Methodology
Stacking ensemble (XGBoost, LightGBM, Ridge) trained on 6 seasons of NBA data with 260+ engineered features — positional defense, teammate impact, travel fatigue, shot tracking, pace adjustments, and advanced stats. Quantile regression models estimate full outcome distributions for model-derived win probabilities. Bet sizing uses volatility-adjusted fractional Kelly Criterion with same-game correlation caps. Walk-forward retraining keeps the model current. Closing line value (CLV) tracked to validate edge quality.
Disclaimer
For informational and entertainment purposes only. Not financial or betting advice. Past performance does not guarantee future results. Must be 21+. Gambling involves risk — only wager what you can afford to lose. Full disclaimer