Soccer Match Prediction

This project tested whether predicting exact goal counts could outperform the standard win/loss/draw classification approach for football match forecasting. Regression models were built in PyTorch and evaluated against classification baselines using Monte Carlo betting simulations.

The regression models showed some promise for match outcome prediction, but goal margin error made them unreliable for real betting profitability. The paper covers the model architecture, evaluation methodology, and visualization of prediction accuracy against simpler strategies.