wider 33

pelvis could facilitate an increased relative end point velocity of the segment as it rotates towards the ball (Smith, 2007).

A relationship between muscle strength and kicking performance would be expected in football because the muscles are directly responsible for increasing the speed of the foot, (Lees & Nolan, 1998). Indeed, kick performance (defined as the distance of the kick) has been found to correlate significantly with hip extensor, hip flexor, knee extensor, and knee flexor strength (Cabri, De Proft, Dufour, & Clarys, 1988). Other researchers have found improvements in ball velocities following strength training programs (De Proft, Cabri, Dufour, & Clarys, 1988; Dutta & Subramanium, 2002;

Manolopoulos, et aI., 2006; Manolopoulos,

Papadopoulos, Salonikidis, Katartzi, & Poluha, 2004; Ta"iana, Grehaigne, & Cometti, 1993; Trolle, Aagaard, Simonsen, Bangsbo, & Klaysen, 1993). Given the superior strength of males compared to females it would be expected that males are able to generate a more forceful and powerful kicking motion than females. Tant and colleagues (Tant, et aI., 1991) attributed the greater ball velocities produced by male players, compared to female players, to their greater leg strength as measured by an isokinetic dynamometer.

Women's football is one of the fastest growing sports in the world, yet the game has received relatively little scientific interest compared to the men's game. To date, descriptions of the female kicking technique in the biomechanics literature are limited to the maximal instep kick of a stationary ball (Barfield, et aI., 2002; Browder, et aI., 1991; Orloff, et aI., 2008; Shan, Daniels, Wang, Wutzke, & Lemire, 2005; Tant, et aI., 1991). The few comparative studies to explore gender differences have indicated that differences exist in kicking techniques between males and females (Orloff, et aI., 2008; Tant, et aI., 1991) and females are generally not capable of achieving ball velocities as high as their male counterparts (Barfield, et aI., 2002; Orloff, et aI., 2008; Shan, 2009; Tant, et aI., 1991). Given the inherent anatomical and physiological differences between genders, and their potential influence on the kicking motion, findings from studies using male participants may not be applicable to females and therefore more detailed research specific to the female population is required. 34

2.2.2.

Mechanics of the foot to ball impact

The quality of the impact between the foot and the ball is important because ball velocity is dependent on the transfer of energy from the body to the ball (Hennig & Sterzing, 2010). The success of the impact depends on the mechanical characteristics of the foot and ball during contact, including the effective striking mass of the foot, stiffness of the foot, mass and pressure of the ball, which part of the foot contacts the ball and where the ball is hit (Lees & Nolan, 1998). Hennig & Sterzing (2010) stated that players with an effective foot/leg striking mass of 3 kg and a foot velocity of 22 m.s- 1 would achieve a ball velocity of 139 m.s- 1 if there was a full momentum transfer and no energy loss. In reality however, the impact is not purely elastic (Tsaousidis & Zatsiorsky, 1996), and top players achieve maximum ball velocities around 35 m.s- 1 (Neilson & Jones, 2004).

Analysis of the foot-ball interaction in toe kicks performed by two amateur male players showed that a foot velocity of 19.6 m.s- 1 at initial ball impact drops to 13.4 m.s- 1 at the point of peak ball deformation (Tsaousidis & Zatsiorsky, 1996). This point is important because until then, the foot velocity was higher than the ball velocity, but after the point of peak ball deformation, ball velocity continued to increase whereas the foot velocity remained fairly constant with a value of 14.1 m.s- 1 at the end of ball contact.

The mechanics of the collision between the ball and the foot can be considered with the following equation (Lees & Nolan, 1998): vball =v foot

where

Vbal/

and

Vfoot

[M] ·[1 +e] [M + m]

is the velocity of the ball and foot respectively, M is the

effective striking mass of the striking object (in this case the shank-foot segment and is affected by muscle activation making the limb more rigid), m is the mass of the ball and e is the coefficient of restitution.

The coefficient of restitution was defined as (Bull Andersen, et aI., 1999): 35

e . (Vf,before where

Vf,before

and

respectively, and

vf,after

Vbal/

Vball,before)

= -( Vf,after -

Vball)

is the velocity of the foot before and after impact

is the velocity of the ball.

The coefficient of restitution (e) relates to the quality of the impact, and depends on the mechanical properties of the boot, ball, ankle and foot upon impact (Bull Andersen, et aI., 1999). A perfect elastic collision would give e = 1 and values in football have been reported between 0.463 and 0.681 (Bull Andersen, et aI., 1999; D6rge, et aI., 2002). Bull Andersen and colleagues (Bull Andersen, et aI., 1999) suggested that increasing the coefficient of restitution from 0.5 to 0.65 would increase ball velocity by 10%.

In an instep kick, the foot contacts the ball around the metatarsal-phalangeal joint and the large impact force passively plantarflexes the foot, which continues until the ankle joint reaches its extreme range of motion and at this point the foot will deform at the metatarsal-phalangeal joint (Lees & Nolan, 1998). There is little to prevent this deformation, which affects the firmness of impact and subsequently the coefficient of restitution, (Lees & Nolan, 1998). Asai and colleagues (Asai, et aI., 2002) demonstrated that the change in horizontal velocity during impact was higher at the end of the foot than at the ankle, indicating greater decelerations, forces and deformations. For powerful ball kicking where speed is the priority, it has been suggested that deformation should be minimised by contacting the ball closer to the ankle joint than the toes (Asami & Nolte, 1983), a view supported by Asai et aI., (2002) and Kellis & Katis., (2007).

Sterzing and

colleagues

(Sterzing,

Kroiher, & Hennig, 2009) found that

participants who could disregard the pain of barefoot kicking produced ball velocities 1.6% higher than when kicking with their own football boot. This remarkable finding was attributed to the football boot limiting the plantarflexion angle at initial ball contact, leading to more 'give' during the collision phase of the kick and thus further forced plantarflexion during the impact. In contrast, the ankle joint in the barefoot condition was fully plantarflexed at ball contact, providing a more rigid impact surface and therefore superior impact mechanics. Pictures taken 36

from high-speed video support this reason for the superior ball velocities in barefoot kicking, with the ankle moving through a greater range of plantarflexion during impact in the shod kicking trials (Sterzing, et aI., 2009).

A ratio of ball-to-foot velocity provides a measure of the effectiveness of the transfer of foot velocity to ball velocity and has been reported to range from 1.06 to 1.65 (Asami & Nolte, 1983; Isokawa & Lees, 1988; Kawamoto, et aI., 2007; Kellis, et aI., 2004; Kellis, et aI., 2006; Lees & Nolan, 2002; Nunome, et aI., 2002; Nunome, Lake, et aI., 2006). Kawamoto and colleagues (Kawamoto, et aI., 2007) found the ratio of foot to ball velocity to be significantly higher in experienced players than inexperienced players, indicating a poor transfer of momentum from the kicking foot to the ball due to deformation at the ankle in the inexperienced group. The foot velocity immediately before ball impact is considered a strong determinant of initial ball velocity, a notion supported by the strong correlations found between foot and ball velocities (r > 0.74) (Asami & Nolte, 1983; Kawamoto, et aI., 2007; Levanon & Dapena, 1998; Nunome, Lake, et aI., 2006; Rodano & Tavana, 1993).

In a study comparing full-instep and pass kicks, the ratio between ball velocity and the velocity of the foot was similar in both kicks (1.33 for the full kick and 1.23 for the pass kick) (Levanon & Dapena, 1998). There was a strong relationship between foot and ball velocity (r = 0.83) which led the authors to conclude that the velocity of the foot prior to the impact was the main determinant of ball velocity after impact and the slower ball velocity observed in the pass kick were almost exclusively due to the slower foot velocity at impact. Nunome and colleagues (Nunome, et aI., 2002) supported these findings with similar lower ball and foot velocities in a side-foot kick when compared with an instep kick.

An investigation into the effects of fatigue on instep kicking found a significant correlation between ball and toe velocities in the non-fatigued condition but a weak correlation after fatigue was induced (Apriantono, Nunome, Ikegami, & Sano, 2006). Lees & Davies (1987) found foot velocity was higher in a fatigued state compared with a non-fatigued state, but ball velocity was lower which they 37

attributed to the fatigued condition providing a less rigid foot surface at ball contact, making the energy transfer from the foot to the ball less effective. Reduced effectiveness of impact mechanics in a fatigued state were supported by Kellis et aI., (2006). Together these findings indicate the importance of the impact mechanics between the foot and the ball in generating ball velocity.

Caution is required when interpreting results concerning foot velocities as the part of the foot used to calculate the velocity is not uniform and in some studies it is often unclear what part of the foot was used (Lees & Nolan, 1998). The study by Asai and colleagues (Asai, et aI., 2002) showed that the toes had a higher velocity at impact than the centre of the foot which in turn was travelling faster than the ankle joint.

2.2.2.1. Mechanics of the foot to ball impact in curve kicks The flight of a football is dependent on the forces applied to it by the kicking foot at impact. Therefore different trajectories require different impact mechanics. The impact mechanics between the foot and the ball when a curved ball trajectory is produced has been investigated across a number of studies. Asai and colleagues (Asai & Akatsuka, 1998; Asai, et aI., 2002) used computer simulation to investigate the effects of the horizontal offset distance (Le. the distance between the location of the foot-ball impact point and the ball centre) on ball spin and ball velocity. The relationship between the horizontal offset distance, spin ratio and ball velocity indicated a trade off between the development of ball spin and velocity (Figure 2-2). That is, as the distance from the point of force application and the ball centre increased, ball spin increased, but ball velocity decreased. For an offset distance that exceeded the ball radius, and therefore only the very edge of the foot contacted the ball, both the spin and velocity decreased as the energy of the impact failed to be transferred to the ball (Asai, et aI., 2002). Asia et aI., (2002) also found that ball spin increased with an increased coefficient of friction between the foot and the ball, yet rotation of the ball still occurred even if the kinetic coefficient of friction was equal to or almost equal to zero. This was attributed to

38

the large deformation of the ball during impact which causes rotation because of the impact force. The offset distance affected ball spin much more than the coefficient of friction (Asai, et aI., 2002).

-e- Spin ratio .;::' 30 I

-0- Ball velocity

(J)

-

E 25

~ 20

'u

.2 15 (I) >

-

10

ctI

.0

"0 I:: t'iS

5

-....

,..

It/) ~

0

-5

. 2-10 t'iS

~-15+---~--~-------'-------'-------'

I::

"E.. -0.16 rJ)

-0.08

o

0.08

0.16

Off-set distance (m)

Figure 2-2. The relationship between the horizontal offset distance, spin ratio and ball velocity. Figure from Asai et al. (2002), p188.

In a later study, Asai and colleagues (Asai, Takano, Carre, & Haake, 2004) used computer simulation to model the angled approach of the foot to the ball and investigate the influence of an 'attacking angle' on ball spin and velocity for curve kicks. The attacking angle was defined as the angle made between the face vector (the normal vector to the line between the heel and toe) and the swing vector (velocity vector of the foot). Results showed that ball velocity decreased as attacking angle increased, while spin rate increased with the attacking angle, but started to decrease at attacking angles greater than 55°. Lees and colleagues (Lees, et aI., 2010) suggested this was due to the foot slipping across the ball. The findings of the study by Asai and colleagues (Asai, et aI., 2004) indicate the importance of the foot orientation at impact in generating ball spin in a curve kick. 39

2.2.2.2. Ball flight of curve kicks The flight trajectory of a ball is influenced by its initial linear velocity, launch angle, spin rate, spin axis orientation and the air density (Kreighbaum & Hunt, 1978). In football many shots and passes are played with spin to produce a curved trajectory (Neilson, et aI., 2004). The curved trajectory results from an imbalance of pressure distribution around the spinning ball causing it to move in the direction of the lowest pressure region as a result of the Magnus effect (Passmore, Tuplin, Spencer, & Jones, 2008). In a direct free kick, typically the defending team sets up a wall of players to deter a straight shot at goal, but for players able to strike a ball with spin, this constraint can be overcome by swerving the ball over or around the wall and into the goal. A well executed free kick gives a goalkeeper little chance of saving a goal, yet despite the prevalence of free kicks in creating goals at an elite level, relatively few studies have investigated the initial launch conditions of the ball that produce these goal-scoring techniques (Bray & Kerwin, 2003).

Theoretical

and

experimental

comparisons

have shown

curve

kicks

are

characterised by a reduced ball velocity, increased spin rate, more vertical spin axis orientation, greater launch angle and an increased flight time compared with instep kicks (Carre, et aI., 2002; Whiteside, Alderson, & Elliott, 2010). For male footballers, instep kick ball velocities have been reported with values up to 34.6 m.s- 1 (Manolopoulos, et aI., 2006; Nunome, Ikegami, et aI., 2006; Nunome, Lake, et aI., 2006). For curve kicks that simulate a direct free kick, resultant velocities ranging from 15.1 to 28.3 m.s- 1 and spin rates between 4.0 and 9.4 revs.s- 1 are reported in the literature (Bray & Kerwin, 2003; Griffiths, Evans, & Griffiths, 2005; Whiteside, et aI., 2010). The difference in linear ball velocity between kick types can be attributed to the trade off between the development of ball spin rate and ball velocity (Asai, et aI., 2002). Neilson & Jones (2004) reported that professional male footballers are capable of producing spin rates up to 14 revs.s- 1 . However, increased spin may not be beneficial because of the ensuing reduction in velocity. Therefore players should aim to create no more spin than is necessary to swerve

40

the ball around the defensive wall, thus ensuring the ball reaches the goal as quickly as possible, giving the goalkeeper minimal time to move and save the ball.

Bray & Kerwin (2003) used a mathematical model to investigate the flight of a football in a direct free kick and determine the constraints the defensive wall places on the kicker when attempting a direct shot at goal. The results revealed how precise a player must be to avoid the defensive wall and score a goal. For a central free kick 18.3 m from the goal, a defensive wall covering the far post, and an initial ball velocity of 25 m.s- 1 , the ball's initial launch angle must be constrained between 16.5° and 17.5° and struck with almost perfect sidespin. If a player is able to strike the ball precisely however, little can be done by the goalkeeper to save it (Bray & Kerwin, 2003).

Carre and colleagues (Carre, et aI., 2002) investigated the effects of spin on the flight of a football by taking controlled measurements of balls launched with varying amounts of spin at different velocities and, in conjunction with their previous data (Asai, et aI., 2002), used the results to simulate typical game situations. Trajectory simulations demonstrated the differences in flight of two kicks taken 18 m from goal and entering the goal in the top right-hand corner. The first was a straight kick with a ball velocity of 26 m.s- 1 , and no spin. The second was struck with the same foot velocity, but the impact position of the foot was 0.08 m to the right of the ball centre, reducing the ball velocity to 18.5 m .S-1, and imparting a spin of 10.2 revs.s- 1 about the vertical axis. The second kick curved as a result of the spin, and because the ball was travelling 30% slower than the straight kick, it had to be launched at a steeper angle in order for them both to land at the same point. The effects of the change in trajectory and the increased drag in the curved kick extended the flight time from 0.9 s for the straight kick to 1.6 s for the curved kick. These simulations demonstrate the practical implication of the trade-off between ball spin and velocity, and the important role played by the defensive wall of players in deviating a direct shot from a free kick, thus providing the goalkeeper with an increased time to move and save the ball.

41

While mathematical simulations used to model ball flight are beneficial in improving our understanding of the precision required to curve a direct free kick successfully around a defensive wall and into the goal, the flight of a football is not necessarily uniform. A football does not have a completely smooth surface and the orientation of the valve on the ball affects ball flight because it shifts the centre of mass by approximately 2.5 mm (Griffiths, et aI., 2005). Research has shown that significantly higher launch angles are achieved when the valve is positioned at the top of the ball rather than the bottom prior to impact (Neilson, et aI., 2004). Griffiths and colleagues (Griffiths, et aI., 2005) used a motion analysis system to track the trajectory and spin of a curve kick kicked with initial ball velocities between 15 and

18 m.s-1. A number of trials showed there can be large variations in spin rate and spin axis orientation angles which affect the ball's trajectory in unpredictable ways, implying an unstable ball flight. Griffiths et aI., (2005) suggested further research was needed on a variety of kickers to allow for inter-participant comparisons of spin rates and stabilities of the spin axis.

The mathematical models of Bray & Kerwin (2003) and Carre et aI., (2002) assumed pure sidespin was imparted to the ball, however in reality that is not the case. Whiteside and colleagues (Whiteside, et aI., 2010) investigated the initial flight characteristics of curve kicks in semi-professional male footballers and found the spin axis orientation was 62.6° to the horizontal (pure sidespin would be 90°). Despite the recent increase in the number of goals scored directly from free kicks in elite women's football, initial flight characteristics of curve kicks by females remain neglected in the literature. Given that gender comparative studies of instep kicks indicate that females are generally not capable of achieving ball velocities as high as their male counterparts (Barfield, et aI., 2002; Shan, 2009; Tant, et aI., 1991), and the trade-off between ball velocity and spin development, it is likely that other flight characteristics would differ between genders also. Research into the ball launch conditions of females performing curve kicks is required.

42

2.3.

Methods in biomechanical studies

In football, individual skills such as kicking are influenced by the free flow of the game, making a game context difficult to simulate in a laboratory setting. In addition, sophisticated data collection techniques are often only possible in a laboratory environment. Therefore, in order to collect quantitative data on the kicking technique, inevitably researchers have simplified the complexity of kicking skills, with almost all published research focusing on the kick of a stationary ball.

Biomechanical analyses on kicking in football have been conducted for many decades and our understanding of the skill has developed with advances in technology and data collection and processing techniques. Researchers have made compromises due to available equipment or laboratory space. For instance, early researchers were limited to two-dimensional analysis methods whereas today three-dimensional analyses are becoming the norm. Other researchers have modified aspects of the kicking motion to suit the laboratory environment. For example, the approach to the ball has been restricted because of available space and participants have worn non-football designed shoes to suit the testing surface. The influence of these restrictions on the data collected are discussed below.

2.3.1.

Data collection techniques

Early kinematic studies of kicking reported data collected in two-dimensions in the sagittal plane. Advances in the technology of the measurement systems have allowed for three-dimensional analyses of motion (Barfield, et aI., 2002; Browder, et aI., 1991; Brown, et aI., 1993; Clagg, Warnock, & Thomas, 2009; Lees, Kershaw, & Moura, 2004; Lees & Nolan, 2002; Levanon & Dapena, 1998; Nunome, et aI., 2002; Prassas, et aI., 1990; Rodano & Tavana, 1993; Scurr & Hall, 2009; Shan, 2009; Shan, et aI., 2005; Shan & Westerhoff, 2005; Tant, et aI., 1991; Teixeira, 1999). Rodano & Tavana (1993) conducted a three-dimensional analysis of professional football players performing an instep kick and analysed the data in both two- and three-dimensions to determine the differences between the two measurement techniques. While good agreement was found for linear velocity of the joints, large discrepancies up to 84% error were found in angular velocities 43

(Rodano & Tavana, 1993). It is still unknown whether the data calculated twodimensionally were over- or underestimations compared with those calculated three-dimensionally. Oorge and colleagues (Oorge, et aI., 2002) collected twodimensional data in the sagittal plane of a maximal instep kick and restricted participants to a straight run-up from a position directly behind the ball in order to minimise movements in the frontal plane. They acknowledged that although an approach angle of 30-45° improved resultant ball velocity as found by Isokawa & Lees (1988), an angled approach would introduce errors to their two-dimensional analysis. Such an approach to the problem however means the kicking motion analysed does not represent the skill in competition and therefore any practical application to the game may be limited. Considering the findings of previous threedimensional studies where rotation of the pelvis and trunk have been found to contribute significantly to the technical execution of the kicking motion, (Browder, et aI., 1991; Lees & Nolan, 2002; Levanon & Oapena, 1998; Shan, 2009; Shan & Westerhoff, 2005; Tant, et aI., 1991), it is now widely acknowledged that the kinematics of kicking a football can only be truly defined using three-dimensional analysis.

2.3.2.

Shoes and surface

The shoes and surface used for laboratory-based kicking tasks are an important consideration as they affect both the foot-ball impact mechanics and the shoesurface interaction, which in turn can affect the ball trajectory and biomechanical data collected. Only one kinematics study reviewed required participants to wear football boots and perform kicks on artificial turf (Rodano & Tavana, 1993). The surface used for laboratory-based kicking tasks could affect the kicking motion compared to playing on grass or artificial surfaces. For example, it is likely that landing on an extremely hard or soft surface would affect the kinematics of the support limb. Shan and colleagues (Shan, 2009; Shan, et aI., 2005; Shan & Westerhoff, 2005) used a 2 cm thick wrestling mat "to mimic the effects of grass," however they did not report on any validation of this statement or how closely the properties of the two surfaces related.

44

In previous laboratory-based studies, often the sutiace used is limited to the laboratory floor sutiace, or the sutiace of the force plates if ground reaction forces are being measured. The participants in the studies by Kellis and colleagues (Kellis, et aI., 2004) and Clagg and colleagues (Clagg, et aI., 2009) landed on the metallic sutiace of a force plate. While this allows for inter- and intra-athlete comparisons under the same testing conditions, it does not provide an understanding of forces

realistic to those

experienced

in

a competitive

environment. Furthermore, it is difficult to compare ground reaction forces across studies because the different shoes and sutiaces affect the measured forces.

The sutiace used dictates the type of footwear that can be used because studded football boots typically used in competition are not suitable for indoor sutiaces. Footwear is important because specially designed football boots influence both the support leg by providing adequate traction, and the foot-ball interaction of the kicking leg (Sterzing & Hennig, 2008). Maximum ball velocity (Hennig & Zulbeck, 1999) and even kicking accuracy have been shown to be influenced by the shoe design (Hennig & Sterzing, 2010).

Studies using standard running shoes may not accurately represent a kick taken with football boots. With the type of footwear and its properties shown to affect shot accuracy, the transfer of the foot velocity to the ball and the resultant ball velocity, caution should be taken when comparing results for these variables across studies using different footwear.

2.3.3.

Consideration of entire body

The majority of biomechanical studies on kicking have focused only on the movements of the lower body, in particular the function of the kicking limb (Apriantono, et aI., 2006; Asai, 2000; Batiield, et aI., 2002; Browder, et aI., 1991; Bull Andersen, et aI., 1999; D6rge, et aI., 2002; Isokawa & Lees, 1988; Levanon & Dapena, 1998; Nunome, et aI., 2002; Prassas, et aI., 1990; Rodano & Tavana, 1993; Tant, et aI., 1991; Teixeira, 1999). Developments in technology and computing power, and the ensuing development of motion analysis systems that 45

can track the body segments in three-dimensional space, have alleviated the need for manual digitising. These sophisticated systems reduce data processing time and make it easier to analyse a greater number of body segments. It is currently more common for researchers to include the role of the upper-body in their analysis (Brown, et aI., 1993; Lees, et aI., 2004; Lees & Nolan, 2002; Orloff, et aI., 2008) and a few have used a full-body model (Shan, 2009; Shan, et aI., 2005; Shan & Westerhoff, 2005). The findings of these studies highlight the role of the musculature of the upper body and upper appendages in creating a better condition for generating a more powerful kick. Without consideration of the upper body in the kick, an understanding of the entire kicking motion and joint coordination remains incomplete.

2.3.4.

Task constraints

Although the multi-factorial nature of the kicking skill and limitations of current data collection techniques mean laboratory-based methods are required to gather accurate biomechanical data, few studies have taken a systematic approach in maintaining as much ecological validity to a game situation as possible. For a full understanding of the kicking motion and for biomechanists to be able to provide information that could improve performance, it is imperative that kicks analysed in laboratory situations are representative of match conditions.

Several studies have not used any accuracy constraints (Barfield, et aI., 2002; Dbrge, et aI., 2002; Isokawa & Lees, 1988; Lees & Davies, 1987; Shan & Westerhoff, 2005) and subjects were instructed to kick the ball as hard as they could. In a game, it is unlikely that players would ever kick the ball as hard as they possibly could with no regard for where it is going. In football there is a requirement for both accuracy, to deliver the ball to the intended target or player, and ball velocity to surprise opponents and give them less time to react. Some studies have used an unguarded football goal for the participants to aim at (Isokawa & Lees, 1988; Kellis, et aI., 2004; Lees, et aI., 2004; Levanon & Dapena, 1998; Nunome, et aI., 2002; Nunome, Ikegami, Asai, & Sato, 2000), providing a much bigger target than there would be in a game situation when players would

46

aim at a small area of the goal in an attempt to beat the goalkeeper. It has been claimed the maximal instep kick corresponds with a penalty kick (Lees & Nolan, 1998), however in reality penalties require a high level of accuracy. Penalty shots placed towards the top of the goal and within one yard (0.91 m) of the goalpost are more likely to result in a goal than those placed towards the bottom and centre of the goal (Hughes & Wells, 2002; L6pez-Botella& Palao, 2007; Morya, et aI., 2004). While maximal instep kick kinematics add to our understanding of how different variables affect ball velocity which is an important performance outcome of any kick, more studies with accuracy constraints are required for a full understanding of situations that are realistic to competition.

Conversely to the maximal kick where the emphasis has been on velocity, studies on sub-maximal kicks have instructed participants to perform the kick towards the target, with no regard for velocity, (Asai, 2000; Levanon & Dapena, 1998; Nunome, et aI., 2002). Marcolin and colleagues (Marcolin, Petrone, & Robazza, 2006) required participants to kick a ball at different areas of the goal from 16 metres away, with and without a goalkeeper. It was found that ball velocities were higher in the series of shots with the goalkeeper. This indicates that studies on sub-maximal kicks may not accurately reflect actions in a game either as the presence of the goalkeeper affected the velocity of the kick.

