Evrim Yazgin
Cosmos science journalist
Sports teams may want to employ artificial intelligence (AI) coaches based on an analysis of the performance of one such algorithmic tactician.
AI system “TacticAI” is produced by the computer programming company DeepMind, owned by Google’s Alphabet Inc. It has been able to accurately predict the outcome of corner kicks in football as well as providing realistic and accurate tactical suggestions in matches, according to an analysis published in Nature Communications.
The authors trained and tested TacticAI using a dataset of 7,176 corner kicks from past Premier League seasons supplied by Liverpool Football Club.
Corner kicks, deep in the opposition’s defence, are an immediate opportunity to score a goal. Hence, coaches spend a lot of time training corner kick routines to try and outsmart defences and increase the chance of scoring, or, on the flip side, to foil a potential attacking team’s routine.
A survey found that experts from Liverpool FC favoured the strategies of TacticAI over existing strategies 90% of the time. The Liverpool experts were 3 data scientists, 1 video analyst and a coaching assistant from the club.
They write that TacticAI was tested on “a number of relevant benchmark tasks: predicting receivers and shot attempts and recommending player position adjustments.”
It also comes after AI “humanoids” were taught to play the world game “from scratch” through algorithms also developed by DeepMind.
TacticAI is based on Geometric Deep Learning – a type of machine-learning algorithm that analyses geometric values such as positions, distances and direction.
“TacticAI incorporates both a predictive and a generative component, allowing the coaches to effectively sample and explore alternative player setups for each corner kick routine and to select those with the highest predicted likelihood of success,” the authors write.
Corner kicks are but one example of a “set piece” in football where AI tactics might be employed. Others include free kicks, goal kicks, throw ins and penalties. The authors say that it could be useful in “other team sports with suspended play situations.”
The results suggest AI could prove to be a valuable new field for sports coaching soon.
