WINSCORE revisited: A model-based evaluation of player performance in the NBA and EuroLeague

In professional basketball, despite progress in Sport Analytics, widely used methods of player evaluation still rely on box-score statistics aggregated without formal justification for the weights used. The popular Performance Index Rating (PIR) is a prominent example. This paper revisits the WINSCORE methodology proposed by \cite{berri2006wages} and reinterprets it within a structural, model-based framework that links individual statistics to team outcomes through the efficient management of possessions. In this framework, the weights assigned to player statistics are derived endogenously from the data, providing a coherent foundation for measuring performance. Using NBA and EuroLeague data, we evaluate the model at the team level through goodness-of-fit evidence, residual analysis, encompassing tests against PIR, and comparisons with alternative advanced measures, including Four Factors and adjusted box-score plus-minus indicators. The results show that the model accounts well for variation in team performance, displays no major residual patterns, and adds explanatory content beyond existing metrics. We then extend WINSCORE to construct player performance measures for season-level and single-game analysis. These measures incorporate opponent adjustments and address playing-time endogeneity. The resulting indicators provide transparent tools for player assessment, coaching, and tactical analysis, with a measurement-oriented rather than causal interpretation.