Search

This paper explores the role of artificial intelligence (AI) in elite sports. We approach the topic from two perspectives. Firstly, we provide a literature based overview of AI success stories in areas other than sports. We identified multiple approaches in the area of Machine Perception, Machine Learning and Modeling, Planning and Optimization as well as Interaction and Intervention, holding a potential for improving training and competition. Secondly, we discover the present status of AI use in elite sports. Therefore, in addition to another literature review, we interviewed leading sports scientist, which are closely connected to the main national service institute for elite sports in their countries. The analysis of this literature review and the interviews show that the most activity is carried out in the methodical categories of signal and image processing. However, projects in the field of modeling & planning have become increasingly popular within the last years. Based on these two perspectives, we extract deficits, issues and opportunities and summarize them in six key challenges faced by the sports analytics community. These challenges include data collection, controllability of an AI by the practitioners and explainability of AI results.

This study identifies dominant and intermediary players in football by applying a play-by-play social network analysis (SNA) on 70 professional matches from the 1. and 2. German Bundesliga during the 2017/2018 season. SNA provides a quantification of the complex interaction patterns between players in team sports. So far, the individual contributions and roles of players in football have only been studied at match-level considering the overall passing of a team. In order to consider the real structure of football, a play-by-play network analysis is needed that reflects actual interplay. Moreover, a distinction between plays of certain characteristics is important to qualify different interaction phases. As it is often impossible to calculate well known network metrics such as betweenness on play-level, new adequate metrics are required. Therefore, flow betweenness is introduced as a new playmaker indicator on play-level and computed alongside flow centrality. The data on passing and the position of players was provided by the Deutsche Fußball Liga (DFL) and gathered through a semi-automatic multiple-camera tracking system. Central defenders are identified as dominant and intermediary players, however, mostly in unsuccessful plays. Offensive midfielders are most involved and defensive midfielders are the main intermediary players in successful plays. Forward are frequently involved in successful plays but show negligible playmaker status. Play-by-play network analysis facilitates a better understanding of the role of players in football interaction.

This study explores the influence of sideout failure on performance in the next sideout in beach volleyball. The sample comprises 965 elite matches in the FIVB World Series 2012–2016 and in the Olympic Games 2012/2016 including 28,974 sideout sequences (12,755 for men and 16,219 for women). A sideout sequence consists of two sideouts by the same player during the same set in a timeframe of four rallies. The first sideout in this sequence is referred to as the previous sideout and the second sideout as the next sideout. After misses, χ2-tests indicate a significantly higher technique alternation rate (from spike to shot or vice versa) in the next sideouts for both men (+32.7%) and women (+40.4%) than the next sideouts after hits. After shot misses, the share of shots in the next sideouts was −12.9% lower for men and −8.3% lower for women than the next sideouts after shot hits. After spike misses, the share of shots in the next sideouts by female players was +5.5% significantly higher, and shot hit rate was −6.5% lower than the next sideouts after spike hits. These findings support the belief that tactical decisions and performance in top-level beach volleyball are influenced by failure in the previous sideouts. They might support coaches and players when analyzing matches and developing game strategies.

This study explores whether positioning systems are a viable alternative to timing gates when it comes to measuring sprint times in ice hockey. We compared the results of a single-beam timing gate (Brower Timing) with the results of the Iceberg optical positioning system (Optical) and two radio-based positioning systems provided by InMotio (Radio 1) and Kinexon (Radio 2). The testing protocol consisted of two 40 m linear sprints, where we measured sprint times for a 11 m subsection (Linear Sprint 11), and a shuttle run (Shuttle Total), including five 14 m sprints. The exercises were performed by six top-level U19 field players in regular ice hockey equipment on ice. We quantified the difference between measured sprint times e.g., by Mean Absolute Error (MAE) (s) and Intra Class Correlation (ICC). The usefulness of positioning systems was evaluated by using a Coefficient of Usefulness (CU), which was defined as the quotient of the Smallest Worthwhile Change (SWC) divided by the Typical Error (both in s). Results showed that radio-based systems had a higher accuracy compared to the optical system. This concerned Linear Sprint 11 (MAEOptical = 0.16, MAERadio1 = 0.01, MAERadio2 = 0.01, ICCOptical = 0.38, ICCRadio1 = 0.98, ICCRadio2 = 0.99) as well as Shuttle Total (MAEOptical = 0.07, MAERadio1 = 0.02, MAERadio2 = 0.02, ICCOptical = 0.99; ICCRadio1 = 1.0, ICCRadio2 = 1.0). In Shuttle Total, all systems were able to measure a SWC of 0.10 s with a probability of >99% in a single trial (CUOptical = 4.6, CURadio1 = 6.5, CURadio2 = 5.1). In Linear Sprint 11 an SWC of 0.01 s might have been masked or erroneously detected where there were none due to measurement noise (CUOptical = 0.6, CURadio1 = 1.0, CURadio2 = 1.0). Similar results were found for the turning subsection of the shuttle run (CUOptical = 0.6, CURadio1 = 0.5, CURadio2 = 0.5). All systems were able to detect an SWC higher than 0.04 s with a probability of at least 75%. We conclude that the tested positioning systems may in fact offer a workable alternative to timing gates for measuring sprints times in ice hockey over long distances like shuttle runs. Limitations occur when testing changes/differences in performance over very short distances like an 11 m sprint, or when intermediate times are taken immediately after considerable changes of direction or speed.

Advertising escape ability could reduce predatory attacks. However, the effectiveness of certain phenotypic cues (e.g., color, shape, and size) in signaling evasiveness is still unknown. Understanding the role of such signals in driving predator learning is important to infer the evolutionary mechanisms leading to convergent evasiveness signals among prey species (i.e., evasive mimicry). We aim to understand the role of the color pattern (white patches on dark background) and morphology (extended butterfly hindwings) in driving learning and avoidance of escaping prey by surrogate avian predators, the European blue tit. These cues are common in butterflies and have been suspected to advertise escape ability in nature. We use dummy butterflies harboring shape and color patterns commonly found in skippers (family Hesperiidae). The prey models displayed the studied phenotypical cues (hindwing tails and white bands) in factorial combinations, and we tested whether those cues were learned as evasive signals and were generalised to different phenotypes. Our results suggest that hindwing tails and white bands can be associated with prey evasiveness. In addition, wild blue tits might learn and avoid attacking prey models bearing the studied phenotypic cues. Although blue tits seem to have an initial preference for the phenotype consisting of white patches and hindwing tails, the probability of attacking it was substantially reduced once the cues were associated with escaping ability. This suggests that the same morphological cues might be interchangeable as preference/avoidance signals. Further investigation of the salience of hindwing tails vs. white bands as cues for escaping ability, revealed that predators can associate both color pattern and shape to the difficulty of capture, and possibly generalize their aversion to other prey harboring those cues. More studies with larger sample sizes are needed to confirm this trend. Altogether, our results highlight the hitherto overlooked role of shape (butterfly hindwing tails) for signaling prey unprofitability.