WAR in Pieces: A Bottom-Up Approach to Player Evaluation in the NBA

Abstract

Basketball is a dynamic sport that requires players to excel in a wide range of skills, including scoring, rebounding, passing, and defense. Traditionally, players have been evaluated based on simple metrics such as points, rebounds, and assists, which fail to capture the full complexity of their performance on the court. Moreover, these metrics do not account for the many contextual factors that can influence a player's performance, such as the quality of their opponents, the pace of play, and the role they play on their team. In recent years, a new generation of metrics have been created to address these limitations. These combine simple metrics with previously untracked aspects of the game to provide a more comprehensive and accurate assessment of a player's performance. Even so, all are flawed in some way, particularly in their failure to meaningfully capture player-player interactions.

In this paper, I create a new metric called Wins Above Replacement (WAR) that seeks to address those shortcomings and provide a more nuanced and reliable evaluation of player performance. The key idea is to create predictions at the lineup level and to rebuild the season from the ground up. I describe the full methodology used to calculate WAR, which is inspired in part by the baseball metric of the same name, and I present the results of its application to NBA players in the past six seasons. The results loosely agree with preexisting evaluation metrics while also raising questions about new ways to think about evaluating players. Overall, this work represents a significant advance in the field of basketball player evaluation and has the potential to revolutionize the way that players are scouted, evaluated, and developed.