AFL clubs close to using artificial intelligence to aid game strategy
Media release
AFL coaches could soon have a tool that crunches data from thousands of past men's matches in seconds to provide data-backed advice on how they can tweak their game plan to improve their team's chances of winning.
Deakin University researchers are putting the finishing touches on their Coach Decision Assistant – an artificial intelligence (AI) tool that will aid in-match decision making by AFL coaches.
AFL clubs are known to be interested in exploring the use of AI to aid strategic judgements made by coaches on game days.
School of Exercise and Nutrition Sciences PhD student Mitch Aarons said he was close to finalising a tool that would revolutionise the way coaches make decisions mid match.
"Our Coach Decision Assistant – or CoDA – compares data from historic match outcomes and in-game team metrics with the current performance characteristics of two teams during a match to provide real-time feedback on a team's chances of winning and how their chances might be improved," Mr Aarons said.
"The best AI is the AI that helps humans to perform better. CoDA will not replace coaches, but it has the capacity to complement the intuition, knowledge, and experience AFL coaches bring to the table to help troubleshoot technical performance-related problems or to validate the strategy they are pursuing. Ultimately, it will help them and their wider coaching team to make better, more informed, data-driven decisions quickly and more effectively."
Project supervisor Dr Dan Dwyer, of Deakin's Centre for Sport Research within the Institute for Physical Activity and Nutrition, said the tool, which functions as a web application, would allow coaches to simulate potential tweaks to their game play to assess the potential impact before implementing changes on the field.
For example, CoDA could identify contested possession as an area of the match in which a team is underperforming and where there is opportunity for improvement. It would then suggest achievable changes to that data that could help turn the tide of the match.
"We see similar technology already used by sporting leagues overseas, including the National Basketball Association and the National Football League in the United States," Dr Dwyer said.
"Coaches are under immense pressure to make on-the-spot decisions during matches that either improve or hinder their side's chances of winning based on countless variables. What our system does is take some of the guess work out of that decision making.
"It is a technical performance indicator-based tool which leverages AI to provide information about a team's chances of winning at any given point in the match. It also provides dynamic suggestions on the magnitude and direction of changes in performance indicators that would lead to the best improvement in win probability."
The first iteration of CoDA will not be available for use by AFLW clubs because the data fed into the tool has been drawn from past men's matches. But Mr Aarons predicted similar decision support systems would soon be used by coaching staff across the AFL's men's and women's leagues and other sporting codes.
"The potential for AI to assist decision making by coaches and their teams is exciting. We're predicting that in the not-too-distant future it will be commonplace across the AFL, AFLW and all sports for this type of technology to be used at the elite and professional level," Mr Aarons said.
