| Predicting NBA Players’ Starting Probability Using Machine Learning Algorithms |
| Paper ID : 1527-SPORTCONGRESS |
| Authors |
|
Ahmad Mahmoudi *1, زهرا اکبری2, سحر بیدختی نژاد3 1AssistanAssistant Professor, Department of Sport Management, Faculty of Sport Sciences and Health, University of Tehran, Tehran, Iran.t Professor 2گرایش مدیریت رویداد ها و گردشگری ورزشی، دانشکده علوم ورزشی و تندرسی، دانشگاه تهران، تهران، ایران 3گرایش مهندسی سیستم های محیط زیست، دانشکده تحصیلات تکمیلی محیط زیست، دانشگاه تهران، تهران، ایران |
| Abstract |
| Introduction: Data-driven approaches have become essential for sports management and team strategy optimization. This study aims to predict basketball players’ likelihood of being in the starting lineup using statistical performance data from the 2023–2024 NBA season. By employing machine learning algorithms, coaches can make informed decisions on player selection and reduce subjective bias(Horvat & Job, 2020). Methods: The dataset was obtained from the public Kaggle repository “2023–2024 NBA Player Stats.” Players starting in at least 50% of their games were labeled as main players (1), and others as bench players (0). After removing duplicates and normalizing numerical and categorical variables, two classification algorithms — Random Forest and XGBoost — were trained using an 80/20 train-test split. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics(Breiman, 2001; Chen & Guestrin, 2016; Vivovinco, 2024). Results:Both models achieved high predictive accuracy, with Random Forest reaching 0.974 and XGBoost outperforming it at 0.991. XGBoost demonstrated superior generalization, with precision and recall values exceeding 0.97 for both classes. According to the XGBoost output, as expected, the result identified the top five players as the optimal lineup, while others were classified as substitutes. Conclusion:The XGBoost model proved to be a robust and reliable tool for predicting basketball player performance and lineup potential. Its ability to capture nonlinear patterns in player statistics supports data-driven coaching strategies. The study highlights the value of integrating artificial intelligence into sports analytics to enhance tactical decision-making and optimize player management. |
| Keywords |
| Machine Learning Algorithms, Basketball, Sport Management, NBA, Team Selection |
| Status: Abstract Accepted (Oral Presentation) |
