Forecasting NBA Players’ Efficiency: Integrating College and Combine Predictors with Beta Regression

Abstract

This study uses beta regression and random intercept beta regression to model normalized Player Efficiency Rating (PER) in the NBA based on pre-draft data from college basketball and the NBA Combine. Using 463 players with complete statistics, we identify key predictors of efficiency, including position, weight, rebounds, field goal percentage, height, and vertical leap. The models account for the bounded nature of PER and capture longitudinal trends with player-specific random effects. Results highlight positional role as the strongest determinant, with guards and wings outperforming centers.