World Congress of Performance Analysis of Sport XI
Will G Hopkins
Sportscience 20, 33-36 2016 (sportsci.org/2016/WCPAS.htm)
The venue for the eleventh biennial conference of the International Society of Performance Analysis of Sport (ISPAS) was the University of Alicante on the sunny Mediterranean coast of Spain. Our hosts made us very welcome, the tapas were terrific, Spanish wine tastes like the real thing in Spain, and winter there felt like summer.
Alicante is dominated by Santa Barbara Castle, an old fort dating back 1000 years to the time of the Arab conquest of the Iberian Peninsula. The climb to the top was rewarded with spectacular views of sea and city, and of a parched hinterland in the grip of a long-term drought. The south of Spain now survives on water piped in from the north, apparently.
I attend conferences in the hope that at least one inspirational idea, method or finding will justify the time, expense and effort of attending (and of writing this report). Here there were two, both addressing the issue of the complex data that come from team sports: the application of genomics analyses to match analyses, and use of self-organizing maps to guide team-sport strategy in real time.
The conference website has links to the program and to videos of the entire morning and afternoon sessions in the main lecture theatre. The abstracts and some keynote slideshows are available via a dropbox, which includes two PDFs (Oral Session IDs, Poster Session IDs) that will allow you to find the abstracts of talks and posters with interesting titles. I have also provided direct links to the abstracts of talks and posters summarized in this article.
I have summarized only those presentations that had an element of performance analysis with a reasonably semantically and statistically clear outcome. I have omitted several presentations focusing only on biomechanics, physiology or psychology. I apologize to those presenters and to any others I have omitted through my failure to understand or value their findings.
In the opening keynote, Mike Hughes reviewed the past 10 meetings of ISPAS before identifying the following areas for "urgent" further research: objective definitions of performance indicators; clarification of the complex statistical procedures in use; attention to issue of "sensitivity" (to smallest important changes) of the measures; and development of more models. He finished with a plea for innovations in gathering and analyzing data, for interpreting rather than just counting actions, for better communication with the coach, and for the ability to listen to others.
You need potentially thousands of variables to characterize performance of all the players throughout a match, but you usually have only hundreds of observations (e.g., game outcomes). Genomics researchers have a similar problem, so their analytical procedures could be useful for performance analysts. Great idea! McInerney
I spoke on the topic of the magnitudes of performance enhancements that increase a top athlete's or team's chances of winning more medals. Different approaches to defining and researching the magnitudes are needed in non-interactive sports, interactive non-team sports, and team sports. The slideshow is a minor updated version of my talk at the ICSEMIS conference. Hopkins
In a review of 12 recent studies using social network analysis, "increased interactions, such as passes between players, lead to success of a team" and one study showed that "a [more] homogenous distribution of links (passes) is associated with a greater probability of success of the team". Roscillo
If you are a practicing performance analyst, you might like to compare your modus operandi with that of Olympic and Paralympic analysts. Nicholls
In a closing keynote, Peter O'Donoghue presented some of his recent work using simulation to show how poor reliability (reproducibility) of performance indicators can seriously impact their interpretation. O'Donoghue.
In NBA basketball, shooting efficacy and chances of winning decrease when fixture congestion results in less than two days between games. Esteves
Several kinds of footwork leading into a jump shot had no significant effect on the successful shooting percentage in eight university-level basketball games. But the poster showed the comparison was done non-parametrically, which reduces power and prevents proper estimation of magnitude. Li-Yao
Two independent discriminant functions based on the usual performance indicators collected in basketball matches revealed clear differences in the playing styles in African, American, Asian and European championships. González-Espinosa
"In contrast to traditional 5x5 basketball, 3x3 competition doesn´t induce fluctuations in heart rate and also, causes more impacts and jumps per minute." Reina-Román
Nic James gave a keynote in which he reviewed performance analysis of football. The game itself has changed from direct play (more shots at goal) to possession-based (more passes). Recent useful metrics include the "high press" (number of passes a team is allowed to make in its own half per defensive action), counter-attacks, and playing styles based on offensive and defensive variables. The future is in data-mining analytics for the big data sets representing movement of all the players in a match and accounting for strategy and tactics.
