A Study of Race Momentum Prediction Based on LSTM Modeling

The aim of this study is to develop an advanced predictive model to capture and predict momentum changes in tennis matches to provide strategic advice to players. By analyzing the 2023 Wimbledon men's singles final, we designed an LSTM-based machine learning model to effectively capture momentum changes during the match and generate visual graphs reflecting the changes. Statistical methods demonstrate the non-randomness of score changes, and heatmaps are utilized to show statistical evidence of the existence of momentum. In addition, an LSTM model is derived through Gray's correlation analysis to identify key features affecting momentum and personalized recommendations are made to adapt to different match scenarios. This study provides new perspectives for understanding and predicting patterns of momentum changes in high-level tennis matches, which can help players and coaches to better control the match dynamics and improve the overall competitive level.