Expected Disruption: Modelling Defensive Interventions in Football

This study introduces a machine learning framework, named Expected Disruption, to predict defensive interventions in professional football using detailed spatiotemporal event data from the 2024–2025 Czech League season. The framework integrates expected threat (xT) modeling with pass characteristics—such as angle, length, and type—and pressing intensity to identify when and where on the pitch defensive actions are most likely to occur. Passes are enriched with context-aware metrics like xPass (expected pass success) and PP90 (pressures per 90 minutes). The model learns from spatial zones, player roles, and event outcomes, using a Random Forest Classifier. Results demonstrate strong predictive power with a ROC AUC of 0.81. This research contributes to modern football analytics by offering tools to evaluate team pressing strategies, identify key defenders, and analyse riskier passing patterns in real time. The primary motivation behind Expected Disruption is to provide a repeatable, scalable way to model defensive value, something historically difficult due to data limitations. By utilising widely available event data, this model bridges a critical gap in the analytics space, empowering coaches, analysts, and scouts with a practical tool to understand defensive efficiency. Moreover, the integration of player-specific pressure statistics (PP90) allows this framework to recognise not just the outcome of an event, but also the behaviours and tendencies that lead to it.