Stuart Morgan

Matchplay characteristics of Grand Slam tennis: implications for training and conditioning

ABSTRACT The purpose of this study was to probe the sex-based differences in the stroke and movement dynamics of Grand Slam hard-court tennis. Player and ball tracking data were collated for 102 male and 95 female players during the 2012–2014 Australian Open tournaments. Serve, serve return, groundstroke and movement data were compared between sexes. Serve statistics were the subject of the largest differences, with males achieving significantly faster speeds, aces and unreturned serves while also winning a greater percentage of service points. When returning serve, women contacted the ball closer to the net, lower to the ground and achieved flatter ball trajectories than males. Groundstroke frequencies were similar between sexes, although males hit with greater speed, flatter trajectories and impacted more shots inside the baseline. Distance covered per set or during points won or lost was not sex dependent, yet men exhibited faster average movement speeds. These findings highlight the need for sex-specific training and practice designs that cater to the different stroke dynamics, particularly in relation to the first serve and serve-return, as well as movement speeds.

Predicting Shot Locations in Tennis Using Spatiotemporal Data

Over the past decade, vision-based tracking systems have been successfully deployed in professional sports such as tennis and cricket for enhanced broadcast visualizations as well as aiding umpiring decisions. Despite the high-level of accuracy of the tracking systems and the sheer volume of spatiotemporal data they generate, the use of this high quality data for quantitative player performance and prediction has been lacking. In this paper, we present a method which predicts the location of a future shot based on the spatiotemporal parameters of the incoming shots (i.e. shot speed, location, angle and feet location) from such a vision system. Having the ability to accurately predict future short-term events has enormous implications in the area of automatic sports broadcasting in addition to coaching and commentary domains. Using Hawk-Eye data from the 2012 Australian Open Mens draw, we utilize a Dynamic Bayesian Network to model player behaviors and use an online model adaptation method to match the players behavior to enhance shot predictability. To show the utility of our approach, we analyze the shot predictability of the top 3 players seeds in the tournament (Djokovic, Federer and Nadal) as they played the most amounts of games.

Applying decision tree induction for identification of important attributes in one-versus-one player interactions: A hockey exemplar

Abstract Decision tree induction is a novel approach to exploring attacker-defender interactions in many sports. In this study hockey was chosen as an example to illustrate the potential use of decision tree inductions for the purpose of identifying and communicating characteristics that drive the outcome. Elite female players performed one-versus-one contests (n = 75) over two sessions. Each contest outcome was classified as either a win or loss. Position data were acquired using radio-tracking devices, and movement-based derivatives were calculated for two time epochs (5 to 2.5 seconds, and 2.5 to zero seconds before the outcome occurred). A decision tree model was trained using these attributes from the first session data, which predicted that when the attacker was moving at ≥ 0.5 m · s−1 faster than the defender during the early epoch, the probability of an attackers win was 1.00. Conversely, when the speed difference at that time was below this threshold the probability of a loss was 0.78. Secondary attributes included defender speed in the lateral direction during the early epoch, and angle of attack (i.e., angle between the respective velocity vectors of the attacker and defender) during the late epoch. The model was then used to predict outcomes of one-versus-one contests from the second session (accuracy = 0.643; area under the Receiver Operating Characteristic (ROC) curve = 0.712). Moreover, decision trees provide an intuitive framework for relating spatial-temporal concepts to coaches, and the suitability of decision trees for analysing the features of one-versus-one exchanges are discussed.

Rankings in professional men’s tennis: a rich but underutilized source of information

Abstract Success in professional tennis is measured, at least in part, by rankings. However, there is little quantitative evidence to inform stakeholders regarding what represents the typical ranking progress of top-ranked players. The objective of this study was therefore to compare the ranking trajectories of male players whom achieved peak professional rankings in the Top 250, 175, 100, 50, 20 and 10. The 11,396 birthdates and weekly professional rankings of all players between 27 August 1973 and 31 October 2011 were collated. The peak ranks for each athlete according to their both chronological age and number of years on tour were identified and athletes were categorised into one of six career–peak ranking bands. One-way analysis of variance tests confirmed distinctive ranking trajectories, which were most pronounced among Top 10 players. The rankings of these players were statistically distinguishable following players’ second year on tour or by 17 years of age. The ranking signature of all Top 100 players emerged as significantly different to players that failed to enter the Top 100 by their fourth year on the tour. Indeed, the representation of ranking as a function of years on tour should be considered for use by tennis policy-makers in the future.

