Inspiration

I wanted to explore process mining techniques on basketball but couldn't find an appropriate dataset. The closest, is the play-by-play data but this is heavily stripped down with a focus on traditionally captured activities such as shot, rebound, or steal. Activities which are frequent like pass and dribble were not captured .

Labour of love

Manually enriching play-by-play data for one game took me about a month. It involved watching a basketball game at 0.25 speed and pausing several times to capture timestamp.

Description

The data is from the NBA final 2024, game 5. The finals winner is the team to win 4 games first, the max amount of games that could be played is 7 but Boston beat Dallas in 5 games. The data covers the winning game.