A trade-off between speed and accuracy is well documented in the literature for sporting movements, including kicking in football (Lees & Nolan, 2002; Teixeira, 1999). The results of these studies indicate that when a player is instructed to perform a fast instep kick with a degree of accuracy, the approach velocity, along with joint rotations, velocities and ranges of motion are lower than in a powerful kick, and movement times are longer. Such observations were attributed to higher velocities increasing the variability of movement and reducing times for corrections of movement through feedback processing

mechanisms

(Teixeira,

1999).

Therefore, reducing the velocity of a movement appears to be a natural strategy to produce ballistic movements within the limits of precision imposed by constraints on accuracy (Teixeira, 1999). For penalty shots in elite male competition, it has been found that as velocity increased, there were less saves but more misses, 47

indicating that if players can improve their accuracy at pace then they will score more goals (Hughes & Wells, 2002).

The approach to the ball has been constrained in many laboratory tests. For example, participants have been restricted to a one-step approach (Clagg, et aL, 2009; Isokawa & Lees, 1988; Kawamoto, et aL, 2007; Kellis, et aL, 2004; Teixeira, 1999), a two-step approach (Barfield, et aI., 2002; Brown, et aL, 1993; Kellis, et aL, 2006; Lees & Davies, 1987; Manolopoulos, et aL, 2006; Rodano & Tavana, 1993), or a run-up 1.5 m away from the ball (Kellis, et aI., 2004; Teixeira, 1999), which may not be indicative of a game situation. In a study of penalty kicks taken in the FIFA World Cup finals and the finals of the European Champions' League, players took between one and ten paces in the approach with an average of five (Hughes

& Wells, 2002). In a laboratory-based study, where no constraints were placed on the approach run for instep and pass kicks, participants typically took three steps (Levanon & Dapena, 1998). The angle of the approach to the ball has also been constrained with participants required to approach the ball from 60° (Clagg, et aL, 2009), within 45° and 60° (Barfield, 1995; Barfield, et aL, 2002) or from a straight run-up (Bull Andersen, et aL, 1999; Dorge, et aL, 2002). Research has shown that ball velocity and accuracy of penalty shots are improved when taken from a selfselected approach angle compared with approach angles of 30°, 45° and 60° (Scurr & Hall, 2009). Therefore, constraints on the approach run in previous biomechanical studies may have affected results.

2.3.5.

Data processing and analysis

The sampling rate and method of noise elimination can markedly affect kinematic data. A wide range of sampling rates have been used to capture kicking kinematics, with the majority of studies using a sampling frequency up to 200 Hz (Asai, 2000; Barfield, et aL, 2002; Browder, et aL, 1991; Brown, et aL, 1993; Isokawa & Lees, 1988; Katis & Kellis, 2010; Kawamoto, et aL, 2007; Kellis, et aL, 2004; Kellis, et aL, 2006; Lees & Davies, 1987; Lees & Nolan, 2002; Levanon & Dapena, 1998; Nunome, et aL, 2002; Nunome, Ikegami, et aL, 2006; Opavsky, 1988; Orloff, et aL, 2008; Prassas, et aL, 1990; Shan, 2009; Shan, et aL, 2005; 48

Shan & Westerhoff, 2005; Tant, et aI., 1991; Teixeira, 1999). A few researchers have used higher sampling rates from 240 to 400 Hz (Bull Andersen, et aI., 1999; D6rge, et aI., 2002; Inoue, Ito, Sueyoshi, O'Donoghue, & Mochinaga, 2000; Lees, et aI., 2004). Typically, the raw kinematic data have been filtered with a cut-off frequency between 6 and 18 Hz (Barfield, et aI., 2002; Bull Andersen, et aI., 1999; D6rge, et aI., 2002; Manolopoulos, et aI., 2006; Nunome, Ikegami, Apriantono, & Sano, 2004; Nunome, et aI., 2000; Prassas, et aI., 1990; Teixeira, 1999). Although these methods are appropriate for describing the swing motion of the limbs, the duration of ball impact has been reported to range between 9 and 12 ms (Asami & Nolte, 1983; Nunome, Ikegami, et aI., 2006; Nunome, Lake, et aI., 2006) and it is therefore unlikely that displacement data in the studies using lower sampling rates (less than 200 Hz) are able to provide enough data points to adequately describe the short-duration, high-frequency movements of the foot and ankle during ball impact, (Nunome, Lake, et aI., 2006).

Digital filtering is a process widely applied in biomechanics to remove excess noise from

a raw displacement signal,

however signals

involving

large

accelerations such as impacts can be prone to error (Georgakis, Stergioulas, & Giakas, 2002). The problem lies with the fact that in conventional filters a single low-pass cut-off is applied for the entire duration of the whole signal, making it difficult to remove noise whilst retaining the high frequency content of the signal (Georgakis, et aI., 2002). This is particularly important in a kicking motion which involves both low (swing phase) and high (ball impact) frequency data. Smoothing through the impact can affect the pattern of kinematic data before, during and after impact, which are presumed to be the actual movement (Knudson & Bahamonde, 2001 ).

A number of methods have been proposed to deal with problems associated with smoothing

through

impacts.

These

include

linear

and

polynomial

data

extrapolation (Knudson & Bahamonde, 2001), deletion of all raw data from one frame pre-impact to five frames post-impact followed by data interpolation (Reid, Elliott, & Alderson, 2008), and low-pass cut-off frequencies up to 50 Hz to keep the high frequencies intact (Bobbert, Schamhardt, & Nigg, 1991; Nunome, Lake, et aI., 49

2006). Time-frequency filtering has been used also which comprises of an automatic filtering algorithm that adapts to the characteristics of the signal (Georgakis,

et

aI.,

2002).

These

methods

have

all

shown

improved

representations of the real data compared with smoothing data through impact with a low cut-off frequency.

Motor control studies have attributed a reduction in limb speed prior to impact to a potential accuracy enhancing strategy (Teixeira, 1999). Other researchers have associated it with data processing techniques causing an oversmoothing of the large acceleration at impact, (Knudson & Bahamonde, 2001). Knudson & Bahamonde (2001) showed that in a tennis forehand, conventional filtering methods through impact caused a false peak prior to the impact and consistent underestimations of distal joint angles and velocities at impact. Nunome and colleagues (Nunome, Lake, et aI., 2006) demonstrated how sampling at 1000 Hz and eliminating noise using an automatic version of the time-frequency filtering algorithm described by Georgakis et aI., (2002) identified rapid, high-frequency movement characteristics of the foot, ankle and shank during ball kicking that were not apparent using traditional data collection methods (250 Hz sampling rate and Butterworth filter with a constant 10Hz cut-off frequency). These authors found that the shank was still angularly accelerating and the ankle was still linearly accelerating until ball contact, and they did not reach their peak angular and linear velocity respectively until after impact. Nunome and colleagues (Nunome, Lake, et aI., 2006) suggested that the majority of previous research on kicking has failed to acknowledge the limitation of data filtering through impacts and it is likely that the kinematics of the kicking leg around impact have been documented incorrectly in the literature.

The study by Nunome and colleagues (Nunome, Lake, et aI., 2006) was the first study to provide evidence to support the general coaching principle of 'kicking through the ball' and it was suggested this was because it provided more representative kinematics of the distal segment motions before, during and after ball contact. However, the participants were required to perform maximal effort kicks and were instructed to kick the ball as hard as possible. Therefore it would 50

be expected that the participants would kick through the ball as hard as they possibly could. If a reduction in limb velocity prior to impact is an accuracy enhancing strategy, it would occur in tasks with accuracy constraints and where a speed-accuracy trade-off is present. Further investigation is required to clarify whether previously reported kinematic data of the kicking leg have been distorted by the data processing procedures or if football players do reduce limb velocity prior to impact when accuracy constraints are introduced.

2.4.

Concluding remarks

There are a variety of performance analysis methods currently used in football to objectively quantify movements and events in training and competition. These range from simple manual video analysis, to GPS technology and sophisticated computerised tracking systems. The actual method used depends on the information required and the environment in which it is collected. With the large expense of automatic tracking systems, and the prohibition of players wearing GPS technology in competition, manual video analysis is still widely used to monitor competition, especially in the women's game. With football being one of the most televised sports in the world, manual video analysis of television coverage provides a cost-effective method of collating information on elite teams.

Elite football is generally low scoring in nature. Competition analysis and statistics from elite men's and women's football competition have shown approximately one third of all goals originate from a set play, and of these, free kicks are the most effective in creating goals. These findings highlight a skill that is not only effective for scoring goals, but also a relatively closed skill that is easy to simulate in practice compared with the more complex situations of open play.

Technique

analysis

can

improve

our understanding

of the fundamental

characteristics of a skill that relate to an improved performance. In football, the majority of biomechanical analyses of the kicking motion have focused on the maximal instep kick of a stationary ball and used male participants, with only a few investigating curve kicks (which simulate a direct free kick swerving around a

51

defensive wall). Studies investigating the trajectory of a curve kick have used mathematical models of the ball flight, computerised analysis of the foot-ball impact and analysis of male players taking successful curve kicks. Given the inherent anatomical differences between males and females, and gender comparative studies have shown that females may not be capable of achieving the ball velocities of their male counterparts, it is likely that the ball flight characteristics for females achieving successful curve kicks differ from males. Differences in kicking kinematics between genders have been reported in the literature indicating that results from studies with male participants may not be applicable to females. Knowledge of the attributes of successful free kicks and the techniques used to achieve them in elite women's football would therefore be beneficial to women's football coaches and female players in learning and enhancing performance of the skill, and ultimately in scoring more goals.

52

CHAPTER THREE 3. Determination of football pitch locations from video footage and official pitch markings Chapter 3 is reproduced with permission from a paper published in the journal of Sports Biomechanics, with the following reference: Alcock, A., Hunter, A. & Brown, N. (2009). Determination of football pitch locations from video footage and official pitch markings. Sports Biomechanics, 8, 129-140.

3.1. Abstract The ability to determine a specific location on a football (soccer) pitch from television footage would provide a cost effective method of obtaining competition specific information on many professional and international teams. This study presents the accuracy and reliability of a new method of calculating ball location from simulated television coverage and known pitch markings. The coordinates of 99 markers of known location on a football pitch were digitised from video. An intersection point was determined from the equations of two lines that form pitch markings and the relationship from this point to other known pitch coordinates was calculated using a curve-fitting based method. Average error between known and reconstructed measures was 0.21 m for pitch width and 0.11 m for pitch length from a view simulating television coverage. Inter- and intra-rater reliability analyses showed researchers could consistently reconstruct pitch locations to within less than half a metre. The accuracy and reliability of this method will be sufficient for most practical uses in an applied sport environment, although the level of accuracy required will depend on the specific application. This method could be applied to other sports to determine specific locations on a pitch or court or to improve current competition analysis systems.

53

3.2.

Introduction

Video analysis provides an unobtrusive way of collecting competition specific information

and

objective feedback on

performance.

Given

the

cost of

sophisticated analysis systems, together with the increasing cost of international travel and the fact that many sports are televised around the world, it would be beneficial to develop video analysis methods from television coverage. There are several possible approaches, including commercial systems and calibration of the field of view from known pitch markings.

Many professional football (soccer) and rugby clubs use sophisticated analysis systems such as Prozone® (ProZone Sports Ltd., Leeds, UK) or Amisco® (Amisco UK, Birmingham, UK), which use eight stationary, synchronised cameras positioned around the top of the stadium to automatically track the movements of all players, the ball, and officials (Oi Salvo, et aI., 2007; Oi Salvo, et aI., 2006). Such systems have been shown to be valid and reliable for tracking player movements (Oi Salvo, et aI., 2006). Prozone MatchViewer® is a mobile system that can be used from television coverage, as it allows analysis of a match from a single camera but requires users to input the type and time of an event, the players involved, and the pitch location (x-y coordinates) (Bradley, et aI., 2007). In a comparison of trained observers who independently analysed an English Premier League game using Prozone MatchViewer®, Bradley et aI., (2007) found the average difference in pitch location recorded by the observers was 3.6 m with a 95% confidence limit of 8.5 m. The authors suggested that accurately plotting the x-y coordinates of an event onto an overhead representation of the pitch is reliant upon the camera location and angle, as well as reference points on the pitch close to the event, such as penalty area markings. The maximum error in the study was 70 m, indicating the observer misjudged the orientation of the pitch on the system (Bradley, et aI., 2007). If coaches are to use information from video analysis to provide detailed feedback to players and to make important tactical and technical decisions about how they will prepare for competition, it is critical to know the amount of error associated with the system and its reliability. Research has shown that the more quantitative and objective the feedback, the greater effect it has on performance (Franks, 1997). The ability to reconstruct real-world pitch 54

location coordinates from camera image-based coordinates would not only provide an objective measure of pitch location but also allow for the error of the measurement to be quantified.

Direct linear transformation (DLT) (Abdel-Aziz & Karara, 1971) is used extensively for three-dimensional analyses of sports activities. A modified version, the 2D-DLT (Walton, 1981), allows two-dimensional reconstruction based on motion occurring in a single plane such as on a football pitch, based on a minimum of four known control points (Kwon, 1999). The DLT method works regardless of the camera angle to the plane of motion (Kwon & Casebolt, 2006). This is a major advantage in sporting arenas where it is not always possible to place the camera perpendicular to the plane of interest (Brewin & Kerwin, 2003) and for reconstructing locations from television footage because knowledge of camera set-up is not required.

Several studies have investigated the influence of camera locations and different calibration configurations on reconstruction accuracy using DL T. Results have shown reconstruction errors increased when control points were located in only one area of the control region and the most accurate results were obtained when control points were evenly distributed throughout the calibration volume (Chen, Armstrong, & Raftopoulos, 1994; Yeadon & Challis, 1994). A minimum of 24 control points are recommended for the best 2D-DLT calibration accuracy (McLean et aI., 2004). Significant inaccuracies in reconstructions have been observed when required points were located outside the calibration volume and extrapolation was needed (Angulo & Dapena, 1992; Chen, et aI., 1994; Hinrichs et aI., 2004; Wood & Marshall, 1986), or if reconstructed coordinates were near the edge of the calibrated area (Brewin & Kerwin, 2003). Thus for 2D-DLT to be used for locating positions on a football pitch, careful camera positioning and a large field of view are required.

Toki & Sakurai (2005) placed a single camera at the highest point of the audience stand of a stadium so that the entire football pitch was in the field of view and used 2D-DLT to calculate player positions. The standard error was 0.4 m in pitch width 55

coordinates and 0.55 m in pitch length coordinates (Toki & Sakurai, 2005). Television coverage of football matches typically uses a zoomed view where only a portion of the pitch is in view at anyone time. Thus 2D-DLT could be useful for locating events inside the penalty area where pitch markings can be used to encompass the area where the movement occurs. Up to 20% of goals are scored from outside the penalty area in international competition (Grant, Reilly, et ai., 1998; Yiannakos & Armatas, 2006), thus knowledge of locations outside this area would also be useful but television camera views often do not support use of the DLT method as all control points (i.e. penalty area markings) are located at one end of the control object and extrapolation is required.

Locating events on a football pitch could assist coaches in making tactical decisions. For example, accurately locating from where goals are scored or direct free kicks are taken (a common scoring method from outside the penalty area) could assist with decisions on when to shoot at goal or how to defend a team approaching the goal. Furthermore, adding temporal information to accurate ball locations could be used to estimate ball velocities or the time available for the goalkeeper to react and save a shot. If this information were available from television coverage, competition specific information on many professional and international teams throughout the world would be readily available and would provide a cost effective method of studying opposition teams. Therefore, the aims of this study were to develop a method of calculating a specific location on a football pitch from digitised video simulating television coverage, especially areas outside the penalty area, and to compare the error of the method with 2D-DLT.

3.3.

Methods

A football pitch measuring 68 m x 105 m was used in this study. The pitch matched the pitch dimensions stipulated for international competitions by Federation Internationale de Football Association (FIFA, the football governing body). Ninety-nine markers were placed at known locations using 100 m and 50 m tape measures (Figure 3-1). Markers were 0.21 m in diameter, similar to the diameter of an official match ball (0.22 m). Two digital video cameras (Sony DCR56

TRV950E, Sony Corporation, Tokyo) were located approximately 7 m above ground level, in line with the middle of the pitch and 3 m from the touchline. One camera remained stationary throughout the study and was positioned to maximise the number of known pitch markings and the amount of the pitch in the field of view (Figure 3-1); this view, containing 69 visible markers, was used to compare the reconstruction error of the developed curve-fitting method with the 2D-DLT method. The second camera panned and zoomed to simulate television coverage and all marker locations were reconstructed using the curve-fitting method.

Distance from Touchline (m)

I

o o o

5

iii o

(!)

G /

o

0

I

G /0 (0 @0 G

E 20

.g

!

/

CD

u

15 25 1;;

/

i5 30 /

35/ /

/0

e e

G G 8

e

/

,40 45

Figure 3-1: Markers with a diameter of 0.21 m were placed on the pitch at each of the 99 known locations. Distances shown are the official dimensions of the penalty area stipulated by FIFA and were confirmed on the test pitch used in this study. Two video cameras were located at approximately x = -3 m, y 52.5 m, and 7 m above the pitch. The dashed line shows the large field of view in which 69 markers were visible and the black circles represent the 11 control points used for 2D-DLT analysis.

=

57

3.3.1.

20-0LT method

With the 2D-DLT technique, the image plane coordinates of 11 control points (Figure 3-1) were digitised in a single video frame from the stationary camera view using 2D-DLT software (Ariel Performance Analysis System, Ariel Dynamics Inc., California, USA). The large field of view allowed for the control points to be well distributed throughout the control area. The software was used to determine the eight DLT camera parameters and subsequently reconstruct the coordinate locations of the 69 visible markers. Reconstruction errors were calculated as the absolute difference between the known and reconstructed real-world coordinates of the markers. For the large field of view, a paired t-test was used to compare the reconstruction errors from the curve-fitting method with the 2D-DL T method.

3.3.2.

Curve-fitting method

To reconstruct the real-world pitch locations from image plane pixel coordinates, the following calibration procedure using known pitch markings as control points was employed. First, video footage was captured and control points from still frames were manually digitised using software specifically written for this application in Visual Basic for Applications (Microsoft Corporation, Washington, USA) (Appendix A) and a digitising window resolution of 1156 x 867 pixels; this corresponds with the 4:3 aspect ratio of the camera. The control points used to reconstruct marker location were dependent on the camera view. A minimum of two parallel lines and two other points along both the x (pitch width) and y (pitch length) axes was required for reconstruction (Figure 3-2). The control points used were the goal line, base of the goal posts, penalty mark, the lines defining the goal area and the penalty area, the two touchlines, and grass mowing lines visible across the pitch, which were measured to ensure they were parallel with the goal line.

58

...

...

... ... ...

...

...

...

...

" ... ,

...

...

...

,

...

...

... ...

... ...

" ...

... ...

... ...

...

D'

A line 2

y=m2 x+b2

Figure 3-2: The method for calculating the camera view using known pitch markings as control points. In this example, the two longest lines used to calculate the intersection point coordinates are line 1 (AA': goal line) and line 2 (DO': front of penalty area). Pixel coordinates of other known pitch markings, B (front right corner of goal box) and C (penalty mark), are used to calculate the gradient from these features to the intersection point (m3 and m4). The slope and magnitude of the line from these control points to the intersection point are entered into a least squares regression analysis to calibrate the field of view.

In the image plane, two lines that are parallel on the pitch will converge and eventually intersect because of perspective. The equations of any two lines parallel with the goal line can be used to determine their intersection point, which in turn can be used to define the equation of a line from the intersection point to any other location in the camera field of view (Figure 3-2). The two points chosen to define the line equation were as far away as possible from each other to reduce error in the calculation of the intersection point. Where more than two parallel lines were visible, the two longest lines in the field of view, as measured in the image plane, were used.

59

The equations of lines 1 and 2 are:

Y = 111lx+bl Y

At the point where these two lines intersect,

= 1112X+b2

111 IX

+ b l = 111 2 X + b 2

(1 ) (2) (3)

Furthermore,

XinterseCI

= (hI -b2) /(1112 -1111)

(4)

and

Yintersect

= In l Xintersect + b l

(5)

The gradients of lines from the intersection point (Xintersect, Yintersect) to the control points were then determined. Where the control point was a line, the point on the line furthest from the intersection point was used to reduce error. The known distance of each control point and the gradient from each point to the intersection point was introduced to LAB Fit software (Silva & Silva, 2006) to determine the calibration factors between image-based and real-world coordinates by means of least squares regression analysis (Levenberg-Marquardt algorithm). For example, the four points (A-D) depicted in Figure 3-2 would be introduced to the software as: A = (m1, 0), B = (m3, 5.5), C = (m4, 11) and D = (m2, 16.5).

For each camera view, all visible known pitch markings points were used as control points to reduce the impact of systematic errors on reconstructed coordinates. The order of the equation that best modelled the data differed across views. Therefore, to ensure consistency across curve models and confidence that locations digitised from television coverage would contain comparable error, all known points were reconstructed based on their image-based coordinates, and if the reconstructed location differed more than 0.15 m from its known location, the model was rejected and points re-digitised until this threshold was met. A value of 0.15 m was set a priori from pilot work based on the ability to reliably achieve this result from a large field of view and this value is less than 0.5% of pitch size. The gradient from the image-based coordinates of the marker to the intersection point was then calculated and introduced into the equation to give the real-world coordinates and thus the real distance to the marker from the goal line.

The same principles were used to determine the distance from the marker to the touchline by using the two longest lines parallel with the touchlines to find the 60

intersection point, before gradients from all visible pitch markings to the intersection point were calculated. The relationship between the gradients and the official distances was determined with LAB Fit Software and the gradient from the marker to the intersection point was input into this equation to determine the distance from the near touchline to the marker.

To investigate the dispersion of reconstruction errors using the curve-fitting method, one way analyses of variance (ANOVAs) with post hoc Scheffe tests were used. Markers that were of equal distance from the goal line were grouped to assess whether reconstruction error changed as the distance from the goal line increased. Next, markers at equal distance from the touchline were grouped to investigate reconstruction error distribution as the distance from the touchline increased.

3.3.3.

Reliability analysis

To evaluate the repeatability of the curve-fitting method, nine balls placed randomly around the penalty area at locations realistic to where direct free kicks may be taken from during competition were digitised. Between view reliability was assessed by comparing the reconstructed coordinates of each ball from two different still frames taken from the simulated television footage. Each ball was reconstructed four times and the mean was used for comparison. Inter-rater reliability was determined by a second researcher estimating the nine ball locations from the two different views (n = 18). To assess the intra-rater reliability and the applicability of the model to television coverage, 70 televised direct free kicks in international matches were captured. Of these, five did not contain enough control points in the field of view for the curve-fitting method to be applied, thus the locations of 65 were reconstructed using the curve-fitting method. Each free kick location was reconstructed three times by the same researcher who was blinded from previous results and with at least three days between analyses. It was noted that only 32 of the 70 free kicks could have been analysed using the 2D-DLT method without extrapolation from control points located in the penalty area. For the intra-rater reliability analysis, reconstruction error was calculated as the

61

absolute mean deviation (the absolute difference between the reconstructed coordinate and the mean of the three trials).

3.3.4.

Sensitivity analysis

The curve-fitting method is designed to be applicable to television coverage, thus it is assumed it is not possible to measure the pitch before a game and confirm the exact distances among official pitch markings. To determine how potential inaccuracies in pitch markings affected the reconstructed ball coordinates, all official pitch dimensions used to reconstruct the positions of six markers in realistic free kick taking locations were manipulated using two methods. First, all control points used for the calibration were adjusted by 0.1 m alternating in both positive and negative directions; for example, the four points depicted in Figure 3-2 would be entered into the curve-fitting software as A 11.1) and D

= (m2,

= (m1, 0.1), B = (m3, 5.4), C = (m4,

16.4). Second, values were adjusted so the first half of the

known points were increased and the other half decreased by 0.1 m; that is, values of A

3.4.

=(m1, 0.1), B =(m3, 5.6), C =(m4, 10.9) and D =(m2, 16.4) were used.

Results

With a large field of view, the reconstruction errors of the curve-fitting method and the 20-DLT method were comparable for pitch width coordinates (Table 3-1; Appendix B). For pitch length, 2D-DLT reconstruction errors were significantly less

(P = 0.02), but the difference of 0.11 m is less than the measurement error of both the curve-fitting and DL T methods (Appendix B). The 2D-DLT method also produced the maximum reported error (Table 3-1) for marker four (Figure 3-1), which was located just outside the calibrated area and therefore required extrapolation.

62

Table 3-1: Mean ± SO reconstruction error in metres resulting from the different methods and different views (maximum absolute errors are reported in parentheses). Large field of view (n=69) Axis

2D-DLT method

x (pitch width) Y (pitch length)

Simulated tv (n=99)

Curve-fitting method

Curve-fitting method

0.35 ± 0.27 (0.91)

0.37 ± 0.32 (1.34)

0.21 ± 0.22 (0.92)

0.26 ± 0.25* (1.39)

0.37 ± 0.25 (1.07)

0.11±0.08 (0.35)

* indicates statistical significance with curve-fitting method (P

=0.02)

The reconstruction error in pitch width calculations increased as the distance from the goal line to the marker increased. Markers located within the penalty area were always reconstructed within 0.21 m of their known location with respect to pitch width, while markers further out had errors of up to 0.92 m (Figure 3-3). Results of the ANOVA revealed reconstruction errors of markers placed 30 m or more from the goal line were significantly greater than those 5-15 m out (P

S;

0.042). Pitch

length reconstruction error of marker locations was never more than 0.35 m, regardless of where the marker was placed on the pitch (Figure 3-4). The ANOVA revealed no difference in reconstruction error of pitch length coordinates as the distance from the camera to the markers increased. On average, the curve-fitting method was able to reconstruct the location of markers within 0.21 m for pitch width and 0.11 m for pitch length of their real-world location, giving an average resultant error of 0.24 m. The maximum resultant error was 0.99 m, which is less than 1.5% of the width and 2.5% of the length of the control area used in this study.

