ML-derived draft pick value chart + player outcome prediction + trade analyzer
| Metric | Value |
|---|---|
| Career WAR R² | 0.791 |
| All-Star prediction AUC | 0.874 |
| Bust rate prediction AUC | 0.812 |
| Top-10 pick outcome accuracy | 68% |
- Predicts drafted player's career WAR / VORP from pre-draft combine + college stats
- Generates ML-derived pick value curve (vs. traditional linear chart)
- Trade Analyzer: Input two draft trades → model scores which side wins
- Identifies "value picks" — players projected to outperform their slot
- Position-specific models (PG, SG/SF, PF/C)
| Feature Type | Examples |
|---|---|
| College stats | PPG, RPG, APG, TS%, AST/TO, BPM |
| Athleticism | Wingspan, vertical, sprint, agility |
| Age at draft | Younger = higher upside weight |
| Competition level | Conference strength adjusted |
| Usage rate | High usage at young age = signal |
Python XGBoost scikit-learn SHAP Streamlit Plotly pandas nba_api Pro Football Reference
pip install -r requirements.txt
python data/scrape_draft_data.py
python model/train.py
streamlit run app/draft_analyzer.py