DyMF-E : encoder-only dynamic graph fusion for win-probability estimation in badminton rallies

Name: Chen, Yen-Hsin

Organization : Chulalongkorn University. Faculty of Science

Email : yenhsinchen09@gmail.com

Name: Naruemon Pratanwanich

Organization : Chulalongkorn University. Faculty of Science

Email : naruemon.p@chula.ac.th

LCSH: Badminton -- Statistical methods

LCSH: Sports -- Data processing

LCSH: Graph theory

LCSH: Machine learning

Abstract: Accurate rally-level win-probability estimation and stroke-wise attribution assist coaches in identifying momentum-shifting shots. Prior shot-influence approaches provide per-stroke attribution but (i) assign credit unilaterally to the hitter and (ii) rely heavily on handcrafted features, limiting scalability. Meanwhile, dynamic graph encoders for movement forecasting explicitly model player–player interactions and evolving play styles; however, they primarily focus on continuous trajectories rather than discrete rally outcomes and do not inherently provide player-aware, stroke-wise attribution. This highlights the need for an interaction-aware, defender-inclusive, and feature-efficient model specifically designed for badminton rallies. This study introduces DyMF-E, an encoder-only dynamic graph network that integrates a relational player–stroke graph with a dynamic player–player graph to generate rally-level win-probability trajectories and shot-level influence through a lightweight prediction head. Experiments conducted on a professional singles dataset show that DyMF-E achieves an AUC above 0.80 with strong calibration performance. The model generalizes effectively from validation to a held-out test set and maintains competitive performance with a reduced four-input Lite-4 variant. This variant lowers data collection requirements while preserving interpretability for performance analysis

King Mongkut's University of Technology North Bangkok. Central Library

Address: BANGKOK

Email: library@kmutnb.ac.th

Created: 2025

Modified: 2026-02-24

Issued: 2026-02-24

บทความ/Article

application/pdf

BibliograpyCitation : In Prince of Songkla University, Phuket Campus. College of Computing. The 9th International Conference on Information Technology (InCIT 2025) (pp.222-229). Phuket : Prince of Songkla University

eng

©copyrights King Mongkut's University of Technology North Bangkok

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	InCIT 2025pp.222-229.pdf	1.21 MB

ใช้เวลา

0.027641 วินาที