Football Predictions — Manuel Cattoni

Football PredictionsA data-driven betting portfolio optimizer for top European football leagues.This project generates sustainable weekly profits using advanced ML predictions across the top 5 European leagues. It incorporates various statistical indicators and team form to compute outcome probabilities and optimize a betting strategy.

Stacking Ensemble Model: combines Random Forest, AdaBoost and Gradient Boosting with a Logistic Regression meta-model, calibrated for reliable probabilities and reaching 81.9% training accuracy on 2025 data.

27-Feature Engineering Pipeline: blends classic metrics like ELO ratings and form with newer additions such as Expected Goals (xG), rest-day analysis, head-to-head patterns and exponential momentum decay.

Tiered Betting Portfolio: allocates each week's stakes across four risk tiers, from Ultra Safe to high-odds picks, sized with Kelly Criterion fractional staking and a built-in probability safety margin.

Tech stack: Python, scikit-learn, pandas, football-data.org APIA project by Manuel Cattoni.