Econometric framework for analyzing NBA player contracts through quantitative methodologies

Abstract

NBA salaries depend not only on talent but also on a mix of future on-court performance, team strategy, and market factors. One contract can redefine a franchise, and therefore, salary forecasting is complex and imperative. In this work, the on-court performance, position, experience, market factors, and popularity that drive NBA salaries are examined in detail. All these factors together make for a challenging environment for the analysis of sports data, with added complexity in player contracts. To overcome such a challenge, three types of regression models have been utilized: Adaptive Boosting Regression (ADA), Bagging Regression (Bagging R), and Decision Tree Regression (DTR). All three have individual strengths in extracting salary trends from past performance, specifications in player contracts, and player performance on the court. As an additional tool, the Wild Geese Algorithm (WGA) is employed to optimize the parameters of each of the three models. Optimizing these parameters leads to improvements in both predictive accuracy and convergence, resulting in more reliable salary predictions. In the two cases of hybrid models, one ranks first (ADWG with 0.989), and the other ranks at a poor 0.927 (DTR). In addition, ADWG obtained an RMSE of 2006791, which was the lowest compared to other models. By combining sophisticated regression techniques with optimization methods, this work aims to gain a deeper understanding of NBA salary dynamics. The findings will help franchisees make informed financial decisions and offer insights into how player value is quantified in the modern NBA, where market forces and team strategies constantly evolve.