In a keynote on fatigue in soccer, we learned that distance covered in the second half of a game is not consistently down on the distance in the first half, but there is evidence of less high-speed running in the second half. Players also slow up for a few minutes following intense bursts of running. Lago (PDF of full talk)
The notion that certain game actions can be regarded as perturbations leading to unstable states ending potentially in scoring is still a promising work in progress in soccer. Kim
The types of goals scored did not differ between winners and losers in 63 qualifying matches in the Spanish futsal league. Agras
I could not understand a potentially important abstract on factors affecting goal-scoring in soccer when the opposition exhibits low defensive control. Tenga
If the usual player performance indicators in soccer really do measure performance, players' mean values ought to have reasonable correlations with the number of times the players have been nominated for the Golden Ball award. The abstract seems to have some ambiguous results, but the positive correlates were number of goals, shots, dribbles, thrown penalties and passes. Prieto-Ayuso
In a study of 750 long corner kicks in the English football premier league, it was evident that "near post guards play an important role when defending corner kicks, as they regularly perform defensive clearances (31.5% of total corners)." Pulling
The majority (64%) of team members in the European football championships had no youth international age-group experience. The authors hint that the football associations should spend less money on youth football. Bacon
See what a survey of 129 professional and semi-professional futsal coaches revealed about strategy when you're near the end of a losing game. Domínguez
In an analysis of turnovers in elite soccer, "teams should press high to win turnovers and create scoring opportunities… Immediate player actions after winning the turnover are critical to the outcome." Lovell
Self-organizing maps were touted as a useful way to visualize match performance indicators in football World Cups, but I missed the presentation and there are no findings in the abstract. Choi
In a season of Spanish 2nd B Division soccer, penalty kicks directed at the top third of the goal are "most effective" (presumably highest rate of success) while those directed at the bottom third are "most used". But all we're given is a chi-squared statistic, and we don't know which direction gives the biggest payoff (frequency´effectiveness). Pérez Aleixandre
Presenting individual-player match performance indicators as percentiles is the right idea, but I can't see how can you develop percentiles with only six soccer players. A full mixed-model analysis with a large number of players is required. Guerra Puente
A comprehensive set of performance indicators for Gaelic football has been developed. Browne
In a qualitative case study of delivery of performance analysis in a professional rugby team, "considerable emphasis was placed on the importance of player learning and engagement in the PA process, [but] delivery was often ad hoc, coach-driven and results-focused." Personal relationships were important for effective delivery. Croft
With only eight men's and eight women's matches from the knockout stage of the last rugby World Cup in the analysis, the identification of performance indicators that discriminate winning and losing is interesting but a long way from definitive. Barnes
Although 12 out of 30 performance indicators were "significantly different across match outcome" in two seasons of international rugby union, apparently only (?) "tackle effectiveness and missed tackles could predict match outcome" in logistic regressions. Why the disparity? Fitzgerald
The findings in a study of strategies in rugby union leading to point scoring for the four most successful teams in the last Rugby World Cup are probably useful, but I missed the presentation and I have been unable to understand the abstract. Barnes
A performance score consisting of unspecified weightings of unspecified offensive and defensive actions had different median scores for various player positions and replaced vs non-replaced rugby players. Smyth
A between-group comparison of gaze behavior with only three athletes in each group (international, senior league, junior) would be a joke, except that there was a lot of repeated measurement and results were so consistent that it changed the advice of badminton coaches: rather than watch the shuttle, they now say to watch the racket head. Dednam
The top eight teams at the men's handball world championships threw the ball with higher speed in the central position than the bottom eight teams. It looks like there was no significant difference in number and effectiveness of throws, but the abstract appears to be poorly edited, and the use of non-parametric inference makes these findings untrustworthy. Cortés
An analysis of the defensive retreat in European championship handball did not seem to result in recommendation for strategy. Sousa
