Google DeepMind has announced its latest artificial intelligence (AI) innovation — TacticAI. The AI assistant uses predictive and generative AI skills to help football (soccer) coaches build effective corner-kick strategies. The AI tool was developed in collaboration with the English Premier League club Liverpool FC, as a part of DeepMind’s multi-year partnership with them. The researchers claimed that in a blind test with football experts, it was found that the strategies shared by the AI were indistinguishable from real corners in previous games.
Introducing the AI assistant in a blog post, Google DeepMind said, “Today, in Nature Communications, we introduce TacticAI: an artificial intelligence (AI) system that can provide experts with tactical insights, particularly on corner kicks, through predictive and generative AI.” The research paper focuses on corner kicks, a particular set-piece play in the game of football which comes into play when the ball crosses the byline after touching a player of the defending team.
Once a corner-kick has been awarded, one player from the attacking side will take a free-kick from the corner of the pitch, and other players from both sides can place themselves at positions they expect to be advantageous to them. As per DeepMind, this is where TacticAI comes in and shares insights on the positioning and the runs the attacking players should make to increase the chances of scoring a goal. It also identifies the receiver of the ball based on multiple metrics.
TacticAI utilises both predictive and generative AI to prepare insights. Using data from corner kick tactics used in older games, it learns about the different tactical aspects such as player position, runs, height of the player, defensive gaps, and more. Then, the AI model predicts corner kick play by applying a geometric deep learning approach. “We directly model the implicit relations between players by representing corner kick setups as graphs, in which nodes represent players (with features like position, velocity, height, etc.) and edges represent relations between them,” the post explained.
DeepMind’s geometric architecture used in TacticAI, then, allows it to focus on three core problems. First, it predicts what will happen for a given corner kick tactical setup. This includes identifying who is most likely to receive the ball, and if there will be a shot attempt. Second, once the play has occurred, it focuses on understanding what happened and if similar tactics have worked well in the past. Third, it finds out the adjustments that need to be made for a particular outcome.
Google says that in a blind test, the football experts from Liverpool FC found that the suggestions given by the AI could not be distinguished from real corners, and were favoured over their original situations 90 percent of the time. The researchers added that TacticAI is both useful and deployable.
