Player Movement Predictions Using Team and Opponent Dynamics for Doubles Badminton
摘要
With the advancement of AI, strategic analysis in team sports has become increasingly valuable. In doubles badminton, the challenge lies in accurately predicting player movements in a fast-paced, dynamic environment. To tackle this, we introduce the novel TS-CVAE specifically designed for doubles badminton. TS-CVAE incorporates Team GAT, which leverages team influence graph over a few strokes to capture rapidly changing team strategies, and Opponent GAT, which holistically analyzes interactions between opposing players. Additionally, a new data augmentation module, SOSA, enhances the understanding of player positioning and strategy by incorporating singles data. Experimental results on real doubles badminton data show that TS-CVAE outperforms state-of-the-art sport forecasting models across multiple evaluation metrics. Visualized results also confirm that TS-CVAE’s predictions closely align with the ground truth.