Uncovering Gender-specific Technical-Tactical Demands in Elite 3x3 Basketball Using XGBoost and SHAP

This study explores the application of artificial intelligence in predicting 3x3 basketball game outcomes and identifies the key tactical and technical indicators that determine victories in both men’s and women’s 3x3 basketball games. Using game data from all international high-level 3x3 basketball competitions from 2019 to 2024, the study applies the XGBoost algorithm to build prediction models for men’s and women’s games and uses the SHAP algorithm to effectively quantify the key tactical and technical factors influencing game outcomes. The study finds that: (1) In both men’s and women’s games, two-point field goal percentage, successful two-point field goals, and successful one-point field goals are critical indicators affecting game outcomes. (2) In men’s games, free throw percentage and assists play a very important positive role in winning. (3) In women’s games, turnovers, unsuccessful one-point field goals, and defensive rebounds have a crucial positive impact on winning. (4) Unsuccessful two-point field goals have a negligible effect on the outcomes of both men’s and women’s 3x3 basketball games. This study employs the integrated machine learning algorithms XGBoost and SHAP to construct game outcome prediction models for different genders, showing excellent performance and strong interpretability, thereby providing coaches with more targeted tactical strategies and critical decision support.