Construction and Validation of a Dynamic Prediction Model for the Event Process under the Influence of Multiple Factors

In sports technology advancement, this study explores real-time performance metrics of tennis players through mathematical modeling and data analysis. We introduce a novel indicator system considering service score factor, physical and mental states. Using logistic regression and machine learning algorithms, LGBM was selected for its highest predictive accuracy. Hypothesis testing and Spearman correlation analysis (coefficient 0.479) disproved the coach’s assumption about random momentum. The model predicts game-wise performance variance and demonstrates broad applicability across various competitions. Testing shows individual ability is critical for real-time performance prediction, and recommendations for coaches are provided.