63

Distance from Touchline {Ill)

I)

5 10 ¢)

c::

:: 15 n:1 0

t'!l

E 20

-

e I II

I.}

c::

.! til is

.07

.14

c

25

.{is

J)ll

{,

.os

J)4

.IlJ

.Hl

.O;t

84

30 .llr

.92

35

,!Ie

.1l9

40

,82

Figure 3-3: Error in metres of reconstructed pitch width coordinates from simulated television coverage. Larger circles reflect larger digitising errors. These measurements represent 27 different still frames from the simulated television coverage.

64

Distance from Touchline (m) 10

li>

22

28

4ll

46

(I

.04

5

.28

.05

.1)3

.!l8

52

-.015

.13

.-

E 10

.01

III t::

:: 15

.00

!Ii

05

\'.)

t!)

E 20

l';

0

.03

01

~ 25 c::

J)2

An

.l:;

... !1iI

J)S

m

:15

.07

,l)5

.01

'()4

04

,04

.liZ

Ifl

030

,1)3

35

.08

40 .01

.1)1

.10

.00

.n .12: .00

.01l

.iJ'l

.02

Figure 3-4: Error in metres of reconstructed pitch length coordinates from simulated television coverage. Larger circles reflect larger digitising errors. These measurements represent 27 different still frames from the simulated television coverage. All markers were calculated within 0.35 m of their known location.

Reliability analyses showed researchers were able to consistently reconstruct locations on a football pitch to within approximately half a metre and the curvefitting method could be applied across multiple views and to real television coverage of international matches. Pitch length reconstructions were more repeatable than pitch width measurements (Table 3-2).

65

Table 3-2: Mean ± SO reconstruction error in metres for reliability analyses (maximum absolute errors are reported in parentheses). Between view

Intra-rater

Inter-rater

reliability (n=9) 1

reliability (n=65) 2

reliability (n=18) 3

x (pitch width)

0.21 ± 0.12 (0.43)

0.16 ± 0.17 (1.00)

0.41 ± 0.38 (1.10)

Y (pitch length)

0.14 ± 0.12 (0.41)

0.08 ± 0.12 (0.65)

0.18 ± 0.15 (0.47)

Axis

1 Nine

locations were reconstructed from two different views each

265 locations from real television coverage were reconstructed three times each 3 The

nine locations and both views used to assess between view reliability were

reconstructed by a different researcher

Adjustment of pitch markings to reflect potential inaccuracies of official pitch markings (± 0.1 m) affected methodological error by less than 1%. This 0.1 m adjustment did not affect the 0.15 m threshold between the known and reconstructed values determined a priori. Calculated locations differed from the original reconstructed coordinates by an average of 0.02 ± 0.15 m in pitch width and 0.06 ± 0.09 m in pitch length measurements from simulated television coverage.

3.5.

Discussion

A method was developed to reconstruct real-world locations on a football pitch from digitised video, without prior knowledge of the camera set-up. The reconstruction error of the curve-fitting method was slightly greater than that of 20OLT, however it was more usable for television coverage where small fields of view are favoured. The 20-0LT method requires a wide field of view so that the control area encompasses the required location and control points are distributed throughout the area to minimise error. Compared with the error found using 20OLTon an entire football pitch (Toki & Sakurai, 2005), the average absolute reconstruction error observed in this study using the curve-fitting method was less because the views used were much smaller so that each pixel covered a smaller real-world distance. Results of this study revealed that while on average the 20OLT method had greater accuracy for the same field of view, the error increased 66

substantially when markers were located outside the control area, which is consistent with the literature (Angulo & Oapena, 1992; Chen, et aI., 1994; Wood & Marshall, 1986). In the sample of 70 free kicks from televised international matches, only 32 were suitable for 20-0L T analysis without extrapolation because mowing lines were not able to be used because the intersection with the touchline was not visible. The main advantage of the curve-fitting method over the 20-0LT method for analysing television coverage is that only one known coordinate is required, thus a point on a line can be used to create the model, whereas 20-0LT requires both the x- and y-coordinates.

The small field of view commonly used in televised coverage is not only advantageous to the curve-fitting method because pixels cover a smaller realworld distance, but also because the camera is far away from the control area and zoomed in, thus reducing the effect of any lens distortion on reconstruction error (Kreighbaum & Barthels, 1996). Relative to the control object dimensions, the maximum resultant reconstruction error of 0.99 m using the curve-fitting method on simulated television coverage equates to less than 1.5% of the pitch width and 2.5% of the pitch length. This is comparable to the error found by Angulo & Oapena (1992), who reconstructed coordinates within an 8 m volume using a specially designed calibration frame and control points located with a maximum error of ± 1.5 mm in any direction.

In an applied sport analysis environment, the error associated with the curve-fitting method would be sufficiently precise for most practical purposes. Although errors of up to 1 m were found when locations further away from the camera were reconstructed or wider and deeper fields of view were used, this level of accuracy could still provide coaches with competition specific information useful for tactical decisions and training simulations. For example, a shot taken 25 m from goal that takes 1 s to reach the goal line would be travelling at 25 m.s-1. With a ball location error of 1 m, the error in the calculated velocity would be 4%, whereas an error of 8.5 m, coinciding with the 95% confidence limits of Prozone MatchViewer® (Bradley, et aI., 2007), would increase the velocity error to 34%. This has implications for comparing competition and training based data, obtaining

67

information on opposition players, and providing accurate and reliable information to the coach that may influence future tactical decisions. Whether the accuracy of the curve-fitting method is acceptable for research studies will depend on the requirements of each particular investigation. If access to the pitch and control over the camera set-up and field of view are possible, 2D-DLT may provide superior accuracy.

The reconstruction of pitch width coordinates was less accurate and reliable than pitch length coordinates, likely due to a number of reasons. First, aside from the touchlines, all control points used to create the model were inside the penalty area and because this area was distant to the camera, these points were more prone to error. To calculate locations of markers outside this area, the equation of the lines from the known pitch markings to the intersection point were extrapolated. Thus any errors in the digitised pixel coordinates of the pitch markings were amplified. For pitch length calculations, markers were always within a digitised area, thus it was never necessary to extrapolate lines and consequently errors across the pitch. Error in pitch width tended to increase as markers were located further away from the near touchline where the camera was located. This is likely because the markers located further away were more prone to large real distance errors and because a larger field of view was required so that the penalty area was visible and could be used to perform calculations. When pitch length coordinates were reconstructed from simulated television coverage, error was similar across the entire pitch, resulting from known points being located over the entire area, allowing zoomed views to be used for each marker.

In the method presented here, the optimal curve fitted to the data was not the same across views, only valid within the limits of the digitised area, and could not be extrapolated. Therefore, for this method to be effective in determining exact locations of balls outside the penalty area, it is necessary to know the width of the pitch and to use mowing lines or other markings at known distances (for example, advertising hoardings) along the pitch length. Although only one pitch was used to determine the error of the curve-fitting method in this study, the mathematical principles should apply to any pitch. The simulated television coverage where the

68

camera panned and zoomed are representative of different camera positions and views, indicating the method works regardless of camera configuration. Because this method was created to be applicable to television coverage, it would not be possible to validate the method on every pitch it is used for, or inspect every pitch marking to ensure the correct dimensions. The sensitivity analysis was conducted to investigate how a deviation of 0.1 m in pitch dimensions would affect reconstructed coordinates and results showed the 0.15 m threshold between inputted and calculated values for each known marking was still met and ball locations were influenced by less than 1%. However it is not possible to say how greater inaccuracies in pitch markings would affect reconstructed coordinates. Thus, for the error reported in this study to be valid, it must be assumed that the pitch markings used were within 0.1 m of the official dimensions stipulated by FIFA.

As with all digitising analysis methods, the more erroneous the image coordinates used, the more erroneous the reconstructed coordinates (Kwon & Casebolt, 2006). Thus, like other methods, the curve-fitting method is dependent on accurately determining the image-based coordinates of the intersection point. The location of this point in turn depends on the two lines used to determine the intersection point being parallel and then accurately digitising two points on these lines. Between view reliability analysis showed locations could be reconstructed from different views and thus different intersection points to within an average of 0.21 m of each other in the pitch width direction and 0.14 m in pitch length. The curve-fitting method does not account for non-linear systematic errors caused by lens distortions, however this investigation provides an insight into the general nature of these errors. The effect of lens distortion is dependent on the quality of the lens (Chen, et aI., 1994) and because the main purpose of this study is for the method to be applicable to television coverage where high quality lenses are used, the error is expected to be no more than that reported in this study. Results showed that the effect of these errors on reconstructed coordinates was not substantial, even in the large field of view, and the accuracy is still sufficient for most practical purposes in an applied sport environment. It is, however, recommended to

69

reconstruct the locations several times to reduce the impact of operator errors (Chen, et a!., 1994), as was done in the reliability analyses in this study.

3.6.

Conclusion

The curve-fitting method presented here provides a tool for calculating ball or player location on a football pitch using known pitch markings and digitised video. Reconstruction error was comparable to that of the 2D-DL T method previously used for similar analyses but had greater utility for television coverage and small fields of view. From simulated television coverage, the reconstruction error was least in and around the penalty area where most digitised points were located, thus the method would be useful for locating from where free kicks are taken and goals are scored. The error found using this method would be sufficiently accurate for most practical purposes in an applied sport environment; however, the accuracy required will depend on the specific application.

Although the aim of this study was to develop a method of calculating the location of the ball in football from previously captured video, the results could be applied by video analysts of other sports to calculate specific locations on a pitch or court or to improve current competition analysis systems. If access to the pitch or court is allowed before competition, known points could be added providing these are in the same plane as the pitch. It is recommended: (1) to maximise the number of known control points and locate control points around the entire area; (2) to use zoomed views and cameras on both sides of the pitch where possible; (3) to measure the pitch beforehand where possible -

some sports allow pitch

dimensions to differ or allow a certain amount of tolerance in their dimensions, thus knowing the exact distance of each feature will further improve accuracy; and (4) to digitise views at least three times and use an average for improved accuracy and reliability.

70

CHAPTER FOUR 4. Analysis of direct free kicks in the Women's Football World Cup 2007 Chapter 4 incorporates a paper published in the European Journal of Sport Science, with the following reference: Alcock, A. (2010). Analysis of direct free kicks in the women's football World Cup

2007. European Journal of Sport Science, 10,279-284.

4.1.

Abstract

The location and outcome of all free kicks taken directly at goal in the 2007 women's football World Cup were assessed to identify areas with the most goal scoring potential and assist with tactical decisions and training design. Video of all free kicks taken directly at goal in the 32 games were captured and the location of the ball on the pitch was calculated from pitch markings and image pixel coordinates using a customised curve-fitting method. The outcome of each free kick was determined and for those that resulted in a goal or were saved, information on the flight time and the placement of the ball relative to the goal was reported. All seven free kicks that resulted in a goal were taken from a central area within 7 m of the penalty circle, placed at the edge of the goal within approximately 1 m of the goalpost, and had an average flight time of 1.09 s, which was significantly faster than those that were saved. All free kicks directed towards the bottom and centre of the goal resulted in straightforward saves for the goalkeeper. It is recommended that teams should consider a direct shot from free kicks awarded within 7 m of the penalty circle. For free kicks from wide areas and areas further from the goal, players should be aware of their individual ability and only take a shot when they perceive the probability of scoring a goal to be high. Otherwise, alternative attacking strategies should be considered to avoid an easy turnover of possession. 71

4.2.

Introduction

In elite football (soccer), approximately one third of goals are scored either directly or indirectly from a set play, irrespective of the tournament (Yiannakos & Armatas, 2006). Therefore, preparation and planning of set plays from both offensive and defensive points of view are important for winning games (Armatas, Yiannakos, Papadopoulou, et aI., 2007). In men's domestic and international football, Carling and colleagues (Carling, et aI., 2005) reported a recent increase in set play efficiency (defined as more goals scored from fewer set plays), and successful teams are more efficient than their opponents at scoring from set plays.

Free kicks are consistently the most effective set play for scoring goals (Carling, et aI., 2005) and analysis of the men's European Championships in 2000 showed that direct shots from free kicks in central areas were more effective than a short pass followed by a shot at goal (Ensum, et aI., 2000). Using a mathematical model of ball flight, Bray & Kerwin (2003) advocated that a well executed direct shot from a free kick gives a goalkeeper little chance of saving a goal. However, no research to date has analysed the characteristics of ball flight and how these relate to the outcome of the free kick. Previous research on penalty kicks among male players has shown the placement of the ball relative to the goal is related to the shot outcome, with those placed towards the top of the goal and within one yard (0.91 m) of the goalpost being more likely to result in a goal than those placed towards the bottom and centre of the goal (Hughes & Wells, 2002; L6pez-Botella & Palao, 2007; Morya, et aI., 2004). The pace of the shot is also important, as it determines how long the goalkeeper has to react and save the ball (Kerwin & Bray, 2006), but if the penalty taker kicks it maximally there is more chance of missing the goal (Hughes & Wells, 2002).

The potential for a direct free kick to score a goal is largely dependent on the pitch location from where it is taken, as this influences the distance the player must kick the ball, the positioning of any defensive wall of players and the angle to the goal. To date, no known scientific study has reported the specific pitch locations of free kicks, although scientists from Liverpool John Moores University, in their report for the Football Association, found a progression in free kicks from the 1991-92 72

English Premier League season to the 1997-98 season (Williams, et aI., 1999). That is, in 1991-92 the pitch locations of free kicks used for direct shots at goal encompassed areas extending to the halfway line and touchlines, whereas in the 1997 -98 season direct attempts were confined to central areas. In addition, twice as many free kicks taken in the 1997-98 season resulted in a goal being scored compared with the 1991-92 season (Williams, et aI., 1999). There has been no research on free kicks in women's football, although statistics from previous World Cup tournaments show a progressive increase in the number of goals scored directly from free kicks (Table 1-1).

Quantitative analysis is required to determine the location on the pitch where free kicks are taken directly at goal, the outcome of the free kicks, where in the goal the ball is placed, and the flight time of the ball in elite women's football. This information would identify areas where female players were capable of scoring from, define areas with the most goal scoring potential, and determine attributes of successful free kicks, which in turn could facilitate decision making on where free kicks should be practised from in training and when a direct shot at goal from a free kick should or should not be attempted. The women's football World Cup provided an opportunity to collect quantitative data on direct free kicks at the current highest level of women's international football.

The aims of this study were to: 1) quantify all pitch locations where players attempted to score directly from a free kick in the 2007 women's World Cup; 2) determine the outcomes of these free kicks (scored, saved, hit the wall or offtarget); and 3) analyse the characteristics of ball flight for all free kicks that resulted in a goal or were saved (flight time and placement relative to the goal).

4.3.

Methods

The procedures of the study were approved by the Human Research Ethics Committee of Southern Cross University (approval number ECN-07-64). A still frame of each free kick taken directly at goal in the 2007 women's football World Cup finals was captured using InterVideo® WinOVO® (Corel Corporation, USA)

73

and video footage of all 32 matches played in the tournament. The intention of the player to shoot directly at goal was determined by a nationally accredited football coach. The location on the pitch from which each free kick was taken was calculated using a customised curve-fitting method as reported by Alcock and colleagues (Alcock, Hunter, & Brown, 2009). Briefly, pixel coordinates of pitch markings which constitute the official pitch dimensions (the goal line, base of the goal posts, goal area, penalty mark, penalty area, halfway line, and the touchlines) and grass mowing lines (gradations in grass colour on the pitch due to grass mowing) parallel with the goal line were determined by digitisation. The coordinates of two parallel lines were used to calculate the intersection point (Figure 3-2). The gradients from the intersection point to other known pitch markings were calculated and the relationship between the known distances and their gradient to the intersection point determined using least squares regression analysis (LAB Fit Curve Fitting Software; (Silva & Silva, 2006)). The distance to the ball from the goal line was then calculated using the equation of best fit provided by the curve fitting software. This method can reliably calculate ball position within an average error of 0.24 m, which is comparable to the diameter of an official match ball (0.22 m) and considered acceptable for the purpose of this study (Alcock, et aI., 2009). All distances were calculated three times and the average taken to reduce operator error.

The locations of 65 direct free kicks were calculated. Due to insufficient pitch markings in the field of view from television coverage, the distance from the ball to the touchline was unavailable for six additional direct free kicks, therefore these locations were estimated and are clearly shown as estimates in the results. The distance from the goal line to the ball had sufficient visible points for analysis of all 71 kicks.

An experienced football analyst classified the outcome of each direct free kick as one of the following: goal, easy save (defined as the goalkeeper was not under any pressure and easily caught the ball), difficult save (goalkeeper was under pressure to save the ball - that is, had to move fast or dive to catch the ball or push it away from the goal), hit the defensive wall of players, hit a player not in the wall,

74

or off-target. Direct kicks that resulted in a goal or were saved were investigated further by recording the position of the ball relative to the goal as it crossed the goal line or, in the case of those that were saved, would have crossed the goal line. To do so, the goal was divided into six areas (Figure 4-1). The flight time of the ball, defined as the time taken from the instant it was kicked to when it crossed the goal line, was estimated from the video footage using slow motion and a resolution of 0.04 s. A one way analysis of variance with post hoc Scheffe tests was used to determine any differences in the flight time of the ball and the different free kick outcomes. All statistical analyses were conducted using the Statistical Package for Social Sciences (SPSS Inc., Chicago, IL) and significance was set at

P < 0.05.

Top 1

Top 2

Top 3

Bottom 1

Bottom 2

Bottom 3

Figure 4-1: The position of the ball relative to the goal as it crossed (or would have crossed) the goal line for all direct free kicks that resulted in a goal being scored (or were saved) were classified into one of these six areas of the goal, as viewed from the penalty mark.

The reliability of the developed notation systems was determined because it is important to quantify the validity of the data collection methods, especially if training and tactical decisions are to be made based on the results (O'Donoghue, 2007). Intra-observer reliability of each measure was determined by repeat analysis of all free kicks. Where there were inconsistencies between the first and second observations, the "correct" outcome was taken as the outcome of a third observation.

Inter-observer

reliability

analysis

limits

operational

bias

or

misinterpretations of operational definitions (James, Taylor, & Stanley, 2007) and

75

was determined

by analysis of all free kicks by a second

researcher,

inexperienced in football (Table 4-1).

4.4.

Results

The intra- and inter-observer reliability of the developed notation systems was high for all performance variables (Table 4-1).

Table 4-1: Intra- and inter-observer reliability following repeat analysis of all kicks. Performance variable

Reliability analysis

Intra-observer

Inter-observer

method

reliability

reliability

96%

97%

96%

96%

0.01 ± 0.02

0.02 ± 0.04

Shot outcome (goal,

Percentage

difficult save, easy

agreement of all

save, hit wall, off-target,

direct free kicks

hit player not in wall).

(n=71 )

Shot placement relative

Percentage

to the goal (classified

agreement of all

into one of six areas).

shots on target (n=23)

Flight time (seconds)

Difference between

(mean ± SO)

two repeat measures (n=23)

In the 32 games played in the 2007 women's World Cup tournament, 359 free kicks were taken in the attacking half of the pitch, of which 71 were directly at goal. Seven of the 111 goals scored in the competition were scored direct from a free kick, equating to one goal every 4.6 games. The seven goals direct from free kicks were taken by six different players (one player scored twice) from six different countries. Four of these countries did not progress past the group stage and two reached the quarter finals.

76

50 • goal Ii. easy save difficult save ... hit wall ~ off target • hit player not in wall

45 40

12m from penalty arc (zone 2)

_ 35 E ........, Q)

~

30

0

E

25

("!l

Ii.

...

7m from penalty arc (zone 1)

~ 20 ro

~

~

~.

.

...

0

.j::

~

~

~

~ ~

... •

~

More than 12m from penalty arc (zone 3)

~

i

Ii.

~

Cii Cl

® ~

~

"

• •

... ... ~

...Ii.

~

t-Ii.

~

~



c

(j)

@

i5 15 ~

10 Q)



5 ::c u :::l 0

0

Goal line

I-

0

5

10

15

20

25 30 35 40 45 Distance from touchline (m)

50

55

60

65

Figure 4-2: Locations and outcomes of all free kicks taken directly at goal in the 2007 women's football World Cup. Those contained within a circle represent estimated distance of the kick from the touchline as the field of view had insufficient information to calculate the location.

The outcome of the direct shot from the free kick was influenced by the location from which it was taken 1. All seven goals were scored from a central area within 7 m of the penalty circle (zone 1) (Figure 4-2). Of the 22 direct free kicks taken from this area, seven resulted in a goal being scored, six hit the defensive wall, and the remaining nine were off target. No direct free kicks taken from this area were saved by the goalkeeper; that is, all those that avoided the defensive wall and were on target resulted in a goal.

Eighteen free kicks were taken directly at goal from the area between 7 and 12 m from the penalty circle (zone 2). Of these kicks, five were saved by the goalkeeper (28%) (three easy saves and two difficult saves), seven were off-target (39%) and 1

Chi-square analysis showed differences were significant (P < 0.001). However the data violated

the assumption that the expected frequency in each cell is greater than or equal to 5.

77

the wall still played an important defensive role by blocking six shots (33%). A further 31 direct free kicks were taken from the area more than 12 m from the penalty circle (zone 3). Of these, three forced the goalkeeper to save under pressure (10%), two hit the defensive wall, both from wide areas (6%), and eight resulted in an easy save for the goalkeeper (26%). Seventeen of the free kicks taken from this area were off target (55%) and one hit a player not in the wall (3%). The outcome of the shot was influenced by where the ball was placed in the goal 1. All the free kicks that scored or forced a difficult save by the goalkeeper were directed around the edge of the goal, within approximately 1 m of the goalposts or crossbar, whereas those directed at the bottom and centre of the goal resulted in an easy save (Figure 4-3).

The flight times of the seven goals were significantly faster (mean ± SO: 1.09 ± 0.14 s) than those of the difficult saves (1.96 ± 0.53 s; P = 0.005) and the easy saves (1.77 ± 0.43 s; P = 0.007). There was no difference in flight time between the easy and difficult saves (P = 0.675).

~, Goal

Easy save

Difficult save

Figure 4-3: Schematic representation of the goal as viewed from the penalty mark to illustrate where shots crossed the goal line or were saved, estimated from video footage. All free kicks that resulted in a goal being scored or forced a difficult save were placed within approximately 1 m of the goalposts or crossbar. All easy saves were directed towards the bottom and centre of the goal, with the exception of one placed at the right of the goal which had a flight time of 2.32 s.

78

4.5.

Discussion

The locations, outcomes and flight characteristics of all free kicks taken directly at goal in the 2007 women's football World Cup were quantified. All free kicks taken from within 7 m of the penalty circle (Figure 4-2) that were on target and avoided the defensive wall resulted in a goal being scored, highlighting an effective area to choose a direct shot at goal and practise this skill. All goals from free kicks had a flight time less than 1.24 s and were scored at the edge of the goal, within approximately 1 m of the goalpost (Figure 4-3). Comparisons of pitch locations and flight characteristics of goals scored from free kicks in previous women's football tournaments or men's football are not possible due to the paucity of scientific literature investigating free kicks.

In a study on penalty kicks in males, Kerwin & Bray (2006) found the corners of the goal could not be protected by the goalkeeper in the available time and players could guarantee scoring if they shot into this "unsaveable zone". Although the flight time of the penalties in their study (0.64 ± 0.01 s) was shorter than that of the free kicks that resulted in a goal being scored in the present study, it is likely there is an "unsaveable zone" for free kicks in the women's game also. Given the success rate of direct free kicks from within 7 m of the penalty circle at this tournament, elite teams should consider a direct shot at goal from this area and aim for the corners of the goal.

No goals were scored from direct free kicks taken between 7-12 m from the penalty circle (Figure 4-2). However, 28% of shots forced a save from the goalkeeper and a further 33% were blocked by the defensive wall, suggesting that this area may have potential for creating scoring opportunities and that defending teams still perceive this a dangerous area for direct shots. However, in contrast to zone 1, all free kicks taken from zone 2 that avoided the defensive wall and were on target were saved by the goalkeeper. This could be because the female players who attempted direct shots from here were unable to kick the ball fast and accurately enough from this area in order to defeat the goalkeeper or because the longer flight time provided the goalkeeper with increased perceptual information of ball flight and thus longer to react and save the ball. It is also possible that in a

79

larger sample of games, goals could or would be scored directly from a free kick from locations on the pitch further from goal. If the ability of female players to score from set plays more than 7 m from the penalty circle could be improved, the probability of scoring directly from a free kick would increase considerably. For example, if the success rate of free kicks in zone 1 (32%) was extended for the 40 free kicks taken across zones 1 and 2, the number of goals direct from free kicks would increase to one every 2.5 games. Of the free kicks taken from more than 12 m from the penalty circle (zone 3, Figure 4-2), more than 80% were either offtarget or an easy save, which provided an easy turnover for the opposition. Players who are unable to consistently perform fast and accurate free kicks that target areas of the goal that make it difficult for goalkeepers to make a save should consider an alternative attacking strategy, such as crossing the ball or passing and building an attack.

