Enhancing Event Analysis at Scale: Leveraging Tracking Data in Sports
summary
Pipeline ingests 250 MB of tracking per match with about 3.5 million coordinates and processes each game in 150 seconds per game using Python workers fed by a message queue into a data lake then Redshift and Tableau dashboards. Architecture scales horizontally with AWS spot instances and autoscaled workers via SQS and CloudWatch, reprocessing hundreds weekly without constant servers, achieving up to 90% savings. Live usage by scouts and staff shows the system actually works and reduces cost versus serverless or always‑on.
AI-written summary · about
chapters (16)
- 0:00Football data analysis intro
- 1:24Types of football analytics data
- 3:10Tracking data volume
- 3:53Pitch control models
- 4:59Pass options and EPV
- 7:51Event data with tracking
- 9:19Tracking metrics challenges
- 11:54Scaling data processing
- 14:23Cost-effective data pipeline
- 20:05On-demand vs spot instances
- 23:27Machine setup automation
- 25:13Dashboards and analytics tools
- 28:04Data for scouting
- 30:29Pregame tactical analysis
- 31:35Postgame performance insights
- 33:00Q&A
related talks
28mMoving Towards Real-Time OLAP - With First Principle Thinking and Guided by Access Patterns
24mGive meaning to 100 billion analytics events a day, analytics at Teads
56mReal Time Data Pipelines
41mThe evolution of basketball with data science
01h 01mDealing With Data Panel
48mHyper Speed! When Big Data Blooms
32mPromoting a data driven culture in a world of microservices
20mBatching vs. Streaming - Scale & Process Millions of Measurements a Second