Project Overview
This project leverages historical college football data to predict game outcomes with high fidelity.
- Data Source: CollegeFootballData API
- Performance: Averaging 73% Accuracy week-to-week.
- Taught the model on metrics such as EPA on offense and defense, PPA per play type, ELO ratings, Season player and team stats, and talent composit ratings
- Implemented Linear regression for game probabilities and used the logistic regression model for spread projections
Linear Regression Logic
ŷ = β₀ + β₁(Elo_diff) + β₂(Off_PPA_diff) + β₃(Def_PPA_diff) + β₄(Home_Field)
View Source Code on GitHub →
Model Outputs & Visualizations
Explore the real-time projections and historical performance charts generated by the pipeline.
Win Probabilities
Raw outcome probabilities for upcoming games.
View Data →Spread Projections
Actual vs. Predicted performance charts.
View Analysis →Logic Visualization
Click image to expand
