“Measuring an NBA Player’s Impact Outside the Box Score”

Abstract:Statistics has become more important in professional basketball than ever. Analyzing and comparing the performance of individual players is a challenge to everyone. For every player, the objective of a basketball game is to win rather than chasing individual statistics. We hope to take advantage of NBA play-by-play data to quantify a player’s contribution in terms of wins. In this paper, we propose a Bayesian linear regression model with shrinkage estimation to measure a player’s impact on his team’s chances of winning. In this model, we completely omit traditional statistics like points, assists, etc., nor do we consider any advanced statistical ratings like Adjusted Plus-Minus, PER, etc. This allows us to measure player’s impact without incorporating potential misleading statistics.