DeepMind, Google’s AI division, has unveiled an AI system developed in collaboration with English football club Liverpool, called “TacticAI”. This system recommends gameplay strategies using a combination of predictive and generative AI, with a focus on corner kicks.
TacticAI was trained on data from 7,176 corner kicks in the Premier League 2020/2021 season, including player positions and physical attributes. This data allows the AI to assess the outcome of each corner kick, predicting who will get the first touch, whether a goal will be scored, and the positioning of players in front of the goal, influencing the likelihood of a goal. Testing revealed that TacticAI correctly predicted the players who would get the first touch (3 players) 78% of the time.
In a Blind Test conducted by DeepMind, experts from Liverpool could not differentiate between game plans devised by AI and humans, with 90% preferring the AI-generated plans. This demonstrates the practicality of AI-generated strategies.
Petar Veličković, a DeepMind researcher, emphasized that TacticAI is not meant to replace humans in football planning but to enhance decision-making for coaches in creating gameplay strategies. He hopes this model can be further developed and applied to other sports as well.
TLDR: DeepMind collaborated with Liverpool FC to develop TacticAI, an AI system for generating football gameplay strategies based on corner kick data. The AI outperformed human predictions and showed potential for application in other sports.
Leave a Comment