It is still possible to beat the tennis bookmakers

A few weeks ago, I picked up the XGBoost model I built at the beginning of 2025 to predict match outcomes for the Grand Slam tournaments. In search of improving my model, I decided to read up on the latest relevant literature. Interestingly enough, I came across this paper by Lawrence Clegg that suggests there might be an exploitable market ineffciciency when it comes to intransitive matchups. He claims to have achieved returns of 3.26% ROI with Kelly staking over 1903 bets on ATP and WTA matches between January 2023 and June 2025. I'm going to try to recreate his findings in this paper, limited to ATP matches due to data availability.