The three zones identified support findings of a comparison of free kicks taken in the men's English Premier League in the 1991-92 and the 1997-98 seasons (Williams, et aI., 1999). In 1991-92 direct shots from free kicks were attempted from areas extending to the halfway line and touch line , while in 1997-98 teams crossed or passed the ball from wide areas and direct shots were confined to a central zone just outside the penalty area. It is possible that coaches and/or players became aware of the efficacy of this area in creating goals and the ineffectiveness of wide areas and areas far from goal given the limits of their skill level. Players should be aware of their unique strengths and weaknesses and look for alternate attacking strategies if the free kick is awarded outside their individual range for a direct shot.

Analysis of flight characteristics of free kicks that were saved revealed that placement in the goal was more important than flight time. That is, on average easy saves reached the goalkeeper faster than difficult saves but because the ball was directed at the bottom and centre of the goal the goalkeeper was able to make an easy catch. No difficult saves were made in the bottom half of the goal and no easy saves were made in the top half. The goalkeepers facing free kicks in this study chose to push balls directed at the top central area of the goal over the 80

crossbar instead of catch it, possibly because they were unable to get their body behind the ball and protect their goal. It is recommended that players taking free kicks should target the top corners of the goal, as this was found to be the most successful area for scoring goals. In addition, if the ball is within the goalkeeper's reach and they are forced to push it over the crossbar, the attacking team will still retain possession. The flight time of the shot was still important and a trade-off between ball speed and accuracy was demonstrated with the easy save made near the right goalpost having a flight time of 2.32 s (Figure 4-3). Free kicks with a long flight time give the goalkeeper more time to react and move to save the ball.

To improve female players' direct free kick ability, the focus should be on increasing ball velocity while maintaining accuracy from areas further from goal. According to Newton's first law of motion, ball trajectory and velocity result from the forces applied to it by the kicking foot at impact, which in turn result from the kicking technique prior to impact. Therefore, a kinematic analysis of players' kicking techniques would facilitate decisions on who should take a free kick and when, as well as enable the coach to train a player to improve aspects that would facilitate a better executed free kick.

Although the research questions of this study relate to attacking strategies of direct free kicks, inferences can be made pertaining to effective defensive strategies. That is, defenders should be aware of the consequences of conceding a free kick in central areas close to the goal and, when facing a direct free kick, understand the importance of a well organised and structured defensive wall. Future research should investigate the position and number of defenders in the wall and the position and movements of the goalkeeper when successfully defending free kicks.

The findings presented here relate to the present sample of games and the players that participated in the 2007 women's football World Cup. It is plausible that different findings could be observed in a different sample of games and players. Differences in skill levels, defensive strategies, and the ability of the opposing goalkeeper can affect game dynamics and outcomes. However, the 81

present results provide an insight into female players' ability to score directly from free kicks at the current highest level of the game. Ball locations were accurate within an average of 0.24 m and flight times within 0.05 s. These errors should be considered when inferring practical implications of the results.

4.6. Conclusion Objective, competition specific information is reported for the locations and outcomes of free kicks taken directly at goal in the 2007 women's football World Cup. Central pitch locations close to the penalty area produced more goals and forced the goalkeeper to make saves under pressure compared with wide areas and those taken from further away, likely due to the improved angle to the goal and reduced flight time. The placement of the free kick in the goal was important, with all free kicks resulting in a goal or a difficult save placed within approximately 1 m of the goalposts or crossbar. Players taking free kicks should target the top corners of the goal because this was the most successful area for scoring goals, and even if the ball does go within the goalkeeper's reach and it gets pushed over the crossbar, the attacking team still retains possession. Players should be aware of their individual ability and only attempt a direct shot from a free kick when they perceive the chance of scoring to be high. If the free kick is awarded outside their individual range for a successful shot, other attacking strategies should be considered to avoid an easy turnover of possession. This information is beneficial for practising free kicks from specific areas in training and decisions on when a direct shot at goal should or should not be attempted in women's football.

82

CHAPTER FIVE 5. Methodology for ball flight characteristics and kinematic analysis

5.1.

Introduction

Ball velocity and accuracy are vital components of a well-executed direct free kick. According to Newton's first law of motion, the flight characteristics of the ball are determined by the magnitude and direction of the forces applied to the ball by the kicking foot at impact, which in turn result from the kicking technique prior to impact. Thus, to achieve different trajectories, different impact characteristics between the foot and the ball are required (Asai, 2000).

In a direct free kick, a defensive wall is set up to prevent a straight shot at goal. In practice, the wall covers one goalpost and extends to approximately 75% of the goal line, leaving a clear sight of the kick for the goalkeeper (Bray & Kerwin, 2003). For players able to strike a ball well with spin, this provides an opportunity to curve the ball around the wall and into the corner of the goal, a skill that has been perfected by a small number of male free kick specialists such as David Beckham, Roberto Carlos and Cristiano Ronaldo and provides a spectacular sight for spectators. Despite the recent increase in the number of goals scored directly from free kicks in elite women's football, initial flight characteristics of curve kicks by females and the techniques employed to achieve them remain neglected in the literature.

A detailed biomechanical analysis of elite female players taking a free kick that would likely score in an elite women's football match will provide information on the initial ball flight characteristics utilised to avoid a defensive wall and score a goal as well as the technique used to achieve that trajectory. Current literature on the female kicking technique has focused on instep kicks. An instep kick, where

83

the ball is hit with the medial-superior portion of the boot (Levanon & Dapena, 1998), is kicked with a straight trajectory and is commonly used for generating a fast ball speed (Nunome, et aI., 2002). Descriptions of the mature kicking technique in the literature generally depict an instep kick (Plagenhoef, 1971; Wickstrom, 1975). A full-body three-dimensional kinematic analysis of both curve and instep kicks will enhance understanding of the technique adaptations employed by elite female players in order to achieve the different ball trajectories. This will provide information on specific coaching points to facilitate development of the curve kick skill.

The purpose of this chapter is to describe in detail the methods used to collect data on the ball flight characteristics and the full-body three-dimensional kinematics of elite female football players performing curve and instep kicks (Studies 3 and 4). Additionally, the treatment of the data including the specific motion analysis modelling techniques and the choice of filtering level for both the ball and the kinematics is discussed. The statistical procedures and specific variable definitions used for analysis are described in the relevant ball flight and kinematics chapters (Chapter 6 and 7 respectively).

5.2.

Laboratory and equipment set-up

A 17 camera Vicon motion analysis system (Vicon Motion Systems, Oxford, UK) sampling at 250 Hz was used to define a volume with approximate dimensions of 6 m wide, 8 m long and 3 m tall (Figure 5-1). The cameras were set at varying heights and angles such that all markers could be seen by at least two cameras at anyone time (Figure 5-2). This configuration allowed the last stride of the run-up, the entire kicking action and the initial ball flight to be captured whilst placing minimal restrictions on the length and angle of the run-up. The cameras were positioned back from the capture volume such that players could take at least five run-up strides from any approach angle. The run-up area was covered with artificial turf (Figure 5-1), consisting of a 37 mm polypropylene pile attached to a nylon backing (EnduroTurf Supreme, Enduroturf Pty Ltd, Australia) to simulate a

84

game. The laws of the game state that football matches can be played on natural or artificial surfaces (FIFA, 2009a, 2009b).

The 17 cameras were calibrated prior to data collection in accordance with standard Vicon procedures. First, a dynamic calibration was performed by waving an L-frame wand with five markers attached at known locations throughout the entire capture volume. All cameras were calibrated with a mean residual of less than 0.5 mm. Next, the same wand was used for a static calibration to define the origin and orientation of the global coordinate system, with the positive y-axis pointing towards the goal, the positive z-axis pointing vertically upwards and the xaxis as the cross product of the two (Figure 5-1). The Vicon system is able to measure known distances between markers to within 1 mm and known angles to within 1.5 0 (Richards, 1999; Windolf, G6tzen, & Morlock, 2008). A digital video (DV) camera (Sony DCR-TRV950E, Sony Corporation, Tokyo) was positioned approximately 7 m behind the ball and approximately 3 m above the ground with the goal in the field of view to monitor the accuracy of the kicks (Figure 5-1). An image of a full size football goal with a 1 m2 target in the top right corner (for right-footed players) was projected on to a white sheet 20 m from the ball (Figure 5-3). The target was directly in front of the ball (Figure 5-1). For the left-footed player, the image of the goal was mirrored so that it represented the top left corner of the goal, but the position of the target relative to the ball start position remained the same. Therefore the tasks for right- and left-footed players were the same. The set-up of the laboratory task was based on the findings from the 2007 women's football World Cup (Chapter 4) that showed direct free kicks with the best potential for scoring a goal in elite women's football were taken from central areas within 27 m of the goal and placed in the top corner of the goal, within approximately 1 m of the goalpost and crossbar. A distance of 20 m was used in the laboratory because that was the maximum distance available that would allow for the entire movement and ball flight to be performed in the laboratory whilst placing minimal constraints on the athlete's run-up.

85

LEGEND ""

\/icon camEra Ot.:tlinE of artifi cia I it.: rf

'.;Vall

9:.15 rn

2:::- rn

,--------------------------- -----------,

,, I

I

, ,.- - - - - - - - - --'

,,

,

,,

I

J

, I _________________ - - - - , I L

D\l carnera

Figure 5-1: A schematic of the laboratory set-up for the data collection procedures for a right-footed player. Not to scale.

86

-..:::"-'"::::M:;;'

',::,

Figure 5-2: View of the laboratory set-up in the frontal plane as viewed from the goal to illustrate the height and angle of the 17 camera positions.

For curve kicks, a defensive wall consisting of four 'players' (The Soccer Wall Company, Lewisville, Texas) 1.83 m tall and 1.78 m wide was placed 9.15 m from the ball to simulate a direct free kick (Figure 5-3). The wall position was determined by the head coach of the national team and placed so that 1.5 mannequins were outside the goalpost when viewed from the ball (Figure 5-4).

Figure 5-3: The laboratory set-up for data collection of a curve kick trial. The set-up was the same for instep kicks with the exception of the defensive wall. The red (1 m2) and yellow squares show the target area, and the four mannequins represent a defensive wall. 87

Figure 5-4: The four mannequins and the position of the defensive wall used for the curve kicks. Note the position of the wall was such that 1.5 mannequins were outside the goalpost when viewed from the ball.

5.3.

Participant information

Fifteen elite female footballers (mean ± SO: age = 20.1 ± 3.2 years; height = 1.69

± 0.06 m; mass = 61.6 ± 5.8 kg) participated in this study. Fourteen were current representatives of the Australian national team who were ranked 14th in the world at the time of data collection. The number of full international appearances ranged from one to 73 with a mean of 17. The remaining participant was a member of the current national squad but had not yet played a full international match. This player was an integral part of the Australian Under 20s team and was identified by the national coaches as possessing a high level of skill in performing curve and instep kicks.

All participants were injury free at the time of testing and played in field-based positions. Based on self-selection, participants were classified as right- or left-foot 88

dominant with only one player considering herself as left footed. Participation was dependent on reading the Participant Information Sheet (Appendix C) and signing the Informed Consent form (Appendix D). Ethical approval was obtained for all testing procedures from both the Human Research Ethics Committee of Southern Cross University (approval number ECN-08-014) and the Australian Institute of Sport Ethics Committee (approval number 20070402).

Participants wore a crop top, cycling shorts and short ankle socks to allow all markers to be attached directly to the skin and therefore alleviate any errors caused by movement of the clothing. Participants wore their own football boots that they used for training and competition. This was important for ecological validity (George, et aI., 2003) and to get as true a representation of the foot-ball contact as possible because the material properties of the boot influences the player-ball interaction (Hennig & Sterzing, 2010), which in turn will affect the ball velocity and trajectory.

5.4.

Overview of the motion analysis model

For three-dimensional motion analysis, it is necessary to assume that each segment can be modelled as a rigid body with a minimum of three anatomical landmarks required to represent bone motion (Campbell, Alderson, Lloyd, & Elliott, 2009). Human movement analysis requires definition of systems of axes associated with each segment that is incorporated in the model (Della Croce, Leardini, Chiari, & Cappozzo, 2005). Early models used to derive threedimensional joint kinematics and kinetics were based on the Conventional Gait Model and developed using the minimum number of markers possible when measurement systems were only capable of detecting a few markers (Baker, 2006).

There

are

two

major problems

with

this

type

of model.

First,

misidentification of anatomical landmarks leads to inaccuracies in modelling joint centres and segments and the hierarchical nature of the model whereby proximal segments must be defined in order for distal segments to be calculated leads to propagation of errors from one segment to the next (Baker, 2006). Secondly, soft tissue artefact, which refers to surface markers incorrectly replicating bone motion 89

due to underlying active and passive tissue movement (Campbell, et aI., 2009) is both task specific (Fuller, Liu, Murphy, & Mann, 1997) and location specific (Schache, Baker, & Lamoreux, 2008) with markers placed over anatomical landmarks particularly susceptible to tissue movement (Bobbert, et aI., 1991; Cappozzo, Catani, Leardini, Benedeti, & Della Croce, 1996). The 'Calibrated Anatomical Systems Technique' or 'CAST' (Cappozzo, Catani, Della Croce, & Leardini, 1995), in which anatomical landmarks are located and then their positions recorded relative to technical markers located on the same rigid segment can be used to minimise soft tissue artefact (Cappozzo, et aI., 1996). The technical markers can be placed arbitrarily on the portion of the segment considered least susceptible to soft tissue movement (Cappozzo, et aI., 1995) and to optimise marker visibility with respect to camera positions and occlusion by the participant (Elliott & Alderson, 2007).

The biomechanics group at the University of Western Australia (UWA) has developed and validated a full-body marker set and model for three-dimensional motion analysis based on CAST (Besier, Sturnieks, Alderson, & Lloyd, 2003; Lloyd, Alderson, & Elliott, 2000). This 'UWA model' uses customised software written in Vicon Bodybuilder and was the model used for this study. The model incorporated two sets of coordinate systems: the technical coordinate system used to define bone movement during the kicking trials and the anatomical coordinate system defined during static calibration trials. Clusters of markers were placed on each segment to create a technical coordinate system (Besier, et aI., 2003; Cappozzo, et aI., 1995). These clusters comprised of three reflective markers attached to semi-malleable pieces of T -shaped plastic. The technical coordinate systems were used during the kicking trials to define the positions and orientation of the segments, but the positioning of the clusters was on the portion of the segment considered to be least susceptible to skin movement rather than having any anatomical relevance (Lloyd, et aI., 2000). Thus, for meaningful kinematic data and intra- and inter-subject reliability, the position of the technical coordinate system had to be referenced to the anatomical coordinate system (Cappozzo, Della Croce, Leardini, & Chiari, 2005; Della Croce, Camomilla, Leardini, & Cappozzo, 2003) which comprised of markers placed on known anatomical 90

landmarks. Static calibration trials were used to record the positions of the anatomical coordinate system markers relative to the technical coordinate system of the cluster located on the same segment. Subsequently for the kicking trials, the markers located on anatomical landmarks were removed and the athlete was relatively unimpeded during performance as clusters were not positioned around joints (Elliott & Alderson, 2007). This allowed markers located on anatomical landmarks to be considered as virtual markers during dynamic trials (Campbell, et aI., 2009). These reconstructed positions relative to the corresponding cluster were then determined during the kicking trials and any error due to soft tissue artefact was reduced because the anatomical coordinate system is less susceptible to errors caused by unwanted marker movement (Elliott & Alderson, 2007).

The convention used to describe joint kinematics, with the exception of the shoulder, corresponded to the standard Euler Z-X-Y order of rotation in accordance with the International Society of Biomechanics (ISB) standard (Wu & Cavanagh, 1995). The sequence of rotations was: flexion/extension about the zaxis of the proximal segment, then adduction/abduction about a floating x-axis and finally an internal/external rotation about the y-axis of the distal segment (Besier, et aI., 2003). The shoulder joint coordinate system was defined with a Y-X-Y order of rotation to comply with a more recent ISB convention (Wu et aI., 2005).

5.4.1.

Motion analysis model marker set

For the static calibration trials 73 reflective markers (Table 5-1), 14 mm in diameter, were attached to the participant using non-allergenic double-sided adhesive tape by a practitioner experienced in this procedure for elite athletes (Figure 5-5 and Figure 5-6). For kicking trials the static markers were not used but the positions defined by these markers were held within the corresponding segment technical coordinate system (Table 5-2).

91

Table 5-1: Marker positions for three-dimensional segment definition during the kicking trials. For positioning of clusters refer to Figure 5-5 and Figure 5-6. Dynamic markers Segment

Marker

Location

Head

LFHD

Left front head

LBHD

Left back head

RFHD

Right front head

RBHD

Right back head

Trunk

C7

yth cervical vertebra

T10

10th thoracic vertebra

CLAV

Clavicular notch

STRN

Xiphoid process of the sternum

Shoulder (left & right)

ACR

Acromion process

Upper Arm

UA1

Superior marker of upper arm cluster

(left & right)

UA2

Middle marker of upper arm cluster

UA3

Inferior marker of upper arm cluster

Forearm cluster

FA1

Superior marker of forearm cluster

(left & right)

FA2

Lateral marker of forearm cluster

FA3

Medial marker of forearm cluster

Hand

CAR

Carpal

(left & right)

HNR

Dorsal radial knuckle

HNU

Dorsal ulnar knuckle

Pelvis

RPSIS

Right posterior superior iliac spine

LPSIS

Left posterior superior iliac spine

RASI

Right anterior superior iliac spine

LASI

Left anterior superior iliac spine

Thigh cluster

TH1

Superior marker thigh cluster

(left & right)

TH2

Middle thigh cluster

TH3

Inferior thigh cluster

Tibia cluster

TB1

Superior marker of tibia cluster

(left & right)

TB2

Middle marker of tibia cluster

TB3

Inferior marker of tibia cluster

92

Non kicking foot

CAL

Calcaneus

MT1

First metatarsal head

MT5

Fifth metatarsal head

CAL

Calcaneus

MT5

Fifth metatarsal head

Kicking foot cluster

FT1

Posterior marker of foot cluster

- lateral side of boot

FT2

Middle marker of foot cluster

FT3

Anterior marker of foot cluster

Kicking foot

Table 5-2: The additional markers attached to participants for the static calibration and the corresponding segment technical coordinate system in which these positions were stored. These markers were removed for the kicking trials. Static markers Joint

Marker

Location

Cluster

in

which

position held Shoulder

ASH

(left & right)

Anterior Shoulder (anterior

Upper arm cluster

glenohumeral joint centre) PSH

Posterior Shoulder (posterior glenohumeral joint centre)

Elbow

LEL

Lateral humeral epicondyle

(left & right)

MEL

Medial humeral epicondyle

Wrist

WRR

Radial styloid process

WRU

Ulnar styloid process

(left & right) Knee

LFC

Lateral femoral condyle

(left & right)

MFC

Medial femoral condyle

Ankle

LMAL

Lateral Malleolus

(left & right)

MMAL

Medial Malleolus

Kicking foot

MT1

First metatarsal head

Upper arm cluster

Forearm cluster

Thigh cluster

Tibia cluster

Kicking foot cluster

93

Figure 5-5: Anterior view of the static calibration trial.

94

Figure 5-6: Posterior view of the static calibration trial.

A modification of the UWA motion analysis model was required for the kicking foot as the marker located over the first metatarsal (MT1) would have obstructed the contact between the boot and the ball. Therefore, a small cluster was attached to the lateral aspect of the boot (Figure 5-7) and the coordinates of the MT1 relative 95

to this cluster were determined during the static calibration trial. This allowed for the MT1 marker to be removed during the kicking trials and a virtual MT1 marker to be reconstructed following data collection.

Figure 5-7: A cluster was affixed to the lateral aspect of the football boot to allow for the first metatarsal marker to be considered as a virtual marker during kicking trials. This ensured that the contact between the boot and the ball was not affected by any external materials.

The validity of this method in accurately recreating the MT1 marker was determined by performing three dynamic trials with the MT1 marker still attached to the boot, reconstructing the marker as if it had been removed and comparing the difference between the actual and the reconstructed marker. For these dynamic trials, a simulated kicking action was performed but no contact with a ball was made. The difference between the actual and reconstructed coordinates of the MT1 marker was calculated for the

X-,

y- and z-axes. The mean difference

between the actual and reconstructed markers across the three axes and for the three trials was 0.84 mm which is comparable to the accuracy of the Vicon motion analysis system (1 mm) and was considered acceptable for the study.

96

5.5.

Ball marker set

Five pieces of reflective tape (approximately 2 cm 2 ) synonymous with that used on the reflective Vicon markers attached to the participants, were adhered to each of four FIFA approved balls (Figure 5-8). Four balls were used to reduce time spent recovering the balls between trials. Whilst the markers were not placed in specific locations on the ball, care was taken such that the markers were not coplanar and three were not collinear.

Figure 5-8: One of the FIFA approved balls used during the test procedures with two markers made from reflective tape visible. Markers were approximately 2 cm 2 •

5.6.

Data collection - kicking trials

Following a standardised warm-up of five minutes on a stationary bike at a self selected pace, participants performed five practise trials of a curve kick. A FIFA approved ball was placed 20 m from and directly in front of the target. Players were instructed to kick with the same focus on accuracy and velocity as in a

97

match; that is, attempting to hit the target whilst also beating the goalkeeper. There was, however, no goalkeeper for the laboratory test. For the curve kicks they were also instructed to curve the ball around the wall as they would in a direct free kick. No constraints were placed on the players with regards to the length or angle of their approach run other than those imposed by the size of the capture volume. That is, no instructions were given as to how the players should approach the ball, but the camera configurations meant the length of the run was limited to approximately five steps for large approach angles (i.e., approaching from the side of the ball) (Figure 5-1). No instructions were provided as to how the ball should be kicked or what part of the foot should be used. Following the five practise trials, participants performed 15 curve kick trials for which full-body three-dimensional kinematic data were collected. Adequate rest was provided between trials. The defensive wall was then removed from the laboratory set-up and five practise instep trials were performed with participants instructed to drive the ball straight at the target. They were again instructed to kick with the same focus on accuracy and velocity as in a match situation. Kinematic data were then collected for 15 instep kick trials. The order of the curve and instep kicks was alternated for each participant to offset any learning or familiarisation effects across participants. After each participant had completed their kicking tasks, the artificial turf was brushed to prevent lean and flattening of the grass pile as synthetic fibres have a tendency to lean in a particular direction or flatten with use (FIFA, 2009a).

5.7.

Data analysis

Each kick was deemed as acceptable for analysis if full-body three-dimensional kinematics were available from before support foot contact until after followthrough, and the ball markers were in view following impact until it had left the capture volume. The primary criterion for inclusion was accuracy, as determined from the DV camera footage. Visual observation of the shot accuracy, as determined from the DV camera footage, revealed that all participants had five or more curve and instep kicks that were close to the target and considered acceptable for the purpose of the study. Therefore, the five most accurate instep

98

kicks and five most accurate curve kicks for each participant were used for analysis (curve n = 75; instep n = 75).

5.7.1.

Filtering data through impacts

Although digital filtering is a process widely applied in biomechanics to remove noise from a raw displacement signal, data involving large accelerations such as impacts can be prone to error due to inadequate data processing (Georgakis, et aI., 2002). Smoothing through impact can affect the pattern of data before, during and after impact, which are presumed to be the actual movement (Knudson & Bahamonde, 2001). Knudson & Bahamonde (2001) showed that conventional filtering methods through impact caused a false peak prior to the impact and consistent underestimations of distal joint angles and velocities at impact for a tennis forehand. Nunome and colleagues (Nunome, Lake, et aI., 2006) suggested that the majority of previous research on kicking has failed to acknowledge this data processing limitation and it is likely that kinematics of the kicking leg around impact have been incorrectly documented in the literature. A number of methods have been reported to deal with problems associated with smoothing through impacts, which were discussed in Chapter 2, section 2.3.5. In choosing an appropriate filter treatment, it is important to remove noise whilst retaining the high frequency content of the signal (Georgakis, et aI., 2002).

5.7.2.

Choice of filter

Initially data were filtered using a quintic spline (Woltring, 1986) which is built in to the Vicon software. To determine the most appropriate filter level for the kicks, a residual analysis was performed whereby raw data were filtered at a range of different mean square error (MSE) values and the residuals between the filtered and raw data at the different MSEs were compared (Winter, 2005). Raw data from six kicking trials (one curve and one instep kick from three different players) were filtered using a quintic spline at an MSE of 5, 10, 15, 20, 25, 30, 40, 50, 75, 100 and 125 prior to being modelled with the kinematic model described in section 5.4. The most appropriate MSE was determined by visual observation of the graphs of the Woltring MSE plotted against the residual. A trend-line was extended from the 99

linear end of the exponential curve to the y-axis and then a line parallel with the xaxis was projected from this y-intercept back to the curve. Finally a perpendicular line from this point on the curve to the x-axis was used to estimate the most appropriate MSE value for each variable of interest (Winter, 2005) (Figure 5-9). This method of visual inspection has been used in the assessment of other elite sporting actions such as tennis serves (Reid, 2009) and cricket bowling (Portus, 2006).

Woltring MSE value

-0.04 -0.042

20

60

40

80

100

120

1 0

-0.044 -0.046

ro

~

-0.048

III

~ -0.05 ~

~ -0.052 c

-0.054 -0.056 -0.058 -0.06

Figure 5-9: Example of the residual analysis to determine the most appropriate MSE value for kicking hip flexion/extenion linear velocity in an instep kick.

