Monte Carlo Pass Search: Using Trajectory Generation for 3D Counterfactual Pass Evaluation in Football

The paper introduces Monte Carlo Pass Search (MCPS), a novel framework for evaluating football passes using a Monte Carlo Tree Search-like approach that integrates a value model, world model, and policy for counterfactual actions. It leverages a new high-fidelity tracking dataset from the Bundesliga and employs an autoregressive trajectory generator adapted from autonomous driving to enhance sample efficiency, achieving superior forecasting accuracy. This development provides practitioners with a robust tool for analyzing pass effectiveness and decision-making in football, facilitating improved tactical insights and player performance evaluation.