Modelling the Influence of Task Constraints on Goal Kicking Performance in Australian Rules Football

Artificial intelligence applications in the football codes: A systematic review

Artificial Intelligence (AI) is increasingly being adopted across many domains such as transport, healthcare, defence and sport, with football codes no exception. Though there is a range of potential benefits of AI, concern has also been expressed regarding potential risks. An important first step in ensuring that AI applications in football are usable, beneficial, safe and ethical is to understand the current range of applications, the AI models adopted and their proposed functions. This systematic review aimed to identify different applications of AI across football codes to synthesise current knowledge and determine whether potential risks are being considered. The systematic review included 190 peer-reviewed articles. Nine areas of application were found ranging from athlete evaluation and event detection to match outcome prediction and injury detection and prediction. In total, 27 different AI models were identified, with artificial neural networks the most frequently applied. Five AI assessment metrics were identified including specificity, recall, precision, accuracy and F1-score. Four potential risks were identified, concerning data security, usability, data biases and inappropriate athlete load management. It is concluded that, though a wide range of AI applications currently exist, further work is required to develop AI for football and identify and manage potential risks.

Shots at goal in Australian Football: Historical trends, determinants of accuracy and common strategies

OBJECTIVES

To understand the historical context of and relationships between, the characteristics of shots at goal and match outcome in the Australian Football League.

DESIGN

Observational.

METHODS

Descriptive statistics of shots at goal were calculated and compared across 20 seasons. The location, type, and outcome of all shots at goal (n = 43,254), by all teams (2017-19 & 21), were compared with match outcome.

RESULTS

The total number of shots per match and the accuracy of shots haven't changed in two decades. Most teams win by having more shots at goal (Wilcoxon-r effect size 0.63) than their opponent (i.e. "majority strategy") and of these, the number of open shots (0.48) is slightly more important than set shots (0.43), followed by shot accuracy (0.29). However, some teams (14 %) win by taking fewer shots at goal from field locations with a higher likelihood of scoring a goal (i.e. "minority strategy"). Arc angle and shot type can be used to predict the outcome of a shot at goal with 60.3 % classification accuracy. The novel shot-outcome prediction model reported here provides a better opportunity to evaluate goal kicking performance of teams and players, because it accounts for the type and difficulty of the shot.

CONCLUSIONS

Teams can evaluate the shot accuracy of their players more fairly, by accounting for shot location using the method reported here. Coaches can compare the two shot strategies identified and implement the one that suits the skill profile of their players and increase their likelihood of winning.

Evaluating the influence of a constraint manipulation on technical, tactical and physical athlete behaviour

Evaluating practice design is an important component of supporting skill acquisition and improving team-sport performance. Constraint manipulations, including creating a numerical advantage or disadvantage during training, may be implemented by coaches to influence aspects of player or team behaviour. This study presents methods to evaluate the interaction between technical, tactical and physical behaviours of professional Australian Football players during numerical advantage and disadvantage conditions within a small-sided game. During each repetition of the game, team behaviour was manually annotated to determine: repetition duration, disposal speed, total disposals, efficiency, and disposal type. Global Positioning System devices were used to quantify tactical (surface area) and physical (velocity and high intensity running) variables. A rule association and classification tree analysis were undertaken. The top five rules for each constraint manipulation had confidence levels between 73.3% and 100%, which identified the most frequent behaviour interactions. Specifically, four advantage rules involved high surface area and medium high intensity running indicating the attacking team's frequent movement solution within this constraint. The classification tree included three behaviour metrics: surface area, velocity 1SD and repetition duration, and identified two unique movement solutions for each constraint manipulation. These results may inform if player behaviour is achieving the desired outcomes of a constraint manipulation, which could help practitioners determine the efficacy of a training task. Further, critical constraint values provided by the models may guide practitioners in their ongoing constraint manipulations to facilitate skill acquisition. Sport practitioners can adapt these methods to evaluate constraint manipulations and inform practice design.