The residual analysis on selected kinematic variables revealed an MSE of 20 would be appropriate (mean

= 17,

range 12-21) which is comparable to other

sporting movements assessed at the Australian Institute of Sport using the same motion analysis system (cricket bowling = MSE 25; hammer throwing = MSE 15; gait analysis = MSE 10). However, for toe and ankle velocities at impact, even an MSE as low as 5 showed a considerable shift in the timing of peak velocity and reduced the velocity at impact by approximately 2 m.s- 1 which equated to 11 % of the raw velocity signal (Figu re 5-10).

100

18

--Raw -MSE5

16

MSE 10 ~MSE

14

15

MSE20

8 6

o

2

4

6

8

10 12 14 16 18 20 22

24 26 28 30

32

34 36

38 40

Frame number (one frame is 0.004 seconds)

Figure 5-10: Example graph of the kicking leg ankle resultant linear velocity for a curve kick. The point of ball impact is at frame 20.

Observation of trials from each of the 15 participants indicated that the velocity pattern immediately prior to and following impact was evident and similar in all individuals irrespective of the type of kick (curve or instep). This demonstrated that this was not random noise and was real data. It was concluded that the Woltring filter was inappropriate in dealing with kicking impact signals where the frequency content varies dramatically with time, and therefore the suitability of a Butterworth filter was investigated.

Raw data from six kicking trials (one curve and one instep kick from three different players) were filtered using a dual-pass second-order Butterworth filter in Vicon Workstation using a Vaquita Plug-In (Vaquita Software, Zaragoza, Spain). Cut-off frequencies of 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100 Hz were implemented prior to being modelled as described in section 5.4. A residual analysis from right foot contact prior to the kick to the right foot contact after follow through (to remove endpoint effects) revealed the appropriate Butterworth filtering level to range from 10 to 50 Hz with a mean of 32 Hz (Table 5-3). As expected, the toe and ankle required higher frequencies to allow the higher signal content through (Table 5-3; Figure 5-11). 101

Table 5-3: Values representing the appropriate cut-off frequency (average of six trials) of the Butterworth filter as determined by residual analysis. Axis

Cut-off frequency (Hz)

5th metatarsal velocity

3D

41.5

Ankle joint centre linear velocity

3D

44.33

Knee joint centre linear velocity

3D

32.17

Hip joint centre linear velocity

3D

14.5

Ankle angular velocity

3D

36.83

Knee angular velocity

3D

33

Hip angular velocity

3D

28.67

Ankle angle

3D

35.33

Knee angle

3D

33.6

Hip angle

3D

24

Foot global position

Z (transverse plane)

31.6

Pelvis rotation angle

Z (transverse plane)

20.5

Variable

32

Mean cut-off frequency for all variables

18

Impact

-Raw

16

-~

.§.

14 12

::- 10

'(3

o

~

8 6

o

2

4

6

8

10

12

14

16

18

~

22

~

~

~

~

~

~

~

38

~

Frame number (one frame is 0.004 seconds)

Figure 5-11: An example of the effect of different cut-off frequencies on the resultant ankle linear velocity of the kicking leg. The point of ball impact is at frame 20. 102

For a Butterworth filter and a cut-off frequency of 50 Hz to be applied to all kinematic data, it was necessary to investigate the effect that it would have on other kinematic variables of interest. For the variables identified in Table 5-3 as having a lower frequency content, and therefore a harsher filter could be more appropriate (e.g. hip joint centre linear velocity), it was found that there was little difference in the data trace whether it was filtered at 20 Hz or 80 Hz (Figure 5-12). Similar results were found for pelvis angles and thorax angles also. That is, the traces were very similar irrespective of the filtering level.

A fast Fourier transformation was performed on the toe and ankle velocity data for six trials (three participants x two trials (one curve and one instep each)). This revealed a small amount of signal content with a frequency above 50 Hz (Figure 5-13). Analysis of the area under the curve for the same six trials showed that a cut-off frequency of 50 Hz allowed on average 96% of the ankle and toe velocity signal through. These findings support Nunome and colleagues (Nunome, Lake, et aI., 2006) who found substantial frequency content up to 60 Hz in their study of kicking kinematics of male participants.

4

____ Raw

3.5

3

~ 2.5 E

~

i!'

2

~

1.5

'0 o

0.5

o

+-----~------~----~------~----~----~------~-----,

o

10

30 40 20 50 Frame number (one frame is 0.004 seconds)

60

70

80

Figure 5-12: Resultant hip joint centre linear velocity for a curve kick. Impact is at frame 50. 103

700 600 500

~ 400 :::J

:!:::

c 0)300 ('C! :E 200

100 0 0

10

20

30

40

50

60

70

80

90

100

Frequency (Hz)

Figure 5-13: Fast Fourier transformation on the resultant linear toe velocity of a curve kick.

In summary, following residual analysis, fast Fourier transformation and visual observation of the effect of different filters and filtering frequencies on the kinematic data at impact, filtering with a second-order Butterworth and a cut-off of 50 Hz was the optimal for both reducing noise and retaining the high frequency content signal. Furthermore, whilst it was sufficient for retaining the high frequency content of the distal end of the kicking limb at impact, it did not compromise the signal at the proximal end of the limb or the pelvis and upper body.

5.7.3.

Ball data treatment

Although five markers were attached to the ball, only the four that were in the view the longest were used to calculate the ball linear and angular velocity (spin). No constraints were placed on the players in terms of how they placed the ball down before the kick, therefore the use of five markers allowed for four to be visible prior to impact even if the participant placed the ball down on one of the markers. Additionally, it provided redundancy for one marker in case one came off the ball during impact or flight. Missing data in the ball marker trajectories were interpolated using the fill gap quintic spline function in the Vicon Nexus software.

104

The markers were not placed in specific locations on the ball. However, care was taken such that they were not coplanar and three were not collinear, which allowed for the ball centre to be calculated. From analytic geometry, there is a unique sphere that passes through four non-coplanar points (Beyer, 1987). Customised software written

in

Visual

Basic for Applications

(Microsoft Corporation,

Washington, USA) (Appendix E) was used to calculate the centre and radius of the ball. Although the radius of the ball was not used specifically for any calculations, it was used as an accuracy check. The mean ± SO of the ball radius for the 150 trials used for analysis was 105.1 ± 2.6 mm. When the radius of the markers used in this study (7 mm) were accounted for, because the Vicon system locates the centre of a spherical marker rather than a flat piece of tape and therefore identifies the ball to be smaller than it actually is (Griffiths, et aI., 2005), this gives a mean ± SO ball radius of 112.1 ± 2.6 mm. The laws of the game (FIFA, 2009b) state that the circumference of a FIFA approved ball must be between 680 and 700 mm, thus the radius of the ball (circumference / TT) must be between 108.2 and 111.4 mm, indicating that this method of calculating the centre and radius of the ball was, on average, accurate to within 1 mm. The average radius reported by Griffiths et al. (2005) (88.4 mm) was considerably smaller than that found in this study, likely because of the large pieces of tape (75 mm across) utilised in their study.

The raw data for the ball centre linear velocity showed very little noise in the signal (Figure 5-14). However, when the ball was modelled to determine its angular velocity and spin axis orientation, it was decided that the data should be filtered. To avoid the effect of impact, and because only post impact ball data were of interest for this study, trials were trimmed from the end of impact to the end of the available ball data. To do this, the graph of the ball centre linear velocity was used (Figure 5-14). The end of the foot-ball impact was taken as the frame following peak ball linear velocity, based on the assumption that the ball can no longer accelerate, and can only decelerate, once the application of the kicking force has ceased. Following the trim, trial lengths ranged from 10 to 42 frames. The trimmed trials were then filtered with a Woltring quintic spline and an MSE of 20 to remove any noise from the signal before being modelled as described below. As the

105

impact phase of the signal had been removed, the Woltring filter was considered appropriate for these data.

30

f

25

-+----+----+--+---+---+--.---+-+---+-""--+-+--"-+-"-- " ~-+.., -..-' --+~...--+

- 20

'in E

-£' o(.)

15

I

Cl)

> 10

j

5

o

~~--~~--~--~~--~~--~~--~~--~~--~~~~

o

2

4

6

8

10

12

14 16 18 20 22 Frame number

24

26 28

30 32 34

Figure 5-14: Raw linear velocity data for the ball centre of an instep kick.

5.7.4.

Calculation of ball flight variables

The linear velocity of the ball centre was calculated using the finite difference method (Winter, 2005):

where Vx; is the velocity at the ith sample, and Llt is the time between samples. The angular velocity of the ball and the spin axis orientation were calculated using a customised ball model reported by Chin (2009) and based on the methods described by Jinji & Sakurai (2006) and Sakurai et aI., (2007). This method has been used previously to report angular velocity vectors of tennis serves (Sakurai, et aI., 2007), cricket spin bowling (Chin, Elliott, Alderson, Lloyd, & Foster, 2009) and men's football (Whiteside, et aI., 2010). First, a local coordinate system was 106

created for the ball with the previously calculated centre at its origin. The angular velocity of the ball was calculated by performing a rotation matrix to translate the local coordinate system of the ball to the global coordinate system of the laboratory and then determine the relative rotation of the ball around that origin in each axis (Craig, 2004). The elevation angle (the angle of the spin axis orientation relative to the horizontal) (Figure 5-15) was calculated using the following equation: . An I . AngVelZ EIevatzon g e= arcslll----=--AbsAngVel

where AngVe/Z was the amount of spin about the z-axis (i.e. sidespin) and AbsAngVel was the absolute angular velocity of the ball.

\

Direction of spin

Figure 5-15: The elevation angle (9) was defined as the angle between the spin axis and the horizontal.

Whiteside and colleagues (submitted) have performed a validation of the ball model used in this thesis. By rotating a tennis ball, cricket ball and football at a variety of known spin rates and spin orientations, they found an average error of 0.75 ± 0.58° in elevation angle values, with a maximum error of 1.56° for the football. The error in spin rate was 0.01 ± 0.01 rev.s- 1 for all ball types.

107

5.7.5.

Shot accuracy

For each of the 75 curve kicks and 75 instep kicks, the distance from the centre of the ball to the centre of the target was determined by digitising and then converting the image-based pixel coordinates into real-world target coordinates using customised

software

in

written

Visual

Basic

for

Applications

(Microsoft

Corporation, Washington, USA). The accuracy measurement was taken one frame (up to 0.04 s) after impact with the target area. The known dimensions of the target (1 m2 ) were used to calibrate the area and the x- and z-coordinates of each kick were recorded with the centre of the target at the origin. The validity of the accuracy measure was 1.76 ± 1.03 cm, determined by calculating the location of known points from 15 trials (one for each participant). One curve and one instep kick for each participant (n

= 30) were

digitised a second time to determine the

reliability of the shot accuracy measure. The mean absolute difference between the two repeat measures was 0.80 ± 0.69 cm. All kicks were within 2.2 m of the target centre with the exception of one instep kick for one player which was 3.97 m away (Figure 5-16).

x

Curve

Instep

I

200 x150

E

x

100

~

x x x

x

x ""tH---'rv"':-:-~ x •

x xx

~

x

x x"'* It

X

Xx

-200 -*50 x100

r.

-

100 \50

x "IX

250

300

350

x'

xx x

~.

-1

200

x

x

x coordinate {cm}

Figure 5-16: Illustration of the curve and instep kick accuracy of the shots used for analysis in this study. The origin is the target centre and the solid black line represents the 1 m2 target. 108

CHAPTER SIX 6. Initial ball flight characteristics of curve and instep kicks in elite women's football The following is from a paper that has been accepted for publication in the Journal of Applied Biomechanics.

6.1. Abstract Initial ball flight characteristics of curve and instep kicks were investigated. Fifteen international female footballers performed curve and instep kicks from a distance of 20 m from goal and at a 1 m2 target. Seventeen Vicon cameras tracked threedimensional coordinates of four reflective markers adhered to the ball. Ball flight characteristics were quantified, and the coordinates of the ball relative to the target centre were recorded. The lateral launch angle and the angle of the spin axis relative to the horizontal best predicted the horizontal placement of the ball relative to the target. The vertical launch angle, antero-posterior velocity and amount of backspin best predicted the vertical coordinate. Regression models demonstrated how carefully controlled the flight characteristics must be to hit the target with launch angles constrained within 3°. Curve kicks were characterised

by

significantly greater (P < 0.05) lateral launch angles (curve: 7.35 ± 0.20°; instep: 3.02 ± 0.36°) and vertical launch angles (curve: 18.95 ± 0.26°; instep: 14.07 ± OAr), increased sidespin (curve: 5A4 ± 0.14 revs.s- 1 ; instep: 1.17 ± 0.26 revs.s- 1 ) and spin about the antero-posterior axis (curve: 2.15 ± 0.09 revs.s- 1; instep: OA8 ± 0.12 revs.s- 1), and a more vertical spin axis (curve: 65.97 ± 0.9r; instep: 16.88 ± 4A9°) compared with the instep kicks. This information is beneficial for training players to achieve the characteristics required to score a goal and for identifying which variables will help to rectify any error if a player consistently kicks to the same area relative to the target.

109

6.2.

Introduction

The flight trajectory of a ball is influenced by its initial linear velocity, launch angle, spin rate, spin axis orientation and the air density (Kreighbaum & Hunt, 1978). In football (soccer) many shots and passes are played with sidespin to produce a curved trajectory in the horizontal plane (Neilson, et aI., 2004). The curved trajectory results from an imbalance of pressure distribution around the spinning ball causing it to deflect as a result of the Magnus effect (Passmore, et aI., 2008). For example, in a direct free kick, a defensive wall is set up to prevent a straight shot at goal, but for players able to strike a ball with spin, this constraint can be overcome by swerving the ball over or around the wall and into the goal. A well executed free kick gives a goalkeeper little chance of saving a goal, yet despite the prevalence of free kicks in creating goals at an elite level, surprisingly few studies have investigated the initial launch conditions of the ball that produce these goal-scoring techniques (Bray & Kerwin, 2003).

Theoretical and experimental comparisons have shown curve kicks to be characterised by a reduced ball velocity, increased spin rate, more spin about the vertical axis, a greater launch angle and an increased flight time compared with instep kicks (Carre, et aI., 2002; Whiteside, et aI., 2010). An instep kick, where the ball is hit with the medial-superior portion of the boot (Levanon & Dapena, 1998), is kicked with a straight trajectory and is commonly used for generating a fast ball speed (Nunome, et aI., 2002). For male footballers, instep kick ball velocities have been reported with values up to 34.6 m.s- 1 (Manolopoulos, et aI., 2006; Nunome, Ikegami, et aI., 2006; Nunome, Lake, et aI., 2006). For curve kicks that simulate a free kick, linear velocities ranging from 15.1 to 28.3 m.s- 1 and spin rates between 4.0 and 9.4 revs's- 1 are reported in the literature (Bray & Kerwin, 2003; Griffiths, et aI., 2005; Whiteside, et aI., 2010). This difference in ball velocity between kick types can be attributed to the trade off between the development of ball spin rate and ball velocity (Asai, et aI., 2002). That is, as the distance from the point of force application and the ball centre increases, ball spin increases, but ball velocity decreases. Neilson & Jones (2004) reported that professional male footballers are capable of producing spin rates up to 14 revs.s- 1 . However, increased spin may not be beneficial because of the ensuing reduction in velocity. Players should 110

therefore aim to create no more spin than is necessary to ensure the ball reaches the goal as quickly as possible, giving the goalkeeper minimal time to move and save the ball.

Despite the recent increase in the number of goals scored directly from free kicks in elite women's football (Alcock, 2010), initial flight characteristics of curve kicks by females in realistic match situations remain neglected in the literature. Gender comparative studies of instep kicks indicate that females are generally not capable of achieving ball velocities as high as their male counterparts (Barfield, et aI., 2002; Shan, 2009; Tant, et aI., 1991). It is therefore likely that other ball launch variables would differ between genders also. Comparing the initial ball flight characteristics with the more commonly investigated instep kick from the same location would quantify the magnitude of the modifications in the initial flight characteristics used by elite females to achieve a curved trajectory that is required to both avoid the defensive wall and still score a goal. This would be useful to coaches who could train players to achieve those required flight characteristics.

In this study, ball linear and angular velocities, the orientation of the spin axis, and launch angles of curve kicks, which simulated a direct free kick, were compared with those of instep kicks for elite female footballers. The aims of the study were: 1) to determine the initial ball flight characteristics that best predicted the final placement of the ball relative to a target centre; and, 2) to compare the initial ball flight characteristics of a direct free kick (curve kick) that would likely score in an elite women's football match with those of an instep kick at goal from the same location. It was hypothesised that the curve kicks would be kicked with more sidespin and greater vertical and lateral launch angles than the instep kicks in order to avoid the defensive wall.

6.3. 6.3.1.

Methods Kicking trials

After providing informed consent to the procedures which were approved by the institutional ethics committee, 15 international female footballers performed 15 111

instep and 15 curve kicks of a stationary ball. All kicks were performed in an indoor laboratory to reduce any air flow effects. Participants wore their own football boots and four FIFA approved Nike® balls were used at random to reduce time retrieving balls between trials. Tasks were performed on artificial turf (Enduroturf Supreme, Enduroturf Pty Ltd, Australia).

The structure of the tasks was based on previous research (Alcock, 2010) that showed direct free kicks with the greatest goal scoring potential in elite women's football were taken from central areas less than 27 m from the goal and placed in the top corner within approximately one meter of the goalpost and crossbar. The ball was kicked from a distance of 20 m at a 1 m2 target positioned in the top corner of a projected image of a full size football goal. The ball was kicked from directly in front of the target. For curve kicks, a wall of four 'players' 1.83 m tall and 1.78 m wide (The Soccer Wall Company, Lewisville, Texas) was placed 9.15 m from the ball to simulate a direct free kick in match conditions. The wall position was determined by the national team coach and placed so that 1.5 mannequins were outside the near goalpost when viewing the target from the ball position. One participant was left-footed, for whom the laboratory set-up was mirrored so that the tasks for right- and left-footed players were the same. Left-footed player data were manipulated such that all results are presented for right-footed

players.

Participants were instructed to drive the ball straight at the target for the instep kick and to curve the ball around the wall for the curve kicks. For both tasks the instruction was to kick with the same focus on velocity and accuracy as in a game situation, that is, attempting to hit the top corner of the goal whilst also beating the goalkeeper. There was however, no goalkeeper for the task. No constraints were placed on the number of steps or angle of the approach to the kick. Four pieces of retro-reflective tape with approximate dimensions 2 cm 2 were attached to each of four footballs. Tape was used instead of spherical markers to minimise any effect on the aerodynamic properties of the balls. Whilst these markers were not placed in specific locations, they were placed such that they were not coplanar and three were not collinear. Markers were tracked threedimensionally by 17 Vicon cameras (Vicon Motion Systems, Oxford, UK) sampling 112

at 250 Hz. A digital video (DV) camera (Sony DCR-TRV950E, Sony Corporation, Tokyo) sampling at 25 Hz was positioned behind the ball with the goal in the field of view to monitor kick accuracy.

6.3.2.

Data analysis

Visual observation of the shot accuracy revealed that all participants had five or more curve and instep kicks that were close to the target and considered acceptable for the purpose of the study. Therefore, the five most accurate curve and instep kicks for each participant were used for analysis to offset individual player bias (curve: n

=75; straight: n =75). To quantify the accuracy of each trial,

the distance from the ball centre to the target centre was determined by digitizing and then converting the image-based pixel coordinates of the DV camera footage into real-world target coordinates using customised software written in Visual Basic for Applications (Microsoft Corporation, Washington, USA). The accuracy measure was taken one frame (up to 0.04 seconds) after impact with the target. Medio-Iateral (x) and vertical (z) coordinates of the placement of the ball relative to the target were recorded with the target centre at the origin. Measures of known points and repeat measures of the same points demonstrated that this method of measuring shot accuracy was valid to 0.02 ± 0.01 m and could reliably locate the same position within 0.008 ± 0.007 m. All kicks were within 2.2 m of the target centre, with the exception of one straight kick for one player which was 3.97 m away. Following an exploration into the effect of this outlier on results, the effects were found to be negligible and it was therefore included for analysis.

Three-dimensional coordinates of the four reflective markers were used to calculate the ball centre. From analytic geometry, there is a unique sphere that passes through four non-coplanar points (Beyer, 1987). All Vicon trials were trimmed from the end of the foot-ball impact until the end of the available ball data (trials ranged from 10-42 frames which is approximately 0.9-3.7 m of ball flight). The end of the foot-ball impact was taken as the frame following peak ball linear velocity, based on the assumption that the ball can no longer accelerate, and can

113

only decelerate, once the application of the kicking force has ceased. Ball linear velocity was calculated using the finite difference method (Winter, 2005): T/:

rX i

= Xi+l

-

X i- 1

2M

m.s

-1

where VXi is the velocity at the ith sample, and Ilt is the time between samples. The trimmed trials were then filtered using a Woltring quintic spline and a Mean Square Error level of 20 (Vicon Nexus V1.4; Vicon Motion Systems, Oxford, UK). Only post foot-ball impact data were treated because of problems associated with filtering through impact and ball flight characteristics were the focus of this study. All results are presented as the average of the first ten frames so that the same time frame was used for all trials, which was important because these variables do not remain constant throughout the flight. The average was used to account for any data oscillation over the initial part of ball flight.

The vertical launch angle (y) was defined as the angle between the x-y plane (laboratory floor) and the vertical velocity vector (Figure 6-1). The angle between the medio-Iateral (x) and antero-posterior (y) velocity vectors represented the lateral launch angle (¢) (Figure 6-1). The angular velocity of the ball and its spin axis orientation were calculated based on the methods reported by Jinji & Sakurai (2006) and Sakurai et aI., (2007). Briefly, a local coordinate system was created for the ball with the ball centre at its origin. The angular velocity of the ball was calculated by performing a rotation matrix to translate the local coordinate system of the ball to the global coordinate system of the laboratory and then determining the relative rotation of the ball around that origin in each axis (Craig, 2004). The elevation angle (8), defined as the angle between the spin axis and the horizontal, and the alpha angle (a), defined as the angle between the ball spin axis and the linear velocity vector were calculated (Figure 6-1). This method has been used previously to determine ball spin and spin axis orientation angles in tennis (Sakurai, et aI., 2007), cricket (Chin, et aI., 2009) and men's football (Whiteside, et aI., 2010). A validation of the ball calculations revealed an average error of 0.75 ± 0.58° in elevation angle values and 0.01 ± 0.01 rev.s- 1 in spin rate (Whiteside, et aI., submitted).

114

z (vertical) Ball Linear Velocity Vector

Ball Spin Axis

1 - - - ; - - - - - . . - - - - ; - - - - - . y (towards target)

, ...

... ...

,,

, ... '...

! ...~

x (media-lateral)

Figure 6-1: The vertical (v) and lateral ($) launch angles were calculated from the ball centre linear velocity vectors. The ball elevation angle (8) is the angle between the spin axis and the horizontal, and the alpha angle (a) is the angle between the spin axis and the ball linear velocity.

All the calculated variables were used in standard least squares regression analyses (JMP version 8.0.1, SAS Institute Inc., Cary, USA) to determine the flight characteristics that best predicted the x- and z-coordinates of the kicks relative to the target. The horizontal and vertical placement of the ball were considered separately due to problems associated with quantifying the absolute distance which does not account for the direction from the target centre. All 75 curve and 75 instep kicks were used for the regression analyses to provide a spread of data. Of the kicks that hit the target (curve: n

= 39;

straight: n

= 33),

initial ball flight

characteristics were compared with an independent t-test. Only accurate kicks

115

were used for this analysis to identify the initial launch conditions required to hit the target. A criterion of P5 c(

0

Figure 7-1: Comparison of kicking limb joint resultant linear and angular velocities for all curve and instep kicks normalised from support foot contact to ball impact. Curve kicks had a significantly greater knee angular velocity, and instep kicks had a significantly greater linear velocity of the hip and knee joint at ball impact. 132

The kicking leg swing motion from support foot contact to ball impact fitted a plane with an r2 of 0.93 ± 0.05 and a root mean square residual of 4.9 ± 1.7 mm. The inclination of the kicking leg plane relative to the horizontal was significantly more vertical in the curve kicks (63.7 ± 6.2°) compared with the instep kicks (54.8 ± 3.9°)

(P < 0.001). The angle between the kicking leg plane and the target line was significantly greater for the curve kicks (28.4 ± 4.4°) compared with the instep kicks (8.7 ± 4.9°) (P < 0.001) (Figure 7-2).

a)

b)

Figure 7-2: Trajectories (thin black lines) of the kicking foot and the angle of the kicking leg plane relative to the horizontal target line for an elite female performing a) a curve kick, and b) an instep kick. The body orientation shows the impact position.

7.5.

Discussion

The elite females in this study achieved greater ball velocities when performing instep kicks with accuracy constraints than those reported in the literature for females performing maximal instep kicks (Barfield, et ai., 2002; Shan, 2009). Previous research has shown evidence of a speed-accuracy trade-off with ball velocity dropping to approximately 80% of its maximal value when accuracy and velocity demands are imposed (Asami & Nolte, 1983). Therefore, it appears that descriptions of the female kicking technique in the literature do not represent an elite population, and more research on expert athletes is required. The ball velocities achieved by the elite female participants in this study were slightly higher than those reported for elite male footballers performing instep kicks at a 1 m 2 133

target (20.4 m.s-\ (Lees & Nolan, 2002)) and curve kicks at a 1.2 x 1.5 m target (19.2 m.s- 1 ; (Whiteside, et aI., 2010)). Although females have been found to produce lower ball velocities than their male counterparts in maximal kicks (Barfield, et aI., 2002; Shan, 2009; Tant, et aI., 1991), the findings from this study suggest that when accuracy constraints realistic to match conditions are imposed, elite females are capable of achieving similar velocities as those reported for elite males.

