Modeling NBA Salaries: A Multiple Linear Regression Approach to Analyzing Player Performance and Predicting Earnings

Abstract

This study investigates the key factors influencing NBA player salaries using a multiple linear regression framework applied to performance and salary data from 1,134 NBA players across the 2016–2019 seasons. The analysis examines the relationship between salary and a range of player attributes, including age, playing time, scoring performance, shooting accuracy, and positional role. Preliminary regression diagnostics revealed violations of linearity, constant variance, and normality assumptions, which were addressed through response transformations, with the Box–Cox transformation providing the best overall improvement. Both manual variable reduction and automated forward selection based on AIC were implemented to refine the model and improve interpretability. The final selected model identified player age, minutes played, and points per game as the strongest positive predictors of salary, while shooting efficiency and player position also demonstrated significant influence. Model evaluation using adjusted $R^2$, AIC, BIC, and VIF indicated that the refined models substantially improved performance while maintaining acceptable multicollinearity levels. The findings suggest that consistent playing time, scoring ability, and accumulated experience are strongly associated with higher earnings in the NBA. In addition to statistical analysis, the study discusses ethical considerations surrounding manual and automated model selection, emphasizing the importance of balancing interpretability, fairness, and bias mitigation in sports analytics research.

Dataset Used

Dataset

Reference

[1] Nikos Chatzistamoulou, Kounetas Kostas, and Antonakis Theodor. “Salary Cap, Organizational Gap, and Catch up in the Performance of NBA Teams: A Two-Stage DEA Model Under Heterogeneity”. In: Journal of Sports Economics 23.2 (2022), pp. 123–155. doi: 10.1177/15270025211022253. url: https://doi.org/10.1177/ 15270025211022253.
[2] Chiang-Ping Chen Chih-Hai Yang Hsuan-Yu Lin. “Measuring the efficiency of NBA Teams: Additive Efficiency Decomposition in two-stage DEA”. In: Annals of Operations Research 217.1 (2014), pp. 565–589. doi: 10.1007/ s10479-014-1536-33.
[3] Kevin Sigler William Compton. “NBA Players’ Pay and Performance: What Counts?” In: The Sport Journal 21 (2018). url: https://thesportjournal.org/article/nba-players-pay-and-performance-what-counts/.
[4] Riccardo Rubini Davide Ratto Alfredo Galli. NBA PLAYERS 2016-2019. Accessed: 2024-12-05. 2019. url: https: //www.kaggle.com/datasets/davra98/nba-players-20162019.
[5] Justin Ehrlich, Shane Sanders, and Christopher J. Boudreaux. “The relative wages of offense and defense in the NBA: a setting for win-maximization arbitrage?” In: Journal of Quantitative Analysis in Sports 15.3 (2019), pp. 213–224. doi: doi:10.1515/jqas-2018-0095. url: https://doi.org/10.1515/jqas-2018-0095.
[6] Sports Reference LLC. Basketball Reference. Accessed: 2024-12-05. url: https://www.basketball-reference. com/.
[7] Rodney Paul, Andrew Weinbach, and Jeremy Losak. “Market Efficiency and the Setting of Salaries for NBA Daily Fantasy on FanDuel and Draft Kings”. In: The Journal of Prediction Markets 14.2 (Dec. 2020), pp. 45–60.
[8] John Robst et al. “Skin Tone and Wages: Evidence From NBA Free Agents”. In: Journal of Sports Economics 12.2 (2011), pp. 143–156. doi: 10.1177/1527002510378825. url: https://doi.org/10.1177/1527002510378825.
[9] Vangelis Sarlis and Christos Tjortjis. “Sports Analytics: Data Mining to Uncover NBA Player Position, Age, and Injury Impact on Performance and Economics”. In: Information 15.4 (2024). issn: 2078-2489. doi: 10.3390/ info15040242. url: https://www.mdpi.com/2078-2489/15/4/242