Unlocking Tennis Dynamics: A Hidden Markov Model Approach to Match Scoring

The aim of this paper is to develop a scoring process model for tennis matches that is applicable to other match scenarios and is able to recognize a player's performance at a specific moment in order to focus on the player's status and match flow. First, we search for specific moments and optimize the model based on the known key moments as well as assigning certain weights to the serving player based on the player's match data and performance metrics. We constructed a model of the scoring process of a tennis match based on a Hidden Markov Model (HMM) to identify specific moments. By converting the match scoring event data into a time series dataset and subsequently assigning service weights, we used the Hidden Markov Model to predict match points. Ultimately, we obtain the probability that each match point belongs to a winning state.