Portal de Periódicos da CAPES

Research On the Momentum Scoring System Model

2024; Volume: 103; Linguagem: Inglês

10.54097/yfhqdd84

2791-0210

Songtao Che,

Power Systems and Technologies

In the 2023 Wimbledon men's singles final, 20-year-old Spanish rising star Carlos Alcaraz defeated Grand Slam player Novak Djokovic to claim the title. In sports, reversals and reversals often occur, while in tennis, this shift is even more pronounced due to the presence of hold and break games, a trend known as momentum. The purpose of this paper is to study the internal mechanism of momentum, understand the causes and influencing factors of momentum, so as to help players and coaches understand the game format from another perspective, so as to improve the level of competition. In response to problem one, a model was developed to capture the flow of the game when scoring occurs and apply it to one or more games. To solve this problem, build a time-series-based model, the Momentum Scoring System model, which will evaluate each point in the match, taking into account the advantage of serve and the impact of consecutive points. The goal of the model is to determine which player is performing better at any point in the game and to quantify the degree of dominance in that performance. In response to question 2, in order to evaluate the tennis coach's skepticism about the role of "momentum", a random process simulation was used to compare the actual match data with the randomly generated match results, and the randomness of the momentum transition was evaluated from the actual momentum score calculation and the random match simulation. Firstly, the momentum score is calculated based on the actual game data, then the distribution of momentum scores is generated by simulating a large number of random games, and finally the statistical hypothesis test is used to judge whether the momentum transition in the actual game is significantly different from the random case by using the null hypothesis (H0) and the alternative hypothesis (H1). In response to problem 3, in order to predict the change of momentum in a match (i.e., when the flow of the game will change from one player to another), a machine learning model is used to analyze the match data and identify possible turning point indicators. Firstly, a gradient boosting tree classifier was trained to predict the transition point of momentum in the game, and then the accuracy of the model on the test set was calculated, and the prediction performance of the model was visualized through the confusion matrix. The confusion matrix provides a comparison between the true class and the model prediction class, helping to understand how the model is performing at the point where the momentum of the prediction is turning. The goal of this paper is to evaluate the generalization ability of the previously developed model on other competition data and explore possible directions for improving the model. In terms of evaluating the generalization ability of models, we focus on data preparation, feature engineering, model testing, and performance evaluation. At the same time, the potential directions for improving the model are also analyzed: 1. More complex characteristics: player fatigue, psychological pressure at key moments of the game, etc. 2. Model tuning: Adjust model parameters according to different types of datasets, or try different machine learning algorithms. 3. Cross-motion generalization: Analyze the applicability of the model in different ball sports, and explore whether it can capture the general momentum transformation law in cross-motion.