Josh Marvald, '21, is majoring in mathematics and statistics with the intention of applying to statistics PhD programs. Josh is also working with Dr. Matt Higham this semester on a continuation of their fellowship project. They are looking into different machine learning methods to predict the outcome of tennis matches.
This project is an attempt to provide an analysis of the serve in professional tennis and its association with win probability. We completed extensive data exploration and wrangling to both find important trends and to shape the data into usable forms to build models. The final model is a Bradley-Terry model that focuses on how first serve percentage (percent of times the first serve is made) is associated with win probability for different players. After completing the model building and testing, we built a Shiny app that allows users to plot the fitted Bradley-Terry model lines for players of interest and specific opponents. The app allows users to examine how the association between the probability of winning a match and first serve percentage changes for different players and first serve percentages.