Momentum Evaluation and Situation Prediction Model Based on Integrated Machine Learning Model - a case of tennis match

The quantitative analysis of momentum is of great guiding significance to the adjustment of competition strategy and state of coaches and athletes in the field of sports. However, the existing momentum research is mainly explained from the perspective of economics, psychology and other theories, without quantitative analysis. Or after quantitative analysis, the model can only predict the results of the whole game, and cannot accurately predict the changes during the game. Therefore, using data from the 2023 Wimbledon men's singles final as a data set, we propose a momentum evaluation model, a state prediction model and an integrated machine learning model composed of XGBoost, LightGBM, GBDT, to resume the match flow, identify which player perform better at a specific time, and predict state fluctuation. Finally, the 2023 Wimbledon men's singles final match is perfectly visualized and the state fluctuation at each time in this match is accurately predicted. The results suggest that models we established have high prediction accuracy and high stability.