Moving forward and penetrating the defensive zones is crucial for goal scoring in soccer games, yet it involves risky tactics. We propose a novel metric called Expected Value of Potential Penetrative Pass, which measures the likelihood of a potential penetrative pass creating scoring/conceding situations. We show how such a pass value accounting for the effects of crossing defense lines can be decomposed into elementary components. Using the UEFA EURO 2020 spatiotemporal dataset, we train several conventional machine learning and deep learning models to estimate these expected values for all potential penetrative pass situations in the dataset. For the best five and worst five teams in the dataset, we provide a trade-off between the ability to perform penetrative passes, and the expected value of goals those create. Finally, we show the impact of different field sections as the starting location of the penetrative pass to be performed and to create goal scoring situations.
© Copyright 2022 Machine Learning and Data Mining for Sports Analytics. MLSA 2022. Communications in Computer and Information Science. Veröffentlicht von Springer. Alle Rechte vorbehalten.