For both kick types, the resultant ankle joint linear velocity and the knee angular velocity peaked at the point of ball impact, supporting the findings of Nunome and colleagues (Nunome, Lake, et aI., 2006) that decelerations of the kicking leg observed prior to impact in previous research were due to inadequate data filtering methods, and not a strategy to enhance accuracy as proposed by Teixeira (1999). Nunome et aI., (2006) were the first researchers to provide biomechanical support for the common coaching recommendation of kicking through the ball in maximal instep kicks. The present study supports and extends the findings of Nunome and colleagues (Nunome, Lake, et aI., 2006) to curve and instep kicks with accuracy constraints representative of match conditions. Although there was no difference in the resultant kicking leg ankle and toe velocities at impact, the instep kick ball velocities were significantly faster than the curve kicks. Previous research has stated that foot velocity at impact is the best determinant of ball velocity, and slower ball speeds were almost exclusively due to slower foot speeds between kick types (Levanon & Dapena, 1998; Nunome, et aI., 2002). However, for a curve kick where the ball must be kicked off-centre to produce the increased spin, and with the trade-off between ball velocity and ball spin (Asai, et aI., 2002), this theory does not hold true. This study supports the suggestion of Whiteside and colleagues (Whiteside, et aI., 2010) that a combination of the foot velocity and the offset distance of the foot impact relative to the ball centre would be a more appropriate predictor of ball velocity.

There was no difference in the resultant ankle and toe velocity at ball impact, although the way in which they were generated and the direction differed between kicks. The instep kick was characterised by a faster approach run, along with an

134

increased linear velocity of the hip and knee joint at ball impact, whereas the curve kicks used a larger knee angular velocity prior to and at impact. This could potentially be due to a control mechanism in the curve kick where players orientate the kicking foot to its optimal position for impact with the ball and then use a large angular knee velocity to generate the foot speed. In a linked segment model such as the kicking limb, knee angular velocity is the major contributor of ankle and toe velocity, and although they were not different in magnitude, the toe was travelling in a more antero-posterior direction for instep kicks and a more medio-Iateral direction for the curve kicks. This implies that the plane of knee extension was more towards the target for the instep kicks, and more across the target for curve kicks, and was demonstrated by the increased angle between the swing leg plane and the target line for curve kicks.

To achieve the different impact mechanics for the curve kicks compared with the instep kicks, elite female footballers took a wider approach angle and placed the support foot pointing to the right of the target. There was no difference in the placement of the support foot heel relative to the ball between kick types. Rather, it was the orientation of the support foot that differed between kick types, and the orientation was comparable to the angle between the swing plane of the kicking limb and the target line. Thus, coaches should advise players to point the support foot in the direction of the kicking leg swing motion, that is, towards the target for instep kicks, and to the right of the target for curve kicks. The differences in the approach angle and support foot orientation resulted in the trunk and pelvis being more rotated to the right of the target in the transverse plane at ball impact for curve kicks compared with the instep kicks. This facilitated the increased angle between the swing plane of the kicking leg and the target line, where the foot trajectory travelled across the face of the goal, and resulted in an increased toe velocity in the medio-Iateral direction at impact for the curve kick. In addition, the foot was travelling vertically upwards at impact, facilitated by an increased backwards lean of the trunk compared with the instep kick where the foot was still travelling towards the ground. The medio-Iateral and vertical direction of the foot trajectory immediately prior to impact in the curve kick is important for applying the

135

forces to the ball that produce spin and the increased vertical and lateral launch angles of the ball necessary to avoid the defensive wall (Carre, et aI., 2002).

The angle between the kicking leg swing plane and the laboratory floor was less for the instep kicks, and the non-kicking side shoulder had greater abduction to stabilise the body during this increased lean to the non-kicking side. This difference in the plane angle could potentially be explained by the orientation of the kicking foot at impact. The ankle of the kicking leg in the instep kick was plantarflexed whereas in the curve kick it was more abducted with little plantarflexion. The increased plantarflexion in the instep kick lengthens the kicking limb and therefore the body must tilt further away from the ball to allow the foot to clear the ground. This supports research on golf club swing planes that showed longer golf clubs have a smaller inclination angle to the horizontal than shorter clubs (Coleman & Anderson, 2007). Differences in kicking leg planes between kick types have previously not been documented. They revealed important coaching implications because the curve and instep kicks require different swing planes in order to achieve the different ball trajectories, and provided a visual representation of the kicking motion that could be useful as a coaching tool. Future research on the kicking leg planes for other kick types would be useful in providing coaching points and visual aids for different kicks.

7.6.

Conclusion

Full-body

three-dimensional

kinematics

of

international

female

footballers

performing curve and instep kicks were quantified. Resultant foot velocities at impact did not differ between kick types, however the way in which the foot velocity was generated differed with instep kicks using a faster approach run to the ball and a greater linear hip and knee velocity at impact, and curve kicks using a greater knee angular velocity. In both kick types, the peak knee angular velocity and peak ankle linear velocity occurred at ball impact providing biomechanical support to the common coaching recommendation of kicking through the ball in curve and instep kicks with accuracy constraints representative of match conditions.

136

This research identified the fundamental coaching points necessary to achieve a curved trajectory of the ball compared with the more commonly described instep kick kinematics. To achieve a curved ball trajectory, players should approach the ball from a wide angle, point the support foot to the right of the intended target, swing the kicking limb through across the face of the goal and impact the ball with the foot moving upwards and in an abducted position. In comparison with an instep kick, curve kicks are characterised by greater kicking foot velocity in the medio-Iateral direction, and along with the increased abduction angle at impact, this allows for a foot-ball contact that is able to produce the necessary ball flight characteristics required to avoid the defensive wall and still score a goal.

137

CHAPTER EIGHT 8. Conclusions and recommendations The objectives of this thesis were to determine the attributes of a successful free kick in elite women's football and the mechanisms involved with successful performance of the skill. This information is potentially useful for designing training programs grounded in scientific research and identifying the fundamental coaching points for achieving a technique that produces the necessary ball flight characteristics from a direct free kick that would likely score in elite women's competition. To accomplish these objectives, the research design used was a combination of performance analysis and biomechanical methods.

In the women's football World Cup 2007, the outcome of a shot at goal direct from a free kick was significantly influenced by the location on the pitch from which it was taken, the placement of the ball within the goal, and the flight time of the shot. All seven goals scored direct from a free kick in the women's football World Cup 2007 were taken from within 7 m of the penalty circle, placed within 1 m of the goalpost and had a flight time of less than 1.24 seconds. Basing the laboratory task on quantitative information collected through the competition analysis not only allowed for the information to be collected on a realistic match situation, but also allowed for referral back to the competition data to determine if the performance in the laboratory was representative of the competition. This provides a vital link between the biomechanical analysis and its application to the competition and coaching environment. In the women's football World Cup 2007, the direct free kicks that resulted in a goal being scored or were saved by the goalkeeper had an average velocity of 20.4 ± 1.89 m.s- 1 based on the flight time and the distance from the free kick location to the goal. This is comparable with the initial linear velocity in the y-direction (towards the goal) of the curve kicks taken in the laboratory (20.3 ± 0.25 m.s- 1). It is acknowledged that calculating the velocities from television coverage is subject to errors up to 1 m.s- 1 due to the error in pitch location and flight time, and the laboratory measure was initial velocity rather than average 138

velocity. However, such a comparison gives an indication that the participants in the laboratory study were performing the laboratory kicks at a similar intensity to direct free kicks taken at the highest level of international women's football. It can therefore be assumed that the initial ball flight characteristics and kinematic data collected in the laboratory are representative of those used in a direct free kick taken in an elite women's football match.

The elite females in this study achieved greater ball velocities when performing instep kicks with accuracy constraints than those reported in the biomechanics literature for females performing maximal instep kicks (19.6 m.s- 1 for "skilled females" from a university team, (Shan, 2009)); 21.5 m.s- 1 for "elite" females although their actual level of play was not documented (Barfield et aI., 2002)). Previous research has indicated a trade-off between speed and accuracy with lower ball velocities associated with increased accuracy constraints (Lees & Nolan, 1998; Teixeira, 1999). Asami and colleagues (Asami, Togari, & Kikuchi, 1976) stated that in kicking, ball velocity drops to approximately 80% of its maximal value when an accuracy as well as a velocity demand is made on a player. In this study, the international female players achieved a mean resultant ball velocity of 22.62 m.s- 1 for an instep kick with a substantial accuracy constraint when all five kicks for each participant were considered. Therefore, it appears that the techniques of females performing maximal instep kicks reported in the literature do not represent an elite population, and more research on expert athletes is required.

Previous research investigating gender comparisons has shown that females are generally unable to produce the ball velocities of their male counterparts in kicks for maximal velocity (Barfield, et aI., 2002; Tant, et aI., 1991). The resultant linear ball velocities achieved by the female participants in this study for instep kicks were lower than those reported in the literature for maximal instep kicks performed by elite and amateur males, but slightly greater than those reported for male professionals and semi-professional aiming at a comparable target (Lees & Nolan, 2002; Whiteside, et aI., 2010). Likewise for curve kicks, the velocities achieved in this study were less than amateur males kicking with no regard for accuracy (23.0 139

- 29.4 m.s- 1) (Asai, et aI., 2004; Bray & Kerwin, 2003), but greater than semiprofessional males performing a task with comparable accuracy constraints (19.2 m.s- 1) (Whiteside, et aI., 2010). These findings suggest that, contrary to the literature, elite females are capable of achieving similar velocities as elite males when accuracy constraints indicative of those required in a match are imposed.

Comparisons of curve and instep kick resultant linear ball velocities produced conflicting results, depending on the inclusion criteria for the trials. In Chapter 6, where the focus was on describing the ball flight characteristics of successful kicks, and therefore only those kicks that hit the target were used for the comparison, no difference in resultant linear ball velocity was found between kick types. However, when comparing the kinematic differences between kick types (Chapter 7), the five most accurate curve and instep kicks were compared for each participant to prevent individual bias as it is known that individual differences exist in technique even at an elite level of play (Lees & Nolan, 2002). When an equal number of trials were used for the comparison, instep kicks had a significantly higher resultant linear ball velocity than the curve kicks. This indicates that a tradeoff between ball velocity and ball spin, as previously documented by Asai and colleagues (Asai, et aI., 2002), was evident in this study, and the reason why no difference was observed in the resultant velocities of the kicks that hit the target was because of individual player bias. For example, the participant who kicked the ball the fastest had five accurate curve kicks travelling at an average of 24.79 m's- 1 but only two accurate instep shots with a mean velocity of 25.44 m·s- 1 . This finding highlights the importance of considering the criteria for inclusion of trials in the research design.

For both kick types, the resultant ankle joint linear velocity and the knee extension angular velocity peaked at the point of ball impact, supporting the findings of Nunome and colleagues (Nunome, Lake, et aI., 2006) that decelerations of the kicking leg observed prior to impact in previous research were due to inadequate data filtering methods, and not a strategy to enhance accuracy as proposed by Teixeira (1999). Nunome and colleagues (Nunome, Lake, et aI., 2006) were the first

researchers

to

biomechanically

support

the

common

coaching 140

recommendation to kick through the ball in a maximal kick because the ankle was still linearly accelerating and the shank was still angularly accelerating until ball impact. The current study is the first to provide biomechanical support for kicking through the ball when accuracy constraints indicative of those required in a match are imposed.

Elite female football players vary their approach angle to the ball for different types of kick. The approach angle of the curve kicks was more than 20° greater than that of the instep kicks. Lees and colleagues (Lees, et aI., 2009) stated that angled approaches are used for both maximal and submaximal kicks and therefore any benefits may not be directly related to ball speed. Rather, the angled approach allows the body to tilt away from the ball, which raises the kicking leg and allows the foot to clear the ground and the knee to extend through impact (Lees, et aI., 2009). In the instep kick, the angle between the kicking leg plane and the horizontal was less, indicating an increased lean to the non-kicking side. This was likely caused by the orientation of the kicking foot at impact. In the instep kick, the kicking foot was plantarflexed, causing a lengthening of the kicking leg and therefore increased lean was necessary for the foot to clear the ground compared with the curve kicks where the kicking foot was abducted at ball impact, with little plantarflexion. The orientation of the kicking foot at impact is important for optimal foot-ball impact mechanics and players should tilt their body away from the ball to allow the kicking leg to swing through and clear the ground in preparation for ball impact.

To achieve the curved ball trajectory, synonymous with that used from a direct free kick to avoid the defensive wall and still score a goal, the elite female football players used a wider approach run to the ball, resulting in the trunk, pelvis, support foot and kicking foot to be more rotated to the right of the target at ball impact compared with the instep kicks. This facilitated the increased angle between the kicking leg plane and the target line, where the foot trajectory travelled across the face of the target, and an increased toe velocity in the medio-Iateral direction at impact for the curve kick. Therefore, the application of the force to the ball was more lateral in the curve kick, resulting in the increased lateral launch angle

141

observed in the initial ball flight characteristics and higher spin rates. In addition, the kicking foot was travelling vertically upwards at impact, leading to the increased vertical launch angle of the ball. The direction the kicking foot is travelling and its abducted orientation at impact are important for creating the optimal application of force to the ball which determines the initial ball flight characteristics. The initial flight characteristics were important predictors of the placement of the ball relative to the target centre and regression analyses demonstrated how carefully controlled the initial ball flight characteristics must be to avoid the defensive wall and hit the target. Conversely to the curve kick, the kicking foot in the instep kick was travelling towards the ground at ball impact, contributing to the increased backspin of the ball, which was also an important predictor of the ball placement relative to the target.

The position of the support foot heel relative to the ball centre was similar for both kick types. Support foot placement has been subjected to little scientific research (Lees, et aI., 2010), although Barfield (1998) did suggest that the antero-posterior placement of the support foot was dependent on whether the intent of the kick was a high or a low trajectory. The findings in this thesis showed that although the curve kicks had a significantly higher trajectory than the instep kicks, the anteroposterior and the medio-Iateral placement of the support foot relative to the ball centre did not change between kick types. Rather, it was the orientation of the support foot that differed between kick types with the instep kick support foot pointing towards the target and the curve kick support foot pointing to the right of the target. The orientation of the support foot was found to be comparable with the angle of the kicking limb plane relative to the target line and coaches should therefore advise players to point the support foot in the direction of the swing motion of the kicking limb. Visual representations of the kicking foot trajectory, such as that provided in Figure 7-2 could be useful as a coaching tool to train players to achieve these movement characteristics, and ultimately to achieve the ball flight characteristics that are necessary to score a goal direct from a free kick.

142

8.1. •

Conclusions It is possible to reliably calculate the location of a football on a pitch to within 0.24 m of its actual location using television coverage, the pitch lines that constitute the official pitch markings, and the curve-fitting method developed in this thesis.



The outcome of a direct shot at goal from a free kick in elite women's football is significantly influenced by the location on the pitch from which it is taken, the placement of the ball in the goal, and the time taken for the ball to reach the goal.



To achieve a curved ball trajectory, that would likely score from a direct free kick in elite women's football, the ball must be launched at a significantly greater vertical and lateral launch angle, with greater sidespin and rotation about the antero-posterior axis, and with a more vertical spin axis compared with a straight kick at goal from the same location.



The vertical launch angle, antero-posterior velocity and the amount of backspin are the best predictors of the vertical distance from the ball to the target centre.



The lateral launch angle and the angle of the spin axis relative to the horizontal are the best predictors of the horizontal distance from the ball to the target centre.



To achieve a curved ball trajectory, players should take a wider approach angle than that used for an instep kick. This causes the trunk, pelvis, support foot and kicking foot to be rotated to the right of the intended target at ball impact, which increases the angle of the kicking leg plane relative to the target line.



At ball impact, the kicking foot travels upwards in a curve kick to achieve the large vertical launch angle whereas in the instep kick, the kicking foot travels towards the ground to impart backspin to the ball.



Players should aim to kick through the ball for both curve and instep kicks.

143

8.2. •

Practical implications for the game Players should practise free kicks from within 7 m of the penalty area and if awarded a direct free kick in this area, teams should consider a direct shot at goal, especially those who have a free kick specialist. Depending on the individual ability of the free kick taker, free kicks may be practised and taken from areas extending up to 12 m from the penalty circle as this area still tested the defending players and goalkeeper.



The placement of the free kick within the goal is important. The ball should not be placed near the centre of the goal as this provides an easy save for the goalkeeper. Players should aim for their shot to arrive at the goal within 1 m of the goalpost for those placed in the bottom corner and within 1 m of both the goalpost and the crossbar for shots placed in the top corner. It is recommended that players taking free kicks should target the top corners of the goal as this is the most successful area for scoring goals, and even if the ball does go within the goalkeeper's reach, it is likely that it will get pushed over the crossbar and therefore the attacking team will still retain possession.



It is important for players to be aware of their individual ability and only attempt a direct shot at goal when they perceive the probability of scoring to be high. If the free kick is awarded outside their individual range for a successful shot, alternative attacking strategies should be considered such as crossing the ball or passing and building an attack in order to avoid an easy turnover of possession for the opposition.



Defensively, teams should avoid conceding free kicks in central areas close to the penalty circle. If a defender is not confident that they can make a legal challenge and win the ball, they should attempt to deviate the attacking player towards the touchline before making a challenge because a free kick conceded in a wide area is not as dangerous in terms of scoring from a direct shot at goal.

144



The initial ball flight characteristics of a direct free kick must be carefully controlled in order to enter the goal within 1 m of the goalpost and crossbar. For example, the vertical and lateral launch angles for a direct free kick awarded 20 m in front of the goal must be within a range of less than 3°. Similarly, the antero-posterior velocity must be within 2.5 m.s- 1 and the backspin rate within 4.3 revs.s- 1 in order to hit the target.



In comparison with the more commonly reported instep kick, the direct free kick is characterised by significantly greater lateral and vertical launch angles, increased sidespin and spin about the antero-posterior axis, and a more vertical spin axis. These initial ball launch conditions are important in order for the ball to avoid the defensive wall and still score a goal.



The vertical launch angle, antero-posterior velocity and the amount of backspin are the best predictors of the vertical distance from the ball to the target centre. The lateral launch angle and the angle of the spin axis relative to the horizontal are the best predictors of the horizontal distance from the ball to the target centre. This information is useful for coaching applications because if a player consistently kicks the ball high or low or continually to the left or right of the target, it is possible to identify which variables will help to rectify that error.



In comparison with an instep kick, the trajectory of the kicking leg prior to ball impact needs to be modified to achieve a curved ball trajectory. Players should approach the ball from a wider angle than that used for an instep kick, and point the support foot to the right of the intended target. This allows the trunk and pelvis to point to the right of the intended target at ball impact, which in turn facilitates the swing plane of the kicking leg being across the face of the target. In addition the kicking foot should be travelling vertically upwards at ball impact for a curve kick. These differences in the kicking foot trajectory prior to impact are important in producing the increased ball vertical and lateral ball launch angles necessary to avoid the defensive wall.

145



In an instep kick the foot should be travelling towards the ground at the instant of ball impact which allows for backspin to be applied to the ball which is important for the vertical placement of the ball relative to the target.



In a curve kick, players should contact the ball with a more abducted foot than in an instep kick. The orientation of the kicking foot is important as it determines the application of the force to the ball which in turn determines the initial flight characteristics of the ball.



Players should aim to kick through the ball for both kick types. Given the strong relationship between foot velocity at impact and the resultant ball velocity, the faster the ball foot velocity at impact the better for achieving a faster flight time and therefore providing the opposing goalkeeper with minimal time to save the ball.

A curved ball trajectory is beneficial for scoring direct from a free kick in elite women's football. The following table (Table 8-1) describes the characteristics of the instep kick technique and the coaching points on how a player should modify that technique in order to achieve a curved trajectory.

146

Table 8-1: Description of the instep kick technique performed by elite female football players and the coaching points required to modify the technique to achieve a curved ball trajectory. Instep kick characteristic

Coaching point for curve kick

Players make an angled approach to the

Take a wider approach run than for an

ball

instep kick (approximately double)

The support leg is placed approximately

Plant the heel of the support foot in the

30 cm to the side of the ball and pointing

same place as for an instep kick but

towards the intended target

point the foot approximately 25 - 30° to the right of the target

The kicking limb swing direction relative

Swing the kicking limb through across

to the target line is approximately 8° to

the face of the goal so that it swings in

the right of the target

a direction approximately 30° to the right of the target (at approximately the same angle the support foot is pointing)

The kicking foot travels towards the

The kicking foot toe should be travelling

ground when it impacts the ball to

upwards as it impacts the ball to

produce backspin

achieve a greater vertical launch angle

The kicking foot is plantarflexed at ball

Abduct the kicking foot at ball impact

impact The trunk and pelvis point slightly (15°)

Rotate the trunk and pelvis

to the right of the target at ball impact

approximately 30° to the right at ball impact

8.3.

Limitations of the study

The following limitations of the studies that comprise this thesis are acknowledged:

Determination of football pitch locations from television footage and pitch markings (Chapter 3): •

The camera views were considered to be representative of television footage, however this could not be validated as it depends on individual game footage.

147



The method does not account for non-linear systematic errors caused by lens distortions, however the lenses used for television footage would be of high quality and therefore it is considered that any effect would be minimal.

Analysis of shots at goal direct from free kicks in the women's football World Cup 2007 (Chapter 4): •

Findings relate to the games played and the players that participated in the 2007 women's football World Cup. It is possible that different findings could be observed in a different sample of games and players.



Ball locations were accurate within an average of 0.24 m and flight times within 0.05 s.

Initial ball flight characteristics of curve and instep kicks in elite women's football (Chapter 6): •

The findings relate to the sample of Nike® footballs used in this study. The laws of the game (FIFA, 200gb) state that the circumference of the ball must be between 68 and 70 cm which means the aerodynamic properties can differ between approved balls. In addition, differences in the ball panel structure and seams can affect the trajectory.



The initial ball flight characteristics identified in this study relate to a free kick taken from a distance of 20 m and directly in front of the intended target area. These may not be representative of free kicks from different distances and / or angles.



The effects of wind or wet conditions on the ball flight were not considered.



The orientation of the ball prior to the kick was not controlled. The position of the valve affects ball flight because the mass of the ball is not uniformly distributed around its centre. However, the aim of the study was to simulate a match situation as closely as possible and therefore participants were given minimal instruction on how to perform the task. No participants were observed to locate the valve on the ball before placing the ball down although this procedure has been seen in male free kick specialists.

148

Curve and instep kick kinematics in elite female footballers (Chapter 7): •

Error introduced as a result of skin movement under the markers was not quantified, however, the use of cluster marker sets reduced this error.



The model used represents body parts as linked segments which does not accurately reflect the body and its anatomical parts.



The curve kick simulated a free kick and not all participants took free kicks in games. However, all players take stationary ball kicks and curve kicks are often used for long passes and delivering balls around defenders. Therefore all participants would perform similar tasks in training and competition to those that were performed in the laboratory.



Variation in maturation and anthropometric strength measures within the sample of players may affect results. No attempt was made to standardise these measures.



The kicks may not reflect those performed in a match due to the laboratory setting, although the ecological validity of the testing environment was fully considered prior to data collection.



Results relate to a free kick taken from a distance of 20 m and directly in front of the goal. They may not be representative of free kicks from different distances and lor angles.

8.4. •

Recommendations for further research Defensively, the wall played an important role in blocking direct shots at goal from free kicks. Future research should investigate the position and number of defenders in the wall and the position and movements of the goalkeeper when successfully defending direct shots from free kicks.



The effect of the ball valve position prior to being kicked and environmental conditions should be considered to determine how they affect the kick outcome.



The flight characteristics and kinematics for free kicks from different locations on the field should be investigated.



The body orientation at ball impact significantly differed between the curve and instep kicks. Future research should consider the joint kinematics

149

throughout the entire movement in order to provide a more detailed explanation of how the athletes achieve those different body orientations at impact. •

The plane of the kicking limb provided a visual representation of the kicking foot trajectory prior to impact that is useful as a coaching tool. The kicking limb plane of other kick types should be considered.



Elite females in this study achieved greater ball velocities for instep kicks with accuracy constraints than those reported for elite females performing maximal instep kicks in the literature (Barfield, et aI., 2002). Given that previous research has shown evidence of a speed-accuracy trade-off and Asami & Nolte (1983) reported that ball velocities reduce to 80% of their maximal value when an accuracy demand is imposed on a player, it appears that a description of the elite female kicking technique has not been adequately documented in the literature and further research is required.



This sample of elite females produced ball velocities comparable to those achieved by elite males in the literature when performing similar tasks. However studies on maximal kicks have shown that females can not produce velocities as high as their male counterparts and therefore it is possible that the range for males to score direct from a free kick may extend further from the goal. A comparison of curve kick kinematics of males and females is required.



A kinetic analysis of the lower body during performance of curve and instep kicks would provide a more detailed understanding of the mechanisms involved in producing the different techniques that are used to achieve the different ball trajectories.

150

REFERENCES Abdel-Aziz, Y. I., & Karara, H. M. (1971). Direct linear transformation from comparator coordinates into object space coordinates in close-range photogrammetry. Paper presented at the ASP Symposium on Close Range Photogrammetry, Falls Church, VA Alcock, A (2010). Analysis of direct free kicks in the women's football World Cup 2007. European Journal of Sport Science, 10(4),279-284. Alcock, A, Hunter, A, & Brown, N. (2009). Determination of football pitch locations from video footage and official pitch markings. Sports Biomechanics, 8(2), 129-140. Andersson, H., Ekblom, B., & Krustrup, P. (2008). Elite football on artificial turf versus natural grass: movement patterns, technical standards and player impressions. Journal of Sports Sciences, 26(2), 113-122. Angulo, R. M., & Dapena, J. (1992). Comparison of filming and video techniques for

estimating

three-dimensional

coordinates

within

a

large

field.

