The evolution of basketball with data science
summary
Chronicle of claims and systems not buzzwords the conference lists the shift from box scores to camera tracking and the concrete effects: SportVU player tracking gives X Y coordinates at 25 frames per second, inflating ~400 rows to roughly a million rows per game, data lands about an hour after play, teams build Python pipelines on AWS, models that infer defender assignments and cluster plays produce actionable metrics (shot charts showing Rockets taking 13% long twos vs 27% league average) and drive more threes though in game use is limited by processing latency.
AI-written summary · about
chapters (10)
- 0:00Welcome!
- 0:20How to think about basketball with stats
- 3:02Analytics in baseball
- 6:35History about Don Knuth (Electronic Coach)
- 8:54Collected data in 1962 (just the boxscore)
- 10:20Collected data since 1996 (play-by-play)
- 13:56Early development in 2000s of tracking players in-game
- 15:31Spatio-Temporal Data
- 22:53Defensive Metrics
- 26:32Questions
related talks
04mInteractive Sports Analytics
41mEnhancing Event Analysis at Scale: Leveraging Tracking Data in Sports
42mDeep Learning for Automatic Basketball Video Production
42mHow Deep Learning is Revolutionizing Sports Analytics
51mVideo Analytics for Football games
29mAI in Sports - Basketball Rule Violation Detection
56mSports Analytics: Big Data in the Big Time
31mBeyond a Bit Fit