Portfolio / NBA analytics
NBA analytics
An end-to-end pipeline that scraped 1,230+ NBA games and found statistically significant in-game patterns.
1,230+games across 5 seasons
127engineered features
6patterns found at 82–93% confidence
The problem
Finding real, repeatable patterns in sports data means collecting large amounts of game data and testing ideas properly, not just eyeballing charts.
What I built
- Web scrapers collecting five seasons of game data
- 127 engineered features per game
- Statistical testing to separate real patterns from noise
- An interactive dashboard to explore the results
The result
Six statistically significant patterns, from a pipeline that can be rerun as new games are played.