International Journal of Sport Biomechanics, 8, 145-151. Apriantono, T., Nunome, H., Ikegami, Y., & Sano, S. (2006). The effect of muscle fatigue on instep kicking kinetics and kinematics in association football. Journal of Sports Sciences, 24(9), 951-960. Armatas, V., Yiannakos, A, Galazoulas, C., & Hatzimanouil, D. (2007). Goal scoring patterns over the course of a match: analysis of women's high standard soccer matches. Physical Training, 1. Armatas, V., Yiannakos, A, Papadopoulou, S., & Galazoulas, C. (2007). Analysis of the set-plays in the 18th football World Cup in Germany. Physical Training, 1. Armatas, V., Yiannakos, A, & Sileloglou, P. (2007). Relationship between time and goal scoring in soccer games: analysis of three World Cups. International Journal of Performance Analysis in Sport 7(2), 48-58. Asai, T. (2000). Comparison of curve ball kick with instep kick in football. In A J. Subic & S. J. Haake (Eds.), The Engineering of Sport. Research, Development and Innovation (pp. 487-494). London: Blackwell Science Ltd.

151

Asai, T., & Akatsuka, T. (1998). Computer simulation of curve-ball kicking in soccer. In S. J. Haake (Ed.), The Engineering of Sport (pp. 433-440). Oxford: Blackwell Science Ltd. Asai, T., Carre, M. J., Akatsuka, T., & Haake, S. J. (2002). The curve kick of a football I: impact with the foot. Sports Engineering, 5, 183-192. Asai, T., Takano, S., Carre, M. J., & Haake, S. J. (2004). A fundamental study of an infront curve kick in football. In M. Hubbard, R D. Mehta & J. M. Pallis (Eds.), The Engineering of Sport 5 (pp. 290-295). Sheffield: International Sports Engineering Association. Asami, T., & Nolte, V. (1983). Analysis of powerful ball kicking. In H. Matsui & K. Kobayashi (Eds.), Biomechanics VII/-B (pp. 695-700). Champaign, IL: Human Kinetics. Asami, T., Togari, H., & Kikuchi, T. (1976). Energy efficiency of ball kicking. In P. V. Komi (Ed.), Biomechanics V-B (pp. 135-140). Baltimore, MD: University Park Press. Baker,

R

(2006).

Gait

analysis

methods

in

rehabilitation.

Journal

of

NeuroEngineering and Rehabilitation, 3(4), doi: 10.1186/1743-0003-11831184. Ball, K. (2008). Biomechanical considerations of distance kicking in Australian Rules football. Sports Biomechanics, 7(1), 10-23. Barfield, W. R (1995). Effects of selected kinematic and kinetic variables on instep kicking with dominant and nondominant limbs. Journal of Human Movement

Studies, 29, 251-272. Barfield, W. R (1998). The biomechanics of kicking in soccer. Clinics in Sports

Medicine, 17(4),711-728. Barfield, W. R, Kirkendall, D. T., & Yu, B. (2002). Kinematic instep kicking differences between elite female and male soccer players. Journal of Sports

Science and Medicine, 1, 72-29. Barros, R M. L., Misuta, M. S., Menezes, R. P., Figueroa, P. J., Moura, F. A., Cunha, S. A., et al. (2007). Analysis of the distances covered by first division Brazilian soccer players obtained with an automatic tracking method. Journal of Sports Science and Medicine, 6,233-242.

152

Besier, T. F., Sturnieks, D. L., Alderson, J. A., & Lloyd, D. G. (2003). Repeatability of gait data using a functional hip joint centre and a mean helical knee axis.

Journal of Biomechanics, 36, 1159-1168. Beyer, W. H. (1987). GRG Standard Mathematical Tables (28th ed.). Boca Raton, FL: CRC Press Inc. Blake, H. (2009). Top ten British earners in sport. from

Retrieved 5th December, 2009,

http://www. telegraph .co.uk/sport/football/6268194/Top-ten-British-

earners-i n-sport. html Bo Kristensen, L., & Bull Andersen, T. (2009). Effect of approach velocity in soccer kicking. In T. Reilly & F. Korkusuz (Eds.), Science and Football VI (pp. 4749). Oxon: Routledge. Bobbert, M. F., Schamhardt, H. C., & Nigg, B. M. (1991). Calculation of vertical ground reaction force estimates during running from positional data. Journal of Biomechanics, 24(12), 1095-1105. Bradley, P., O'Donoghue, P., Wooster, B., & Tordoff, P. (2007). The reliability of ProZone MatchViewer: a video-based technical performance analysis system. International Journal of Performance Analysis in Sport, 7(3), 117129. Bray, K., & Kerwin, G. (2003). Modelling the flight of a soccer ball in a direct free kick. Journal of Sports Sciences, 21, 75-85. Brewin, M. A., & Kerwin, D. G. (2003). Accuracy of scaling and DL T reconstruction techniques for planar motion analyses. Journal of Applied Biomechanics, 19,79-88. Browder, K. D., Tant, C. L., & Wilkerson, J. D. (1991). A three dimensional kinematic analysis of three kicking techniques in female soccer players. In C. L. Tant, P. E. Patterson & S. L. York (Eds.), Proceedings of the Ninth

ISBS Symposium (pp. 95-100). Iowa: ISBS. Brown, E. W., Wilson, D. J., Mason, B. R., & Baker, J. (1993). Three dimensional kinematics of the direct free kick in soccer when opposed by a defensive wall. In J. Hamill, T. R. Derrick & E. H. Elliott (Eds.), Biomechanics in Sport XI (pp. 334-338). Massachusetts: ISBS.

Bull Andersen, T., D6rge, H. C., & Thomsen, F. I. (1999). Collisions in soccer kicking. Sports Engineering, 2, 121-125. 153

Bunc, V., & Psotta, R. (2004). Functional characteristics of elite Czech female soccer players. Journal of Sports Sciences, 22, 528. Cabri, J., De Prott, E., Dufour, W., & Clarys, J. P. (1988). The relation between muscular strength and kick performance In T. Reilly, A Lees, K. Davids & W. J. Murphy (Eds.), Science and Football (pp. 186-193). London: E & FN Spon. Campbell, A. C., Alderson, J. A, Lloyd, D. G., & Elliott, B. C. (2009). Effects of different technical coordinate system definitions on the three dimensional representation of the glenohumeral joint centre. Medical and Biological

Engineering and Computing, 47,543-550. Cappozzo, A, Catani, F., Della Croce, U., & Leardini, A (1995). Position and orientation in space of bones during movement: anatomical frame definition and determination. Clinical Biomechanics, 10(4), 171-178. Cappozzo, A, Catani, F., Leardini, A, Benedeti, M. G., & Della Croce, U. (1996). Position and orientation in space of bones during movement: experimental artefacts. Clinical Biomechanics, 11, 90-100. Cappozzo, A., Della Croce, U., Leardini, A, & Chiari, L. (2005). Human movement analysis using stereophotogrammetry. Part 1: theoretical background. Gait

and Posture, 21, 186-196. Carey, D. P., Smith, G., Smith, D. T., Shepherd, J. W., Skriver, J., Ord, L., et al. (2001). Footedness in world soccer: an analysis of France '98. Journal of

Sports Sciences, 19( 11), 855-864. Carling, C., Bloomfield, J., Nelsen, L., & Reilly, T. (2008). The role of motion analysis in elite soccer. Sports Medicine, 38(10), 839-862. Carling, C., Williams, A. M., & Reilly, T. (2005). Handbook for soccer match

analysis. A systematic approach to improving performance. London: Routledge. Carre, M. J., Asai, T., Akatsuka, T., & Haake, S. J. (2002). The curve kick of a football II: flight through the air. Sports Engineering, 5, 193-200. Chen, L., Armstrong, C. W., & Rattopoulos, D. D. (1994). An investigation on the accuracy of three-dimensional space reconstruction using the direct linear transformation technique. Journal of Biomechanics, 27(4),493-500.

154

Chin, A. (2009). Anatomical and functional based upper limb models: methods for kinematic

analysis

of

cricket

spin

bowling.

Unpublished

Doctoral

Dissertation. University of Western Australia. Chin, A., Elliott, B., Alderson, J., Lloyd, D., & Foster, D. (2009). The off-break and "doosra": Kinematic variations of elite and sub-elite bowlers in creating ball spin in cricket bowling. Sports Biomechanics, 8(3), 187-198. Clagg, S. E., Warnock, A., & Thomas, J. S. (2009). Kinetic analysis of maximal effort soccer kicks in female collegiate athletes. Sports Biomechanics, 8, 141-153. Coleman, S., & Anderson, D. (2007). An examination of the planar nature of golf club motion in the swings of experienced players. Journal of Sports

Sciences, 25(7), 739-748. Coutts, A. J., & Duffield, R. (2010). Validity and reliability of GPS devices for measuring movement demands of team sports. Journal of Science and

Medicine in Sport, 13,133-135. Craig, J. J. (2004). Introduction to robotics: mechanics and control (3rd ed.). Englewood Cliffs, NJ: Prentice Hall. Davids, K., Lees, A., & Burwitz, L. (2000). Understanding and measuring coordination and control in kicking skills in soccer: implications for talent identification and skill acquisition. Journal of Sports Sciences, 18, 703-714. Davis, J. A., & Brewer, J. (1992). Physiological characteristics of an international female soccer squad. Journal of Sports Sciences, 10(2), 142-143. Davis, J. A., & Brewer, J. (1993). Applied physiology of female soccer players.

Sports Medicine, 16(3), 180-189. De Leva, P. (1996). Adjustments to Zatsiorsky-Seluyanov's segment inertia parameters. Journal of Biomechanics, 29(9), 1223-1230. De Pratt, E., Cabri, J., Dufour, W., & Clarys, J. P. (1988). Strength training and kick performance in soccer. In T. Reilly, A. Lees, K. Davids & W. J. Murphy (Eds.), Science and Football (pp. 108-113). New York: E & FN Spon. Della Croce, U., Camomilla, V., Leardini, A., & Cappozzo, A. (2003). Femoral anatomical frame: assessment of various definitions. Medical Engineering

and Physics, 25,425-431.

155

Della Croce, U., Leardini, A, Chiari, L., & Cappozzo, A (2005). Human movement analysis using stereophotogrammetry. Part 4: assessment of anatomical landmark misplacement and its effect on joint kinematics. Gait and Posture, 21,226-237. DeWitt, J. K. (2002). Biomechanics of kicking. Performance Conditioning Soccer, 8(7),4-8. DeWitt, J. K., & Hinrichs, R N. (2002). Factors related to ball velocity during an

instep

soccer kick.

Paper

presented

at

the

World

Congress

of

Biomechanics, Calgary, Alberta. Di Salvo, V., Baron, R, Tschan, H., Calderon Montero, F. J., Bachl, N., & Pigozzi, F. (2007). Performance characteristics according to playing position in elite soccer. International Journal of Sports Medicine, 28,222-227. Di Salvo, V., Collins, A, McNeill, B., & Cardinale, M. (2006). Validation of Prozone: A new video-based performance analysis system. International Journal of

Performance Analysis in Sport, 6( 1), 108-119. Di Salvo, V., Gregson, W., Atkinson, G., Tordoff, P., & Drust, B. (2009). Analysis of high intensity activity in Premier League soccer. International Journal of

Sports Medicine, 30,205-212. D6rge, H. C., Bull Andersen, T., S0rensen, H., & Simonsen, E. B. (2002). Biomechanical differences in soccer kicking with the preferred and nonpreferred leg. Journal of Sports Sciences, 20, 293-299. Duffield, R, Reid, M., Baker, J., & Spratford, W. (2010). Accuracy and reliability of GPS devices for measurement of movement patterns in confined spaces for court-based sports. Journal of Science and Medicine in Sport, 13, 523-525. Dutta, P., & Subramanium, S. (2002). Effect of six weeks of isokinetic strength training combined with skill training on football kicking performance. In W. Spinks, R. Bower, A. Murphy, T. Reilly & R Smith (Eds.), Science and

Football IV (pp. 333-340). New York: Routledge. Egan, C. D., Verheul, M. H., & Savelsbergh, G. J. P. (2007). Effects of experience in the coordination of internally and externally timed soccer kicks. Journal of

Motor Behaviour, 39(5), 423-432.

156

Elliott, B., & Alderson, J. (2007). Laboratory versus field testing in cricket bowling: A review of current and past practice in modelling techniques. Sports Biomechanics, 6(1), 99-108.

Engelhorn, R. (1997). Speed and accuracy in the learning of a complex motor skill. Perceptual and Motor Skills, 85, 1011-1017.

Ensum, J., Williams, M., & Grant, A. (2000). Analysis of attacking set plays in Euro 2000. Insight, 4, 36-39.

FIFA. (1991). FIFA Women's World Cup China '91 - Technical report and statistics.

Retrieved

2nd

2009,

September,

from

http://www.fifa.com/mm/document/afdeveloping/technicaldev p/SOlO8/19/ww c 91 tr part2 260.pdf FIFA. (2006). Report and Statistics. 2006 FIFA World Cup Germany.

Retrieved

June 2007, from www.fifa.com FIFA. (2007). Report and Statistics. FIFA Women's World Cup China 2007. Retrieved May 2007, from www.FIFA.com FIFA. (2008). Technical Study Group: Technical Report List Women's Compo Retrieved

2

September,

2009,

from

http://www.fifa.com/aboutfifa/developing/technicaldevelopmentltechnicalwo menreport. html FIFA. (2009a). FIFA Quality concept for football turf.

Retrieved 21st October from

2009,

http://www.fifa.com/mm/documentlafdeveloping/pitchequip/fqc football turf folder 342.pdf FIFA. (2009b). Laws of the game 2009/10.

Retrieved 20 October 2009, from

http://www.fifa.com/mm/docu ment/affederation/federation/81/42/36/1awsofth

egameen.pdf FIFA. (2010a, 1Sth April 2010). 80 things you need to know about the FIFA World Cup.

Retrieved

7th

August,

2010,

from

http://www.fifa.com/mm/documentlfifafacts/mencompwc/S1/97/89/statisticai kitfwc201 O. 80th ingsyou needtoknowabouUhefifa. pdf FIFA. (2010b). FIFA's growth since 1904.

Retrieved 7th May, 2010, from

http://www.fifa.com/mm/documentififafacts/organisation/S2/00/16/fs100 03a growth.pdf 1S7

FIFA (2010c). FIFAICoca-Cola Women's World Ranking.

Retrieved 7th May,

2010,

from

http://www.fifa.com/woridfootbaillranking/iastranking/gender=f/fullranking.ht ml Franks, I. M. (1997). Use of feedback by coaches and players. In T. Reilly, J. Bangsbo & M. Hughes (Eds.), Science and Football 11/ (pp. 267-278). London: E & FN Spon. Franks, I. M. (2004). The need for feedback. In M. Hughes & I. M. Franks (Eds.), Notational analysis of sport: systems for better coaching and performance in sport (2nd ed., pp. 8-16). London: Routledge. Fuller, J., Liu, L. J., Murphy, M. C., & Mann, R. W. (1997). A comparison of lowerextremity skeletal kinematics measured using skin- and pin-mounted markers. Human Movement Science, 16,219-242. Georgakis, A, Stergioulas, L., & Giakas, G. (2002). Automatic algorithm for filtering kinematic signals with impacts in the wigner representation. Medical and Biological Engineering and Computing, 40, 625-633. George, K., Batterham, A., & Sullivan, I. (2003). Validity in clinical research: a review of basic concepts and definitions. Physical Therapy in Sport, 4, 115121. Grant, A, Reilly, T., Williams, M., & Borrie, A (1998). Analysis of the goals scored in the 1998 World Cup. Insight, 1(2), 18-20. Grant, A, & Williams, M. (1998a). Analysis of corner kicks. Insight, 1(3),25-26. Grant, A, & Williams, M. (1998b). An analysis of penalty kicks before and after the recent rule change. Insight, 1(2),21. Grant, A, & Williams, M. (1999). Analysis of the final 20 matches played by Manchester United in the 1998-99 season. Insight, 3(1),42-45. Grant, A, Williams, M., Reilly, T., & Borrie, A (1998). Analysis of the successful and unsuccessful teams in the 1998 World Cup. Insight, 1(2), 21-24. Grehaigne, J. F., Mahut, B., & Fernandez, A (2001). Qualitative observation tools to analyse soccer. International Journal of Performance Analysis in Sport, 1,52-61.

158

Griffiths, I., Evans, C., & Griffiths, N. (2005). Tracking the flight of a spinning football in three dimensions. Measurement Science and Technology, 16, 2056-2065. Hay, J. G. (1993). The biomechanics of sports techniques (IV ed.). New Jersey: Prentice-Hall. Helgerud, J., Hoff, J., & Wisl0ff, U. (2002). Gender differences in strength and endurance of elite soccer players. In W. Spinks, T. Reilly & A. Murphy (Eds.), Science and Football IV (pp. 382-383). London: Routledge. Hennig, E. M., & Sterzing, T. (2010). The influence of soccer shoe design on playing performance: a series of biomechanical studies. Footwear Science, 2(1),3-11. Hennig, E. M., & Zulbeck, O. (1999). The influence of soccer boot construction on ball velocity and shock to the body. Paper presented at the 4th Symposium

on Footwear Biomechanics, Canmore, Alberta, Canada. Hewitt, A., Withers, R., & Lyons, K. (2007). Match analyses of Australian international women soccer players using an athlete tracking device. Journal of Sports Science and Medicine, 6(S10), 107.

Hinrichs, R. N., Morrison, B., Vint, P. F., DeWitt, J. K., Mitchell, J., & McLean, S. P. (2004). Predicting out-of-plane point locations using the 20-0L T. Paper

presented at the 28th Annual Meeting of the American Society of Biomechanics, Portland, OR. Horn, R., Williams, M., & Grant, A. (2000). Analysis of France in World Cup 1998 and Euro 2000. Insight, 4(1),40-43. Hughes, M. (1988). Computerised notation analysis in field games. Ergonomics, 31, 1585-1592. Hughes, M. (1991). Notation analysis in football. In T. Reilly, J. Clarys & A. Stibbe (Eds.), Science and Football" (pp. 151-159). London: E & F Spon. Hughes, M., & Bartlett, R. (2002). Performance analysis. Journal of Sports Sciences, 20,735-737.

Hughes, M., Cooper, S.-M., & Nevill, A. (2002). Analysis procedures for nonparametric data from performance analysis. International Journal of Performance Analysis in Sport, 2(1),6-20.

159

Hughes, M., & Franks, I. (1997). Notational Analysis of Sport. London: E & FN Spon. Hughes, M., Hughes, M. T., & Behan, H. (2007). The evolution of computerised notational analysis through the example of racket sports. International

Journal of Sports Science and Engineering, 1(1), 3-28. Hughes, M., & Wells, J. (2002). Analysis of penalties taken in shoot-outs.

International Journal of Performance Analysis in Sport, 2, 55-72. Inoue, S., Ito, T., Sueyoshi, Y., Q'Donoghue, R. K., & Mochinaga, M. (2000). The effects of lifting the rotational axis on swing speed of the instep kick in soccer. In Y. Hong & D. P. Johns (Eds.), Proceedings of the XVIII

International Symposium on Biomechanics in Sports (pp. 39-42). Hong Kong: The Chinese University of Hong Kong. Iso kawa , M., & Lees, A (1988). A biomechanical analysis of the instep kick motion in soccer. In T. Reilly, A Lees, K. Davids & W. J. Murphy (Eds.), Science

and Football (pp. 449-455). London: E & FN Spon. James, N. (2006). Notational analysis in soccer: past, present and future.

International Journal of Performance Analysis in Sport, 6,67-81. James, N., Mellalieu, S. D., & Hollely, C. (2002). Analysis of strategies in soccer as a function of European and domestic competition. International Journal of Performance Analysis in Sport, 2, 85-103. James, N., Taylor, J., & Stanley, S. (2007). Reliabillity procedures for categorical data in performance analysis. International Journal of Performance Analysis

in Sport, 7, 1-11. Jinji, T., & Sakurai, S. (2006). Direction of spin axis and spin rate of the pitched baseball. Sports Biomechanics, 5, 197-214. Jones, P. D., James, N., & Mellalieu, S. D. (2004). Possession as a performance indicator in soccer. International Journal of Performance Analysis in Sport, 4,98-102. Katis, A, & Kellis, E. (2010). Three-dimensional kinematics and ground reaction forces during the instep and outstep soccer kicks in pubertal players.

Journal of Sports Sciences, 28(11),1233-1241.

160

Kawamoto, R., Miyagi, 0., & Fukashiro, S. (2007). Kinetic comparison of a sidefoot soccer kick between experienced and inexperienced players. Sports Biomechanics, 6(2), 187-198.

Kellis, E., & Katis, A (2007). Biomechanical characteristics and determinants of instep soccer kick. Journal of Sport Science and Medicine, 6, 154-165. Kellis, E., Katis, A, & Gissis, I. (2004). Knee biomechanics of the support leg in soccer kicks from three angles of approach. Medicine and Science in Sports and Exercise, 36(6),1017-1028.

Kellis, E., Katis, A, & Vrabas, I. S. (2006). Effects of an intermittent exercise fatigue protocol on biomechanics of soccer kick performance. Scandinavian Journal of Medicine and Science in Sports, 16, 334-344.

Kerwin, D. G., & Bray, K. (2006). Measuring and modelling the goalkeeper's diving envelope in a penalty kick. In E. F. Moritz & S. Haake (Eds.), Engineering of Sport 6. Volume 1: Developments for Sports (pp. 321-326). New York:

Springer Science+Business Media. Knudson, D., & Bahamonde, R. (2001). Effect of endpoint conditions on position and velocity near impact in tennis. Journal of Sports Sciences, 19,839-844. Konstadinidou, X., & Tsigilis, N. (2005). Offensive playing profiles of football teams from the 1999 Women's World Cup Finals. International Journal of Performance Analysis in Sport, 5,61-71.

Kreighbaum, E., & Barthels, K. M. (1996). Biomechanics: A qualitative approach for studying human movement. Boston, MA: Allyn & Bacon.

Kreighbaum, E., & Hunt, W. A

(1978). Relative factors influencing pitched

baseballs. Biomechanics of Sports and Kinanthropometry, 227-236. Krustrup, P., & Bangsbo, J. (2001). Physiological demands of top-class soccer refereeing in relation to physical capacity: effect of intense intermittent exercise training. Journal of Sports Sciences, 19(11),881-891. Kwon, Y. (1999). Object plane deformation due to refraction in two-dimensional underwater motion analysis. Journal of Applied Biomechanics, 15, 396-403. Kwon, Y., & Casebolt, J. B. (2006). Effects of light refraction on the accuracy of camera calibration and reconstruction in underwater motion analysis. Sports Biomechanics, 5(2), 315-340.

161

Lees, A (2002). Technique anlaysis in sports: a criticial review. Journal of Sports

Sciences, 20, 813-828. Lees, A (2007). Qualitative biomechanical assessment of performance. In M. Hughes & I. Franks (Eds.), The Essentials of Performance Analysis (pp.

162-179). London: Routledge. Lees, A, Andersen, T. B., Nunome, H., & Sterzing, T. (2010). The biomechanics of kicking in soccer: a review. Journal of Sports Sciences, 28(8), 805-817. Lees, A, & Davids, K. (2002). Coordination and control of kicking in soccer. In K. Davids, G. Savelsbergh, S. J. Bennett & J. V. d. Kamp (Eds.), Interceptive

Actions in Sport (pp. 273-287). London: Routledge. Lees, A, & Davies, T. (1987). The effects of fatigue on soccer kick kinematics.

Journal of Sports Sciences, 6, 156-157. Lees, A, Kershaw, L., & Moura, F. (2004). The three-dimensional nature of the maximal instep kick in soccer. Journal of Sports Sciences, 22, 493-494. Lees, A, & Nolan, L. (1998). The biomechanics of soccer: a review. Journal of

Sports Sciences, 16,211-234. Lees, A, & Nolan, L. (2002). Three-dimensional kinematic analysis of the instep kick under speed and accuracy conditions. In W. Spinks, T. Reilly & A Murphy (Eds.), Science and Football IV (pp. 16-21). London: Routledge. Lees, A, Steward, I., Rahnama, N., & Barton, G. (2009). Understanding lower limb function in the performance of the maximal instep kick in soccer. In T. Reilly

& G. Atkinson (Eds.), Proceedings of the 6th International Conference on Sport, Leisure and Ergonomics (pp. 149-160). London: Routledge. Levanon, J., & Dapena, J. (1998). Comparison of the kinematics of the full-instep and pass kicks in soccer. Medicine and Science in Sports and Exercise,

30(6),917-927. Liebermann, D. G., Katz, L., Hughes, M. D., Bartlett, R. M., McClements, J., & Franks, I. M. (2002). Advances in the application of information technology to sport performance. Journal of Sports Sciences, 20(10), 755-769. Lloyd, D. G., Alderson, J., & Elliott, B. C. (2000). An upper limb kinematic model for the examination of cricket bowling: A case study of Mutiah Muralitharan.

Journal of Sports Sciences, 18, 975-982.

162

L6pez-Botella, M., & Palao, J. M. (2007). Relationship between laterality of foot strike and shot zone on penalty efficacy in specialist penalty takers.

International Journal of Petiormance Analysis in Sport, 7(3), 26-36. Luhtanen, P. (1994). Biomechanical aspects. In B. Ekblom (Ed.), Handbook of

Sports Medicine and Science Football (Soccer) (pp. 59-77). Oxford: Blackwell Scientific Publications. Luhtanen, P., Belinskij, A., Hayrinen, M., & Vanttinen, T. (2001). A comparative tournament analysis between the Euro 1996 and 2000 in soccer.

