The most exciting young coach in football might not be at Bayer Leverkusen, or Stade de Reims, or even Bologna FC. It might be at Google DeepMind.
For the last few years, the search giant’s artificial intelligence division has been working with Liverpool Football Club to bring AI to the world’s most popular sport. In 2021, DeepMind researchers developed a model that could predict where players would hit a penalty based on their outfield position. In 2022, they developed one that analyzed video footage of games to predict where players would run next, even when they went off screen. “But none of these systems were a complete prototype that could feasibly give useful suggestions to coaches in the real world,” says Petar Veličković, a staff research scientist at Google DeepMind and co-author of a paper published today in Nature Communications. “We wanted to actually build something that could lead to a feasible system.”
Enter TacticAI. It started life as a predictive system for open play—one that could analyze a game and tell coaches who was most likely to receive a pass, or what their chances of creating a dangerous goalscoring opportunity might be. But the data analysts and coaches at Liverpool wanted something simpler. “In open play you can’t make a lot of useful on-the-spot changes, because there’s 22 players, and it’s very dynamic, and if you try to make changes in the heat of the moment you might end up confusing people” says Veličković.
Instead, Liverpool suggested that DeepMind’s researchers focus on corner kicks. Roughly ten times a game, the action on the pitch is effectively frozen and the attacking team gets an opportunity to swing the ball into the box. But only one in 50 corners actually results in a goal. Elite clubs already spend a huge amount of time in the lead up to games preparing corner routines and defensive plans with elaborate running routes and blocking schemes. “If you can increase your chances to score or defend better at corners, that integrates over an entire season to really give you a competitive edge,” says Veličković.
Working with player-tracking data from 7,176 corners taken in the Premier League during 2020 and 2021, the researchers began by representing the arrangement of players as a graph, with the players’ position, movement, height, and weight encoded as nodes on the graph, and relationships between players as the lines between them. Then they used an approach called geometric deep learning, which takes advantage of the symmetry of a football pitch to shrink down the amount of processing the neural network needed to do. (This isn’t a new strategy—a similar approach was used in DeepMind’s influential AlphaGo research.)
The resulting model led to the creation of a number of tools that could be useful to football coaches. Based on the arrangement of players at the moment the kick is taken, TacticAI can predict which player is most likely to make the first contact on the ball, and whether a shot will be taken as a result. It can then generate recommendations for the best ways to adjust player position and movement to either maximize the chance of a shot being taken (for the attacking team) or minimize it (for the defending team)—shifting a defender across to cover the near post, for instance, or putting a man on the edge of the area.
The football experts at Liverpool particularly liked how TacticAI’s recommendations could pinpoint attackers who were critical for the success of a particular tactic, or defenders who were “asleep at the wheel,” Veličković says. Analysts spend hours sifting through video footage looking for weak points in their opponents’ defensive setups that they can target, or trying to find holes in their own team’s performances to double down on in training. “But it’s really hard to track across 22 people, across lots of different situations,” Veličković says. “If you have a tool like this it immediately helps you see which players are not moving in the right way, which players should be doing something different.”
