G-RAPM: revisiting player contributions in regularized adjusted plus-minus models for basketball analytics

Identification and analysis of players ranking have a central role in the sports analytics. An essential tool in this framework is the Regularized Adjusted Plus-Minus (RAPM) model. When player and lineup effects are included simultaneously, the interpretation of the RAPM model results can be cumbersome. The present work aims at estimating a modified version of the RAPM model, adopting a one-sided assumption for player effects. The proposed specification allows for a direct performance interpretation. The model can be estimated feasibly within the Bayesian framework, allowing for straightforward generalisations.