An assessment measurement study based on the momentum of the tennis ball in question

Abstract

In tennis, it is very important to always be aware of the movement and trend of the match. Accurately grasping the change of momentum is of great practical significance in predicting the future fluctuation of the match. Based on the relevant match data provided by Wimbledon, three models are constructed in this paper. Model one is a win-loss probability model based on binomial distribution, which deflates the impact of each small game win on the whole game, and accurately calculates the overall win rate of current players by judging the difference in players' level as well as the number of games won and lost. Model 2 is a momentum calculation model, which accurately describes the player's game status by analyzing the number of consecutive scores, score gap, number of key balls, and number of errors in a multi-dimensional analysis, and then accumulating and calculating the momentum data. Model 3 is a match fluctuation prediction model constructed based on Gradient Boosting Tree, which is utilized to predict match fluctuations by comparing predicted fluctuations with actual fluctuations on the test set in the tournament. And it evaluates which factor is the most important in predicting the future trend.