A Study of Momentum in Tennis Based on Multiscale Momentum-Success Test Model and Swings Prediction Model

In tennis, "momentum" is one of the most important factors affecting the results of the game. First, this paper establishes the Multiscale Momentum-Success Test Model and calculates the proportion of the winning side’s momentum at four different scales (score, game, set and match). The proportions of the four scales are 70.3%,78.1%,84.6%,93.5%. Therefore, it can be proved that momentum plays a role in the game. Also, the fluctuation and success in the game are not random. Afterward, this paper establishes the Support Vector Machine Model (SVM) and the Feed-Forward Neural Network Model (FNN). Three correlations are analyzed for factors such as scoring ratio, serve, consecutive scores (lost points), highlight scores, major lost points, and physical condition. It is concluded that the serve, consecutive scores (lost points), highlight scores, and major lost points will play a role in the occurrence of Swing Points. By the two models, we create the Swings Prediction Model. The prediction accuracy of SVM and FNN Model are 84.85% and 67.4%. Finally, based on the momentum changes, match suggestions can be made for the coaches.