Hsu, Ming-Hwa¹, Kondrič Miran², Fan-Chiang, I-Yun³, Wu, Jiunn-Lin³

¹Graduate Institute of Sports and Health Management, National Chung Hsing University
²Racket Sports Department, Faculty of Sport, University of Ljubljana
³Department of Computer Science and Engineering, National Chung Hsing University

A New Table Tennis Match Stroke Forecasting Method Using Transformer-Based Deep Neural Networks

Monten. J. Sports Sci. Med. 2026, 15(1), Ahead of Print | DOI: 10.26773/mjssm.260306

Abstract

This paper proposes a novel approach for forecasting stroke outcomes in table tennis matches using a transform- er-based deep neural network architecture. Table tennis rallies’ highly dynamic and fast-paced nature makes tra- jectory and stroke prediction a particularly challenging. Our model employs dual encoder-decoder structures to extract contextual features from rally sequences and individual players separately, addressing this issue. The mod- el uses attention mechanisms to evaluate the relative importance of stroke techniques and landing positions. Key game-specific attributes, including the ball speed and spin, are incorporated to enhance strategic insight and prediction accuracy. We further introduce a stroke-level event stream representation to convert raw match re- cords into a structured and consistent format, significantly improving interpretability and enabling more efficient analysis. A feature fusion network is employed to integrate rally dynamics and player-specific traits, allowing the model to accurately forecast the type and landing zone of the next stroke. The “Intellectual Tactical System in Competitive Table Tennis” system database provided the table tennis match data collected in this study. This database collects the data of Lin Yun-Ju's matches against male opponents (23 matches, 121 games, 2,225 rallies, totaling 10,517 hits, an average of 4.7 hits per rally). Experimental results show that the proposed architecture significantly improves prediction performance. On the dataset, it achieved top-1 accuracies of 57.2% for stroke type and 42.8% for landing zone (spot), with top-5 accuracies of 98.2% and 91.8%, respectively. Furthermore, we visualize prediction outcomes alongside known stroke data, providing a novel per- spective for tactical analysis. This visualization facilitates intuitive understanding for coaches and players, offering a valuable tool for performance evaluation and strategic development in professional table tennis.

Keywords

Table Tennis, Deep learning, Attention Mechanism, Stroke forecasting, Technical and tactical analysis

View full article
(PDF – 768KB)

References

Arevalo, J., Solorio, T., Montes-y-Gomez, M., & Gonzalez, F. A. (2017). Gated multimodal units for information fusion. arXiv preprint, arXiv:1702.01992.

Chiang, S., & Denes, G. (2023). Supervised learning for table tennis match prediction. arXiv preprint, arXiv:2303.16776. https://doi.org/10.48550/arXiv.2303.16776

Decroos, T., Van Haaren, J., & Davis, J. (2018). Automatic discovery of tactics in spatio-temporal soccer match data. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 223–232 (London, August 19–23, 2018). https://doi.org/10.1145/3219819.3219832

Giuliari, F., Hasan, I., Cristani, M., & Galasso, F. (2021). Transformer networks for trajectory forecasting. In 2020 25th International Conference on Pattern Recognition (ICPR), Milan, Italy, pp. 10335–10342. https://doi.org/10.1109/ICPR48806.2021.9412190

Huang, W., Lu, M., Zeng, Y., Hu, M., & Xiao, Y. (2021). Technical and tactical diagnosis model of table tennis matches based on BP neural network. BMC Sports Science, Medicine and Rehabilitation, 13(1), 54. https://doi.org/10.1186/s13102-021-00281-y

Hung, C. C., Chou, T. C., & Hsu, M. H. (2020). The impact of 40+ competitive table tennis connecting techniques on the tactics of high-level athletes. Quarterly of Chinese Physical Education Society of Physical Education, 34(4), 273–286. https://doi.org/10.6223/qcpe.202012_34(4).0006

Kulkarni, K. M., & Shenoy, S. (2021). Table tennis stroke recognition using two-dimensional human pose estimation. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Nashville, TN, USA, pp. 4571–4579. https://doi.org/10.1109/CVPRW53098.2021.00515

Liu, J.-W., Hsu, M. H., Lai, C. L., & Wu, S. K. (2024). Using video analysis and artificial neural network to explore association rules and influence scenarios in elite table tennis matches. The Journal of Supercomputing, 80(4), 5472–5489. https://doi.org/10.1007/s11227-023-05399-6

Liu, J. W., Hsu, M. H., Lai, C. L., & Wu, S. K. (2025). A novel framework for table tennis match analysis: combining 3S theory, video analysis, and big data analytics. Journal of Mechanics in Medicine and Biology, 25(5), 2540045. https://doi.org/10.1142/S0219519425400457

Tsai, A. L., Hsu, M. H., & Chiu, C. H. (2023). Analyzing the impact of the competitive performance of Olympic medal-winning table tennis players from the perspective of victory and defeat. Sports Coaching Science, 71, 35–50. https://doi.org/10.6194/SCS.202309(71).0004

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. In Proceedings of the 31st International Conference on Neural Information Processing Systems, pp. 6000–6010.

Wang, W.-Y., Shuai, H.-H., Chang, K.-S., & Peng, W.-C. (2022a). Shuttlenet: Position-aware fusion of rally progress and player styles for stroke forecasting in badminton. In Proceedings of the AAAI Conference on Artificial Intelligence, 36(4), 4219–4227. https://doi.org/10.1609/aaai.v36i4.20341

Wang, W.-Y., Chan, T.-F., Peng, W.-C., Yang, H.-K., Wang, C.-C., & Fan, Y.-C. (2022b). How is the stroke? Inferring shot influence in badminton matches via long short-term dependencies. ACM Transactions on Intelligent Systems and Technology, 14(1), 1–22. https://doi.org/10.1145/3545311

Wang, W.-Y., Peng, W.-C., Wang, W., & Yu, P. S. (2023). ShuttleSHAP: A turn-based feature attribution approach for analyzing forecasting models in badminton. arXiv preprint, arXiv:2312.10942.

Wu, E., & Koike, H. (2020). Futurepong: Real-time table tennis trajectory forecasting using pose prediction network. In Proceedings of the Extended Abstracts of the 2020 CHI Conference on Human Factors in Computing Systems, Honolulu, HI, USA, pp. 1–8. https://doi.org/10.1145/3334480.3382853

Zhang, Z., Xu, D., & Tan, M. (2010). Visual measurement and prediction of ball trajectory for table tennis robot. IEEE Transactions on Instrumentation and Measurement, 59(12), 3195–3205. https://doi.org/10.1109/TIM.2010.2061450