DEVELOPMENT OF A REAL-TIME AUTOMATED WRESTLING MATCH ANALYSIS SYSTEM BASED ON YOLO26X-POSE AND ATHLETE RE-IDENTIFICATION

Damir Tassybayev, Odiljon Dadabayev, Валентин Савицкий

2026

Closed access Metadata (CSL-JSON)
DOI 10.62931/2959-6335_2026_1_60
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

In this work a real-time wrestling match analysis system built on the YOLO26x-Pose detector is described. The system runs four modules in sequence: two-dimensional pose estimation for both wrestlers; appearance-based re-identification (ReID) using MobileNetV3-Small embeddings with Hungarian assignment and three-speed exponential moving average (EMA) updates; a heuristic finite-state machine (FSM) that assigns one of five match states: SEPARATION, CLINCH, TAKEDOWN_ATTEMPT, PAR_TERRE, DANGER from skeletal geometry; and a homography-based top-down minimap. The ReID module includes an entanglement-aware EMA freeze, anchor-drift protection, per-zone passivity timing, and per-frame CSV logging. On a standard GPU, without any wearable sensors, the system achieves a mean Identity Preservation Rate (IPR) of 90.1%, FSM accuracy of 85.0%, and 27.7 FPS throughput.