RESEARCH-BASED MODEL

Stacked ensemble extending Allen & Savala (2025) — Logistic Regression, XGBoost, and MLP Neural Network base models combined via a learned meta-learner, with isotonic calibration for reliable win probabilities.