Identifying playstyles in football through spatial networks
summary
Claims an automated pipeline converts tracking and event data (25 Hz) into spatial networks, computes seven metrics like convex hull, opponent/covering degree and betweenness, and clusters ~2500 shot moments with fuzzy C-means to yield five attacking and five defending playstyle fingerprints plus a low‑confidence bucket. Results show per‑team cluster preferences, co‑occurrence matrices with goal rates, and domain‑expert video checks for validation, so the method actually produces interpretable playstyle labels and actionable summary statistics.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:56Football analysis limits
- 3:41Play styles before shots
- 4:45Data types and prep
- 6:08Player positions to networks
- 7:11Metrics for spatial networks
- 9:01Betweenness centrality explained
- 10:27Convex hull area metric
- 11:09Metric trends before shots
- 12:52Selecting metrics statistically
- 13:54Clustering play styles
- 15:38Play style fingerprints
- 16:21Validating with experts
- 18:07Attacking and defending styles
- 19:11Team tactical identity
- 20:34Play style consistency
- 22:01Summary
- 23:06Question and answer session
related talks
29mFootball (soccer) data analysis: A pedagogic introduction
32mCutting Edge Football Analytics in Python
04mInteractive Sports Analytics
14mModeling Individual and Team Behavior through Spatio-temporal Analysis
41mThe evolution of basketball with data science
19mFootball Analytics with the Soccermatics Package
35mFootball video analysis using Deep Learning
27mAnalysis of sports scenes using computer vision models