70 likes | 183 Vues
This study introduces a novel approach to ranking NBA players based on their individual offensive and defensive skills through a Bayesian framework. Using data from a complete game between CLE and DET on March 7, 2007, we analyze player performance over traditional box scores, focusing on offensive skills (measured to be higher the better) and defensive skills (measured to be lower the better). Our findings suggest that smoothing and Bayesian priors can significantly improve player assessment. Future work will extend this methodology to more comprehensive datasets, enhancing player ranking accuracy.
E N D
NBA: Bayesian Skill Ranking Leland Chen, Joseph Huang, Ryan Thompson
Gaussian Logistic
Case Study Offensive/Defensive Skills Offensive Skills: Higher the better Defensive Skills: Lower the better 1 complete game; 183 total possessions • Traditional Box Score CLE @ DET, March 07, 2007. 101-97 OT.
Future Work Conclusion New way to rank players based on individual contributions • Smoothing/Bayesian Prior • Coach Skill • Even More Data
Southwest Division 2008-2009 Top 2 pt Off./Def. Players (300+ possessions) Offensive Skills: Higher the better Defensive Skills: Lower the better 35 games; 3152 1st half possessions