Abstract
Abstract
The increasing availability of positional data in invasion sports like football drives the development of advanced performance metrics. Two widely adopted examples are
space control
, which quantifies spatial influence, and
expected possession value
(EPV), which estimates the value of match situations. However, space control lacks theoretical and empirical grounding while EPV struggles with hypotheticals and demands prohibitively extensive data. We introduce a physically grounded alternative to both measures using pass completion maps generated from simulated passes. Our model is fitted and validated using only three matches of open data with minority-class oversampling. Its individual pass outcome predictions are well-calibrated and produce more plausible completion maps than a leading learning-based approach. By integrating valued completion maps, we derive
dangerous accessible space
(DAS), which measures threat potential by the amount of dangerous space that can be accessed through passes and carries. DAS captures elusive performance aspects such as defensive positioning, timing of attacking runs, and strategic decision-making in a combined spatial and value-based manner. It achieves 74% accuracy (80.4% excluding ball height tests) on the OJN-Pass-EPV benchmark (OJN: Overmeer, Janssen, Nuijten), rivaling cutting-edge EPV methods in identifying valuable match situations. The model and validation are fully available as the open-source Python package on PyPI.