Searching for the GOAT of tennis win prediction

Abstract

Sports forecasting models-beyond their interest to bettors-are important resources for sports analysts and coaches. Like the best athletes, the best forecasting models should be rigorously tested and judged by how well their performance holds up against top competitors. Although a number of models have been proposed for predicting match outcomes in professional tennis, their comparative performance is largely unknown. The present paper tests the predictive performance of 11 published forecasting models for predicting the outcomes of 2395 singles matches during the 2014 season of the Association of Tennis Professionals Tour. The evaluated models fall into three categories: regression-based, point-based, and paired comparison models. Bookmaker predictions were used as a performance benchmark. Using only 1 year of prior performance data, regression models based on player ranking and an Elo approach developed by FiveThirtyEight were the most accurate approaches. The FiveThirtyEight model predictions had an accuracy of 75% for matches of the most highly-ranked players, which was competitive with the bookmakers. The inclusion of career-to-date improved the FiveThirtyEight model predictions for lower-ranked players (from 59% to 64%) but did not change the performance for higher-ranked players. All models were 10-20 percentage points less accurate at predicting match outcomes among lower-ranked players than matches with the top players in the sport. The gap in performance according to player ranking and the simplicity of the information used in Elo ratings highlight directions for further model development that could improve the practical utility and generalizability of forecasting in tennis.