Google DeepMind and Liverpool FC create TacticAI

In collaboration with Liverpool FC, the team has introduced TacticAI, a sophisticated model that leverages both predictive and generative AI technologies to furnish coaches with strategic insights specifically tailored for corner kicks.

The journey towards this application of AI in football began three years ago, marking the start of a multi-year research collaboration with Liverpool FC. The initial focus of the collaboration was on understanding the potential of AI in augmenting football tactics, with studies ranging from the analysis of penalty kicks to predictive systems capable of anticipating off-camera player movements.

TacticAI, the result of this collaboration, stands out as a full-fledged AI system that marries predictive and generative models to offer tactical insights. Its development was detailed in the journal, Nature Communications, highlighting its capacity to predict and generate tactical setups for corner kicks with remarkable accuracy.

The system employs a geometric deep learning approach, which is adept at modelling the complex dynamics of a football match by representing corner kick scenarios as graphs. This technique captures the intricate relationships between players, considering variables such as position, velocity, and height. By applying the principles of symmetry to football pitch representations, TacticAI can generalise models with a fraction of the data typically required, showcasing state-of-the-art results in predicting corner kick outcomes.

TacticAI’s ability to simulate various tactical setups and predict their outcomes provides coaches with invaluable insights, effectively enabling a more strategic approach to corner kicks. The AI's recommendations were overwhelmingly preferred by expert raters from Liverpool Football Club.

Moreover, TacticAI simplifies the tactical planning process, allowing for the efficient identification of key patterns and crucial players in successful strategies. Its generative model offers tangible recommendations for optimising player positions to achieve desired outcomes, a feature that was validated through extensive qualitative studies.

The success of TacticAI in providing tactical recommendations that rival real game strategies not only demonstrates its immediate utility but also paves the way for future advancements in sports AI. The project underscores the dynamic potential of AI to assist in sports strategy, offering a glimpse into a future where AI and human expertise converge to elevate the game.

As the collaboration with Liverpool FC progresses, the possibilities for expanding AI's role in football and beyond continue to unfold. The lessons learned from TacticAI extend into broader applications of assistive technology, aiming to blend AI analysis with human expertise across various domains.