TacticAI: an AI assistant for football tactics | Nature Communications

Summary

This article describes "TacticAI," a graph neural network-based deep learning pipeline developed to analyze football match data, particularly corner kicks, by representing the spatial-temporal data as graphs. The process includes training and evaluating models for predicting outcomes like the receiver of a corner kick or shot opportunities, with a focus on the geometric deep learning framework to address symmetries and regularities in sports data. The tags like "General topology," "Machine learning," "Tensors," "Filter," "Calculus on finite weighted graphs," and "Algebraic signal processing" all relate to the methodologies and mathematical foundations used in constructing and training the models within TacticAI.