A Study on Momentum Prediction Based on Random Forest Model and Monte Carlo Simulation

In recent years, more and more attention has been paid to the study of players' "momentum". In this paper, in order to assess the performance of players at specific moments, we constructed an assessment index based on feature engineering and factor analysis using hierarchical analysis (AHP) and verified its feasibility. In addition, in order to verify the effect of momentum on the game results, this study used Monte Carlo simulation to generate the results that are not affected by momentum based on the statistical characteristics of the relevant factors in the original data. The Kolmogorov-Smirnov (K-S) test was then used to test the difference in distribution between the simulated and original data. Subsequently, the problem was transformed into a machine learning classification task in order to explore the dynamics of race transitions and their influencing factors. In order to fully account for the temporal characteristics of the data, a sliding window was employed to track the frequency of significant events and scoring situations at each time point over a recent period of time, which were then predicted using a random forest model.