International Journal of Petiormance Analysis in Sport, 1, 74-82. Manolopoulos, E., Papadopoulos, C., & Kellis, E. (2006). Effects of combined strength and kick coordination training on soccer kick biomechanics in amateur players. Scandinavian Journal of Medicine and Science in Sports, 16,102-110. Manolopoulos, E., Papadopoulos, C., Salonikidis, Katartzi, E., & Poluha, S. (2004). Strength training effects on physical conditioning and instep kick kinematics in young amateur soccer players during preseason. Perceptual and Motor

Skills, 99, 701-710. Marcolin, G., Petrone, N., & Robazza, C. (2006). Correlation between support foot placement and goal accuracy for instep kicks in the soccer field. In E. F. Moritz & S. Haake (Eds.), The Engineering of Sport 6. Volume 1:

Developments

for

Sports

(pp.

345-350).

New

York:

Springer

Science+Business Media. Marieb, E. N. (2001). Human Anatomy & Physiology (5th ed.). San Francisco: Benjamin Cummings. Maslovat, D., & Franks, I. M. (2008). The need for feedback. In M. Hughes & I. M. Franks (Eds.), The essentials of petiormance analysis: an introduction (pp. 1-7). Routledge: Oxon. McLean, B. D., & Tumilty, D. (1993). Left-right assymetry in two types of soccer kick. British Journal of Sports Medicine, 27(4),260-262. McLean, S. P., Vint, P. F., Hinrichs, R. N., DeWitt, J. K., Morrison, B., & Mitchell, J. (2004).

Factors affecting the accuracy of 2D-DL T calibration. Paper

presented at the 28th Annual Meeting of the American Society of Biomechanics. 163

Mohr, M., Ellingsgaard, H., Andersson, H., Bangsbo, J., & Krustrup, P. (2004). Physical demands in high-level female soccer: application of fitness tests to evaluate match performance. Journal of Sport Sciences, 22, 552-553. Morya, E., Bigatao, H., Lees, A., & Ranvaud, R. (2004). Evolving penalty kick strategies: World Cup and club matches, 2000-2002. Journal of Sports

Sciences, 22(6), 512-513. Mujika, I., Santisteban, J., Impellizzeri, F. M., & Castagna, C. (2009). Fitness determinants of success in men's and women's football. Journal of Sports

Sciences, 27(2),107-114. Neilson, P., & Jones, R. (2004). Dynamic soccer ball performance measurement.

Journal of Sports Sciences, 22(6),497. Neilson, P., Jones, R., Kerr, D., & Sumpter, C. (2004). An image recognition system

for

the

measurement

of

soccer

ball

spin

characteristics.

Measurement Science and Technology, 15,2239-2247. Nunome, H., Asai, T., Ikegami, Y., & Sakurai, S. (2002). Three-dimensional kinetic analysis of side-foot and instep soccer kicks. Medicine and Science in

Sports and Exercise, 34(12), 2028-2036. Nunome, H., Ikegami, Y., Apriantono, T., & Sano, S. (2004). Kinetic comparison of instep soccer kick between preferred and non-preferred leg in highly skilled players. In M. Lamontagne, D. Gordon, E. Robertson & H. Sveistrup (Eds.),

Proceedings of XXII International Symposium of Biomechanics in Sports (pp. 265-271). Canada: University of Ottawa. Nunome, H., Ikegami, Y., Asai, T., & Sato, Y. (2000). A kinetic comparison of the inside soccer kick between high performance players and junior players. In Y. Houng & D. P. Johns (Eds.), XVII' International Symposium on

Biomechanics in Sports (Vol. II, pp. 751-754): The Chinese University of Hong Kong. Nunome, H., Ikegami, Y., Kozakai, R., Apriantono, T., & Sano, S. (2006). Segmental dynamics of soccer instep kicking with the preferred and nonpreferred leg. Journal of Sports Sciences, 24(5), 529-241. Nunome, H., Lake, M., Georgakis, A., & Stergioulas, L. (2006). Impact phase kinematics of instep kicking in soccer. Journal of Sports Sciences, 24(1), 11-22. 164

O'Donoghue, P. (2007). Reliability issues in performance analysis. International

Journal of Performance Analysis in Sport, 7, 35-48. O'Donoghue, P., & Robinson, G. (2009). Validity of the ProZone3® player tracking system: A preliminary report. International Journal of Computer Science in

Sport, 8(1),37-53. O'Donoghue, P. G. (2002). Time-motion analysis of work-rate in English FA Premier League soccer. International Journal of Performance Analysis in

Sport, 2(1), 36-43. Opavsky, P. (1988). An investigation of linear and angular kinematics of the leg during two types of soccer kick. In T. Reilly, A. Lees, K. Davids & W. J. Murphy (Eds.), Science and Football (pp. 456-459). London: E & FN Spon. Orloff, H., Sumida, B., Chow, J. Y., Habibi, L., Fujino, A., & Kramer, B. (2008). Ground reaction forces and kinematics of plant leg position during instep kicking in male and female collegiate soccer players. Sports Biomechanics, 7(2),238-247. Passmore, M. A., Tuplin, S., Spencer, A., & Jones, R (2008). Experimental studies of the aerodynamics of spinning and stationary footballs. Journal of

Mechanical Engineering Science, 222, 195-205. Plagenhoef, S. (1971). Patterns of Human Motion. Englewood Cliffs, NJ: PrenticeHall. Portus, M. (2006). Technique effects on low back pain in cricket bowlers. Unpublished Doctoral Dissertation. University of Western Australia. Prassas, S. G., Terauds, J., & Nathan, T. A. (1990). Three-dimensional kinematic analysis of high and low trajectory kicks in soccer. In M. Nosek, D. Sojka, W. E. Morrison & P. Susanka (Eds.), Proceedings of Vlllth International

Symposium of Biomechanics in Sports (pp. 145-149). Prague: Conex Company. Price, D. S., Neilson, P. J., Harland, A. R, & Jones, R (2004). Measurement of out-of-balance forces in soccer balls. Journal of Sports Sciences, 22(6), 497-498. Putnam, C. A. (1991). A segment interaction analysis of proximal-to-distal sequential segment motion patterns. Medicine and Science in Sports and

Exercise, 23, 130-144. 165

Randers, M. B., Mujika, I., Hewitt, A, Santisteban, J., Bischoff, R., Solano, R., et al. (2010). Application of four different football match analysis systems: A comparative study. Journal of Sports Sciences, 28(2), 171-182. Reid, M. (2009). Loading and velocity generation in the high performance tennis serve. Unpublished Doctoral Dissertation. University of Western Australia. Reid, M., Elliott, B., & Alderson, J. (2008). Lower-limb coordination and shoulder joint mechanics in the tennis serve. Medicine and Science in Sports and Exercise, 40(2), 308-315. Reilly, T. (2001). Assessment of sports performance with particular reference to field games. European Journal of Sport Science, 1(3),1-12. Reilly, T., Bangsbo, J., & Franks, A (2000). Anthropometric and physiological predispositions for elite soccer. Journal of Sports Sciences, 18, 669-683. Reilly, T., & Thomas, V. (1976). A motion analysis of work-rate in different positional roles in professional football match-play. Journal of Human Movement Studies, 2, 87-97. Richards, J. G. (1999). The measurement of human motion: A comparison of commercially available systems. Human Movement Science, 18, 589-602. Rodano, R., & Tavana, R. (1993). Three-dimensional analysis of instep kick in professional soccer players. In T. Reilly, J. Clarys & A Stibbe (Eds.), Science and Football" (pp. 357-361). London: E & F N Spon. Sakurai, S., Jinji, T., Reid, M., Cuitenho, C., & Elliott, B. (2007). Direction of spin axis and spin rate of the ball in tennis serve. Journal of Biomechanics, 40(Suppl 2), S197. Schache, A G., Baker, R., & Lamoreux, L. W. (2008). Influence of thigh cluster configuration on the estimation of hip axial rotation. Gait and Posture, 27, 60-69. Schmitt, S. R. (2009). Center and radius of a sphere from four points.

Retrieved

2nd July, 2009, from http://home.att.net/-srschmittlsphere4pts.htmi Scott, D. (1999). Analysis of all goals scored during the 1999 Women's World Cup Finals. Insight, 3(1), 52-54. Scoulding, A, James, N., & Taylor, J. (2004). Passing in the soccer World Cup 2002. International Journal of Performance Analysis in Sport, 4, 36-41.

166

Scurr, J., & Hall, B. (2009). The effects of approach angle on penalty kicking accuracy and kick kinematics with recreational soccer players. Journal of

Sporls Science and Medicine, 2,230-234. Seabra, F., & Dantas, L. E. P. B. T. (2006). Space definition for match analysis in soccer. International Journal of Performance Analysis in Sporl, 6, 97-113. Setterwall, D. (2003). Computerised video analysis of football - technical and commercial possibilities for football coaching. Unpublished Masters Thesis. Stockholms Universitet. Shan, G. (2009). Influence of gender and experience on the maximal instep soccer kick. European Journal of Sporl Science, 9(2), 107-114. Shan, G., Daniels, D., Wang, C., Wutzke, C., & Lemire, G. (2005). Biomechanical analysis of maximal instep kick by female soccer players. Journal of Human

Movement Studies, 49, 149-168. Shan, G., & Westerhoff, P. (2005). Full-body kinematic characteristics of the maximal instep soccer kick by male soccer players and parameters related to kick quality. Sporls Biomechanics, 4, 5-72. Shinkai, H., Nunome, H., Ikegami, Y., & Isokawa, M. (2009). Ball-foot interaction in impact phase of instep soccer kicking. In T. Reilly & F. Korkusuz (Eds.),

Science and Football VI (pp. 41-46). Oxon: Routledge. Silva, W. P., & Silva, C. M. D. P. S. (2006). LAB Fit Curve Fitting Software (Nonlinear Regression and Treatment of Data Program).

V 7.2.34.

Retrieved 2008-01-15, 2008, from www.labfit.net Smith, C. E. M. (2007). Application of a three-dimensional kinematic model of the lofted instep kick to technique development in soccer players. Unpublished Doctoral Dissertation. Southern Cross University. Sterzing, T., & Hennig, E. M. (2008). The influence of soccer shoes on kicking velocity in full-instep kicks. Exercise and Sporls Science Review, 36(2), 9197. Sterzing, T., Kroiher, J., & Hennig, E. M. (2009). Barefoot kicking superior to shod kicking? In T.Reilly & F.Korkusuz (Eds.), Science and Football VI (pp. 5056). Oxon: Routledge. Stoner, L. J., & Ben-Sira, D. (1981). Variation in movement patterns of professional soccer players when executing a long range and a medium 167

range in-step soccer kick. In A. Morecki, K. Fidelus, K. Kedzior & A. Wit (Eds.), Biomechanics VI/-B (pp. 337-341). Baltimore, MD: University Park Press. Ta"lana, E., Grehaigne, J. F., & Cometti, G. (1993). The influence of maximal strength training of lower limbs of soccer players on their physical and kick performances. In T. Reilly, J. Clarys & A. Stibbe (Eds.), Science and

Football /I (pp. 98-103). London: E & FN Spon. Tant, C. L., Browder, K. D., & Wilkerson, J. D. (1991). A three dimensional kinematic comparison of kicking techniques between male and female soccer players. In C. L. Tant, P. E. Patterson & S. L. York (Eds.),

Proceedings ofthe Ninth ISBS Symposium (pp. 101-105). Iowa: ISBS. Teixeira, L. A. (1999). Kinematics of kicking as a function of different sources of constraint on accuracy. Perceptual and Motor Skills, 88, 785-789. Toki, S., & Sakurai, S. (2005). Quantitative match analysis of soccer games with two dimensional DLT procedures XXth Congress of the International

Society of Biomechanics (pp. 911). Clevland, Ohio: International Society of Biomechanics. Trolle, M., Aagaard, P., Simonsen, J., Bangsbo, J., & Klaysen, K. (1993). Effects of strength training on kicking performance in soccer. In T. Reilly, J. Clarys & A. Stibbe (Eds.), Science and Football /I (pp. 95-98). London: E & FN Spon. Tsaousidis, N., & Zatsiorsky, V. (1996). Two types of ball-effector interaction and their relative contribution to soccer kicking. Human Movement Science, 15, 861-876. Tumilty, D. (1993). Physiological characteristics of elite soccer players. Sports

Medicine, 16(2), 80-96. Tumilty, D. (2000). Protocols for the physiological assessment of male and female soccer players. In C. J. Gore (Ed.), Physiological Tests for Elite Athletes (3rd ed., pp. 356-362). Champaign, IL: Human Kinetics. Walton, J. S. (1981). Close-range cine-photogrammetry: A generalized technique for quantifying gross human motion. Unpublished doctoral dissertation. Pennsylvania State University, USA.

168

Wang, J., & Wiese-Bjornstal, D. M. (1994). Mechanical and anatomical analysis of the soccer instep shot. Strength and Conditioning, Dec, 34-38. Whiteside, D., Alderson, J., & Elliott, B. (2010). 'Bend it like Beckham': Ball rotation in the curved football kick. In R. Jensen, W. Ebben, E. Petushek, C. Richter & K. Roemer (Eds.), 28th International Society of Sports Biomechanics

Conference (pp. 297-300). Michigan, USA: International Society of Sports Biomechanics. Whiteside, D., Middleton, K., & Chin, A. (submitted). Talking a lot of spin: The validation of a three-dimensional ball rotation model. Manuscript submitted

for publication. Wickstrom, R. L. (1975). Developmental kinesiology. Exercise & Sport Sciences

Reviews, 3, 163-192. Wickstrom, R. L. (1983). Fundamental Motor Patterns (3rd ed.). Philadelphia: Lea & Febiger.

Williams, A. M., & Hodges, N. J. (2005). Practice, instruction and skill acquisition in soccer: challenging tradition. Journal of Sports Sciences, 23(6), 637-650. Williams, J., Hughes, M., Q'Donoghue, P., & Davies, G. (2007). A reliability study of a real-time and lapsed-time application for rugby. International Journal of

Performance Analysis in Sport, 7(1), 80-86. Williams, M., Lee, D., & Reilly, T. (1999). A quantitative analysis of matches played in the 1991-92 and 1997-98 seasons. The Football Association, 128. Willmott, A., & Dapena, J. (2008). Quantifying the planarity of the field hockey hit. Paper presented at the North American Congress of Biomechanics, Ann Arbor. Wilson, S. (2010). A decade of performance analysis development.

Retrieved 4th

August 2010, 2010, from http://www.prozonesports.com/news-article-adecade-of-performance-analysis-development.html Windolf, M., G6tzen, N., & Morlock, M. (2008). Systematic accuracy and precision analysis of video motion capturing systems - exemplified on the Vicon-460 system. Journal of Biomechanics, 41,2776-2780. Winter, D. A. (2005). Biomechanics and Motor Control of Human Movement. New Jersey: John Wiley & Sons, Inc. 169

Wisl0ff, U., Helgerud, J., & Hoff, J. (1998). Strength and endurance of elite soccer players. Medicine and Science in Sports and Exercise, 30(3),462-467. Woltring, H. J. (1986). A Fortran package for generalized, cross-validatory spline smoothing and differentiation. Advanced Engineering Software, 8(2), 104113. Wood, G., & Marshall, R N. (1986). The accuracy of DLT extrapolation in threedimensional film analysis. Journal of Biomechanics, 19(9),781-785. Worsfold, P., & Macbeth, K. (2009). The reliability of television broadcasting statistics in soccer. International Journal of Petformance Analysis in Sport, 9,344-353. Wu, G., & Cavanagh, P. R (1995). ISS recommendations for standardization in the reporting of kinematic data. Journal of Biomechanics, 28, 1257-1261. Wu, G., van der Helm, F. C. T., Veeger, H. E. J., Makhsous, M., Van Roy, P., Anglin, C., et al. (2005). ISS recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion - Part II: shoulder, elbow, wrist and hand. Journal of Biomechanics, 38, 981992. Yeadon, M. R, & Challis, J. H. (1994). The future of performance-related sports biomechanics research. Journal of Sports Sciences, 12,3-32. Yiannakos, A., & Armatas, A., V. (2006). Evaluation of the goal scoring patterns in European

Championship

in

Portugal

2004.

International Journal of

Petformance Analysis in Sport, 6, 178-188.

170

APPENDIX A - VISUAL BASIC FOR APPLICATIONS CODE FOR CURVE-FITTING METHOD

171

To determine the location of a football on a pitch, customised software was written in Visual Basic for Applications (Microsoft Corporation, Washington, USA) and used to determine the pixel-coordinates of known pitch dimensions. The pixel coordinates were then used in the curve-fitting method to determine the location of the football on the pitch, as described in Chapter 3. To allow future researchers to use this method, the software code and an example of the digitising window used are presented in this section.

Initially the software program was developed to open video (.avi) files. This allowed for the video footage used for the validation of the method (Chapter 3) to be opened directly in the program. The Userform is shown in Figure A-1. The Load button is used to load the required video file for analysis. The control buttons to the right of the Load button allow for manipulation of the video (play, stop, step forwards and backwards). The blank white box is where the time point on the video would be shown, and the two white windows with numbers show the x- and y-pixel coordinates of the point the cursor is clicked on. Clicking the Add button writes the x- and y-coordinates to the Excel worksheet.

Figure A-1: The Userform window used for video analysis.

172

The following piece of code written in Visual Basic for Applications (Microsoft Corporation, Washington, USA) was used in conjunction with the Userform in Figure A-1 to perform the analysis. Comments explaining the code are written in

bold.

Private Sub CmdButton_Load_ClickO 'Subroutine to load a video for analysis. filename1 = Application.GetOpenFilename 'Opens browser for file selection. WindowsMediaPlayer1.URL

=filename1

'Loads video in media player

window. End Sub

Private Sub CmdButton_LargeStepBack_ClickO WindowsMediaPlayer1.Controls.currentPosition

=

WindowsMediaPlayer1.Controls.currentPosition - 0.04 * 5 'Steps video backward

5 frames for sampling rate of 25 Hz. TextBox_Time

=WindowsMediaPlayer1.Controls.currentPosition 'Displays the

time for the current position of the video. End Sub

Private Sub CmdButton_StepBack_ClickO WindowsMediaPlayer1.Controls.currentPosition = WindowsMediaPlayer1.Controls.currentPosition - 0.04 'Steps video backward 1

frame for sampling rate of 25 Hz. TextBox_Time = WindowsMediaPlayer1.Controls.currentPosition End Sub

Private Sub CmdButton_StepFwd_ClickO WindowsMediaPlayer1.Controls.currentPosition = WindowsMediaPlayer1.Controls.currentPosition + 0.04 'Steps video forward 1

frame for sampling rate of 25 Hz. TextBox_Time

=WindowsMediaPlayer1.Controls.currentPosition

End Sub

173

Private Sub CmdButton_LargeStepFwd_ClickO WindowsMediaPlayer1.Controls.currentPosition = WindowsMediaPlayer1.Controls.currentPosition + 0.04 * 5 'Steps video forward

5 frames for sampling rate of 25 Hz. TextBox_Time

=WindowsMediaPlayer1.Controls.currentPosition

End Sub

Private Sub CmdButton_Stop_ClickO WindowsMediaPlayer1.Controls.stop 'Stops the video. TextBox_Time

=WindowsMediaPlayer1.Controls.currentPosition

End Sub

Private Sub CmdButton_Play_ClickO WindowsMediaPlayer1.Controls.Play 'Plays the video. TextBox_Time

=WindowsMediaPlayer1.Controls.currentPosition

End Sub

Private Sub WindowsMediaPlayer1_Click(ByVal nButton As Integer, ByVal nShiftState As Integer, ByVal fX As Long, ByVal fY As Long) 'Subroutine which

stores current X, Y values of the cursor position when the video is clicked. TextBox_X = fX 'Puts X value into memory. TextBox_Y

=fY 'Puts Y value into memory.

End Sub

Private Sub CmdButton_Add_ClickO 'Subroutine toadd data to excel worksheet Row1

=4

'Sets the first possible row to put the data into.

Do Until Cells(Row1, 2) Row1

=Row1

=

1111

'Checks if current rows cells are empty.

+ 1 'If row was not empty go to the next row.

Loop 'Once an empty cell has been found exits the loop.

=TextBox_X 'Puts X value into cell. Cells(Row1, 3) =TextBox_Y 'Puts Y value into cell.

Cells(Row1, 2)

End Sub

174

For the application of the method to the World Cup coverage, a still frame of each free kick taken directly at goal in the 2007 women's football World Cup finals was captured using InterVideo® WinDVD® (Corel Corporation, USA) because access to the footage in .avi format was not possible. The program was therefore developed so that it allowed for both still images and video files to be opened and analysed.

The following Userform (Figure A-2) and code were used to analyse still images for analysis of the women's football World Cup 2007.

Figure A·2: The Userform window used for analysis of a still frame.

Private Sub CmdButton_Load_ ClickO 'Subroutine to load image for analysis. filename1 = Application.GetOpenFilename 'Opens browser for file selection. UserForm2.lmage1.PictureSizeMode

=fmPictureSizeModeZoom

'Changes

the scaling mode for the frame which displays the image so that it zooms. UserForm2.lmage1.Picture

=LoadPicture(filename1) 'Loads image in frame.

End Sub

175

Private Sub Image1_MouseDown(ByVal Button As Integer, ByVal Shift As Integer, ByVal X As Single, ByVal Y As Single) 'Subroutine which stores the current X, Y values of the cursor position when the image is clicked.

=1 Then 'Checks that the left mouse button was clicked. TextBox_X =Round((X * 10) / 6,0) 'Puts X value into memory. TextBox_Y =Round((Y * 1 0) / 6, 0) 'Puts Y value into memory.

If Button

End If End Sub

Private Sub CmdButton_Add_ClickO 'Subroutine to add data to excel

worksheet. Row1 = 4 'Sets the first possible row to put the data into. Do Until Cells(Row1, 2)

="" 'Checks if the current rows cells are empty.

Row1 = Row1 + 1 'If the row was not empty go to the next row. Loop 'Once an empty cell has been found exits the loop. Cells(Row1, 2) = TextBox_X 'Puts X value into cell. Cells(Row1, 3) = TextBox_Y 'Puts Y value into cell. End Sub

176

APPENDIX B - COMPARISON OF CURVEFITTING METHOD ACCURACY WITH 2-D DLT

177

The accuracy of the curve-fitting method described in Chapter 3 was compared with 20-0LT, which has been previously used for similar applications (Toki & Sakurai, 2005). As Chapter 3 is taken directly from a journal publication, which had limitations on the figures and word count, only the descriptive statistics were provided. Visual representations from this comparison are presented in this section, providing more detailed information on the error for each marker and the pattern of error distribution across and along the pitch using the two different analysis methods.

With a large field of view, the reconstruction errors of the curve-fitting method and the 20-0LT method were comparable for pitch-width coordinates (Figure B-1 and Figure B-2).

Distance from Touchline (m) 4

10

16

-5-1-----'----t-ll.47

35----

0.Q1

-Q1i-

--8----8~--8--

--8-----·--

~

i

__ ~ •• ~ __ ~.~~ .-1

""" _" "____ ~"_" __ "~_~ """ __ "~ __ " _______ "~_~_

0.76i------------+----+-----~__+_

0.03

Figure B-1: Error in reconstructed pitch-width coordinates for the 2D-DLT method from a large field of view.

178

Distance from Touchline (m)

10

1(3

I .30 -5-1-----t---------c --- ----1-9.19----0.27 ~-

I W .=

I .--60:09----0:3 --~q.1 0~00:34-

~~---o~------

101------+----------------1-~-.. -OM.-----q'"---Q,,~-~,-,.-

•.., ,., --i'0i--

i

i~ ---II----~.06--------0..-'---e--e-e-r.. " '" -" -+.-0,. -cv

'H>-t---f-----;--{

0.06

20

"---,---Q,,--8

o

1-----Q:35--8- -8----8~--E)-----

30---0.

-0,,-- ----- -

re"' --Q,,- -----

--- -

.34

03

i

Figure 8-2: Error in reconstructed pitch-width coordinates for the curvefitting method from a large field of view.

Error in pitch-width estimations increased as the distance from the goal line increased because the majority of the known pitch markings used to create the model were inside the penalty area (Figure B1 and Figure B2). Therefore, to calculate locations of markers outside this area, the equation of the lines from the known pitch markings to the intersection point were extrapolated for both the curve-fitting method and the 2D-DLT method. Hence any errors in the digitised pixel coordinates of the pitch markings were amplified. Furthermore, because the penalty area was distant to the camera, these markings were more prone to digitising errors and these errors became more apparent from 30 m onwards as the lines were extrapolated further.

179

For pitch length, 20-0L T reconstruction errors were significantly less (P

= 0.02),

but the difference of 0.11 m is less than the measurement error of both the curvefitting and 20-0LT methods (Figure B-3 and Figure B-4). The 20-0LT method produced the maximum reported error for any marker (the pitch-length coordinate of the marker located on the goal line and 22 m from the near touchline). This location was just outside the calibrated area and therefore required extrapolation.

Distance from Touchline (m)

4

10

16

52

cti.11

e

5

-=

8

S 10 III

Cri.12\

~ °0.14

°0.18

Q.1S

0

0.40

°0.14

°0.18

Cb.11

Q.28

0.02

0 0.28

C6.13

0.33

a.35 e e e

Q.23

0.36 037

~.os

q18

Q.14

0 0.05

Suggest Documents