Digital Coaching for Tennis Serve with Machine Learning

Sammanfattning

This project develops a novel approach to analysis and understanding of tennis shot and serve performance, with the help of machine learning for digital coaching purposes. Using the PlayReplay system, two models are developed to generate shot return probabilities and quantified serve quality. Shot return probability is used to provide insight into the decisiveness of a shot and its effectiveness. Built on that, serve quality is measured as the impact of such serves on subsequent shots, allowing for a more comprehensive understanding of serve quality. Key metrics for both shots and serves are computed on multiple samples obtained by rigorous data collection in other to maximize models’ performance. Ultimately, this mentioned process allows the development of a final tool that offers specific feature analysis of a given serve, providing a suggestion on performance improvement of serve. The results show the potential of machine learning and data driven techniques to be implemented in real world scenarios and be used by players and coaches to improve player performance.