A Query Language for Exploratory Analysis of Video-Based Tracking Data in Padel Matches

Area of Science:

• Sports Science
• Computer Vision
• Data Analysis

Background:

• Advances in video analysis enable automatic annotation of sports videos.
• Deep learning methods allow robust player and ball tracking in racket sports.
• Traditional data analysis methods struggle with complex spatio-temporal sports data and visualization.

Purpose of the Study:

• To address limitations in analyzing video-based sports tracking data.
• To develop a system for efficient analysis of padel match data.
• To facilitate coaches and analysts in understanding complex game strategies.

Main Methods:

• Utilized deep learning for player detection, tracking, and pose estimation from single videos.
• Developed a domain-specific query language for compact representation of spatio-temporal queries.
• Enriched data visualization by linking data points to specific video segments for detailed analysis.

Main Results:

• Created a system capable of analyzing padel match video data.
• Demonstrated the system's flexibility by converting literature-based padel strategies into readable queries.
• Provided a more comprehensive view of game data, including player motion and court positioning.

Conclusions:

• The proposed system overcomes limitations of traditional table-based methods for sports data analysis.
• The domain-specific query language enhances usability for coaches and analysts.
• Linking data to video segments provides richer insights into game dynamics and player performance.