"Hurling teams demonstrating higher skill levels with 'clean' sequences are likely to create more shots and ultimately win more games." Clear
Self-organizing maps have been applied to netball games to identify game styles associated with winning and losing and to implement and monitor (with a dashboard interface) strategies to change the style and thereby apparently increase the chance of winning. Groundbreaking! Croft
Differences in passing revealed by network analysis in three "critical" matches in the Commonwealth Games "suggest that tactical approaches differ between the World’s top four international netball teams." Uncertainty in standardized estimates would be preferable to chi-squared statistics in conveying some sense of magnitude. Butterworth
Knowing the shot and error rates for different parts of the squash court at the international level could be useful, but it seems to me that it needs a sophisticated analysis that accounts for differences between individual players. Ghani
It looks like service points, block points, service errors and attack errors are unsurprisingly important predictors of outcomes in (men's?) world-league volleyball, but I can't really understand the abstract. Vicente
Differences in success of attacks between positions might help inform strategy in men's world-championship volleyball. Millán-Sánchez
The "set" area and tempo in men's World Championship volleyball affected the number of players blocking the attacker, and therefore potentially the success of the attack (although this aspect was not reported). Fernández-Echeverría
There were various significant differences between winning and losing teams in shots at goal in 69 Olympic water-polo matches, but it was all done non-parametrically, so there are no magnitudes and the non-significant effects aren't necessarily trivial. Menescardi
A six-month archery-specific resistance-training program involving rubber bands improved competition scores by 2.9%, whereas a control resistance-training group and a control usual-training (?) group improved by only 1.0% and 0.2% respectively, in this randomized balanced controlled trial of 24 archers with at least 4 y experience. If the data after the "±"signs are standard deviations, the within-athlete competition-to-competition SD was 1.2/Ö2 = 0.8%, which makes the specific-training effect very large and, remarkably, the non-specific training effect moderate. Monzoni
On the basis of analysis of the run-up to the first hurdle in the Spanish and world indoor championships, the authors propose use of midpoint marks at 4.65 and 4.55 m for approach run phase for eight-steps hurdlers. González Frutos
The authors argued for the use of Kohonen feature maps (a neural-net data-mining technique) to combine various physical tests to predict throwing performance of 35 javelin throwers on two occasions separated by 1 y, but the evidence doesn't seem to be presented. Maszczyk
Raúl Arellano presented his long experience of tracking and predicting performance of swimmers in a keynote address. There is no abstract. A video is available, but unfortunately the view of the speaker obstructs part of the slides.
Automatic tracking of swimmers using eight cameras and algorithms for image recognition appears to estimate split times with high validity. Ruiz-Teba
Even splits is definitely not the way to swim 400 m freestyle: the lap times of Kathleen Ledecky, the best of the 20 elite swimmers analyzed here over their competitive careers, displayed the most negative quadratic curvature and the slowest first lap. "There were otherwise no clear associations between ability of the swimmers and the pacing parameters," but up to one-half of the swimmers might improve their times by up to 1% by changing one or more parameters of their pacing profile. Lipinska
Sequences of tactical actions in 75 Olympic male matches of taekwondo "can provide an insight for athletes’ patterns" but it's not clear how they can inform strategy or tactics. Menescardi
In an analysis of nearly 30,000 points in matches between tennis star Novak Djokovic and his opponents, Djokovic's scoring was modified by game status (winning, losing, equal score, breaking opponent and facing breakpoint) but apparently not by different opponents, rounds, sets or server. It would be good to know how many points need to be analyzed for this approach to provide useful strategies for other players. Cui
Transgender female athletes lose the testosterone advantage after a year or two and end up with age-graded scores (percent of world record for their age) similar to what they had as males. "The data offer further support for the recent IOC decision to allow transgender women to compete against cisgender women in the 2016 Olympics after one year of hormone-replacement therapy." Harper
Acknowledgements: WCPAS paid for my registration and accommodation; High Performance Sport NZ provided financial support through a consultancy.
Published Nov 2016