Quantifying player versatility through lineup-based playing styles in elite basketball.

This article presents a comprehensive framework to quantify player versatility in elite basketball by analyzing variations in playing styles across different lineups. Using data from 11,978 NBA games over 10 seasons, the study applies Non-negative Matrix Factorization to reduce 15 performance metrics into six distinct playing style bases, followed by a two-step clustering algorithm based on cosine distance to identify typical player performance patterns within lineups. To compare versatility between players, the Wasserstein distance from Optimal Transport Theory is employed, accounting for both playing style distributions and playing time across lineups. This approach offers practical applications for coaches and management in lineup optimization, player development, recruitment, and injury management, while acknowledging limitations related to data availability and the treatment of absolute performance contributions.