DeepMind and Liverpool FC Collaborate to Develop AI Model for Corner Kick Tactics

DeepMind, the leading artificial intelligence research and development company owned by Google, has joined forces with Liverpool Football Club to create TacticAI, an innovative artificial intelligence model. TacticAI aims to revolutionize the prediction and optimization of corner kicks in football matches, assisting coaching teams in devising effective offensive and defensive strategies. By enhancing players' ability to make successful contact with the ball and creating opportunities to shoot, TacticAI aims to elevate team performance while minimizing the chances of opponents scoring.

This collaboration marks the beginning of a three-year research partnership between DeepMind's Petar Veličković and Liverpool Football Club.

In football, when the ball goes out of bounds across the goal line due to the last touch of a defending player, the attacking team is awarded a corner kick. Traditionally, corner kicks have represented prime scoring opportunities for attacking teams. Consequently, coaches invest significant effort into developing intricate tactics for various scenarios, which they communicate to players prior to matches.

TacticAI has been trained using data from 7,176 corner kicks that occurred during the 2020/2021 Premier League season. This dataset includes real-time information on player height, weight, and on-field positioning. Leveraging this wealth of data, the model is capable of predicting the player who is most likely to make the first touch during a corner kick. In testing, TacticAI accurately predicted the top three first-touch players in 78% of cases.

Coaches can leverage the insights provided by TacticAI to devise strategies for both offensive and defensive corner kicks, thereby maximizing the chances of specific players touching the ball and influencing shooting opportunities. By analyzing real-world corner kick scenarios, the AI model identifies patterns and recommends tactical adjustments to accomplish predefined objectives.

In a blind test, Liverpool's football experts found it challenging to discern between AI-generated tactics and those designed by human coaches, showing a preference for the AI solution 90% of the time.

Despite TacticAI's impressive performance, Veličković emphasizes that the AI system is not intended to replace coaches. Instead, it serves as a valuable tool to augment their decision-making capabilities. Furthermore, Veličković highlights the broader applications of this research beyond the realm of sports. Particularly, in situations involving uncertainty, AI systems can provide decision-making guidance and suggestions across various fields.