Swimmer Localization from a Moving Camera

At the highest level of competitive sport, nearly all performances of athletes (both training and competitive) are chronicled using video. Video is then often viewed by expert coaches/analysts who then manually label important performance indicators to gauge performance. Stroke-rate and pacing are important performance measures in swimming, and these are previously digitised manually by a human. This is problematic as annotating large volumes of video can be costly, and time-consuming. Further, since it is difficult to accurately estimate the position of the swimmer at each frame, measures such as stroke rate are generally aggregated over an entire swimming lap. Vision-based techniques which can automatically, objectively and reliably track the swimmer and their location can potentially solve these issues and allow for large-scale analysis of a swimmer across many videos. However, the aquatic environment is challenging due to fluctuations in scene from splashes, reflections and because swimmers are frequently submerged at different points in a race. In this paper, we temporally segment races into distinct and sequential states, and propose a multimodal approach which employs individual detectors tuned to each race state. Our approach allows the swimmer to be located and tracked smoothly in each frame despite a diverse range of constraints. We test our approach on a video dataset compiled at the 2012 Australian Short Course Swimming Championships.

Horizontal positioning error derived from stationary GPS units: A function of time and proximity to building infrastructure

This investigation quantified Horizontal Positioning Error (HPE) from stationary Global Positioning System (GPS) units. In experiment 1, GPS units were placed within close proximity of each other and data were collected at three times within the same day. In experiment 2, the GPS units were configured in a straight line from the edge of the field (nearby to a stadium), towards the centre. Circular Error Probable of 95% (CEP95) was used to quantify HPE, mean numbers of satellites recorded. CEP95 and the number of satellites were inversely related in both experiments. Changes in satellite availability throughout the day led to significant variability in HPE across trials in experiment 1. In experiment 2, the largest CEP95 was found among units located closest to the stadium. Collectively, these findings identify important considerations for using GPS to map athletes’ positions that have not been adequately addressed in the sports science literature. The number satellites, time between repeated measures testing, and the proximity of nearby buildings, can cause unpredictable changes in measurement error and should be considered in the interpretation of data. Further, we recommend that CEP95 be reported in GPS-based experiments, which may help prevent the misleading interpretation of inaccurate or unstable data.

Forecasting Events Using an Augmented Hidden Conditional Random Field

In highly dynamic and adversarial domains such as sports, short-term predictions are made by incorporating both local immediate as well global situational information. For forecasting complex events, higher-order models such as Hidden Conditional Random Field (HCRF) have been used to good effect as capture the long-term, high-level semantics of the signal. However, as the prediction is based solely on the hidden layer, fine-grained local information is not incorporated which reduces its predictive capability. In this paper, we propose an “augmented-Hidden Conditional Random Field” (a-HCRF) which incorporates the local observation within the HCRF which boosts it forecasting performance. Given an enormous amount of tracking data from vision-based systems, we show that our approach outperforms current state-of-the-art methods in forecasting short-term events in both soccer and tennis. Additionally, as the tracking data is long-term and continuous, we show our model can be adapted to recent data which improves performance.

Has player development in men's tennis really changed? An historical rankings perspective

Abstract Tennis federations are regularly faced with decisions regarding which athletes should be supported in financial terms, and for how long. The financial investments can be considerable, given the cost of competing on tour has been estimated at a minimum

Understanding and analyzing a large collection of archived swimming videos

121,000 per year and only the top 130 professionally ranked athletes earned enough prize money to cover this cost in 2012. This study investigates key points of progression in tennis players’ careers, to determine how these have changed over time and how that evolution may inform talent development. Approximately 400,000 weekly rankings for 273 male professional tennis players between 1985 and 2010 were compiled, and historical trends in the time taken to reach career milestones were investigated by least-squares regression. The time between earning a first professional ranking point and entry into the Top 100 significantly increased over time for all considered athletes. This was at the detriment of time spent within the Top 100 for some athletes. Career peak Top 50–100 athletes have shown an increase in longevity. These results assist tennis federations in assessing the progress of developing athletes and highlight the evolving nature of the competition for top players.

Discovering frequently recurring movement sequences in team-sport athlete spatiotemporal data

In elite sports, nearly all performances are captured on video. Despite the massive amounts of video that has been captured in this domain over the last 10-15 years, most of it remains in an “unstructured” or “raw” form, meaning it can only be viewed or manually annotated/tagged with higher-level event labels which is time consuming and subjective. As such, depending on the detail or depth of annotation, the value of the collected repositories of archived data is minimal as it does not lend itself to large-scale analysis and retrieval. One such example is swimming, where each race of a swimmer is captured on a camcorder and in-addition to the split-times (i.e., the time it takes for each lap), stroke rate and stroke-lengths are manually annotated. In this paper, we propose a vision-based system which effectively “digitizes” a large collection of archived swimming races by estimating the location of the swimmer in each frame, as well as detecting the stroke rate. As the videos are captured from moving hand-held cameras which are located at different positions and angles, we show our hierarchical-based approach to tracking the swimmer and their different parts is robust to these issues and allows us to accurately estimate the swimmer location and stroke rates.