Comparing predictive models for strategic insights in volleyball match prediction

This study predicts volleyball match outcomes using Monte Carlo simulation and machine learning models with actual match data from the Korean Professional Volleyball V-League. Using 43 performance metrics from three seasons (2021-2024), four predictive frameworks were constructed: Monte Carlo simulation and three machine learning approaches (independent, paired, and variable selection). Monte Carlo simulation achieved the highest accuracy, particularly for women's matches. The paired Random Forest model with variable selection (Model 4) showed the lowest Expected Calibration Error (ECE), indicating excellent predictive reliability with fewer variables. Variable importance analysis revealed gender-specific factors, highlighting the need for customized strategies. This study demonstrates the value of integrating simulation and machine learning for match analysis and provides practical insights for data-driven tactical development.