Playground

This repository is a lightweight playground for exploring how flagrant fouls correlate with NBA outcomes. The flagrant_fouls/ subdirectory holds the core dataset plus the notebooks used for data collection and the current point differential regression.

Layout

• flagrant_fouls/nba_flagrant_fouls.csv – one row per game, recording the home/away teams, flagrant fouls, and final scores.
• flagrant_fouls/data_collection.ipynb – helper notebook that pulls box-score data from the NBA API.
• flagrant_fouls/point_differential_analysis.ipynb – linear regression modeling of point differential against foul counts.
• AGENTS.md – operating guidelines for contributors (especially AI agents).

Data snapshot

• Dataset stored in flagrant_fouls/nba_flagrant_fouls.csv; refer to that file for the current row count and season coverage.
• Seasons captured: game IDs with leading prefixes 0012, 0022, 0032, 0042, and 0052.
• Games with no play-by-play data are tracked in flagrant_fouls/nba_skipped_games.csv.
• Schema: game_id, home_team, away_team, home_flagrants, away_flagrants, home_score, away_score.

Getting started

1. From the project root, install dependencies with uv sync.
2. Inspect AGENTS.md for repository conventions.
3. Change into flagrant_fouls/ and open the notebooks with jupyter notebook (or lab).
4. Re-run point_differential_analysis.ipynb any time the CSV is updated.