A machine learning-based decision model for optimal return height in tennis

The tactical role of return height in tennis rallies has been largely overlooked in existing research, which tends to focus on placement or overall outcomes. This study addresses this gap by developing a machine learning based decision model to optimize return height in prolonged rallies. Using data from seven matches of the 2024 Australian Open Men’s Singles Champion, 7,200 shot-level records were extracted and analyzed. Three algorithms—support vector machine (SVM), artificial neural network (ANN), and random forest (RF)—were applied to predict optimal return heights, with RF achieving the highest accuracy (84.5%) and F1-score (94.0%). Furthermore, a personalized model incorporating K-means clustering improved accuracy to 81.2% for specific player styles. Feature importance and sensitivity analysis confirmed that return height is a key tactical determinant. This study contributes a data-driven framework for analyzing return strategy in tennis and provides practical implications for coaching, player training, and real-time tactical support.