NFL Predictor
Project Overview
NFL Predictor is a personal project I did over winterbreak of 2025.
This was my first ever coding intensive project, but I thoroughly enjoyed it, and it sparked my interest to pursue a minor in computer science!
NFL Predictor is a machine learning model that calculates win probability of each team in a match up based on 10 years of historic data.
The model uses logistic regression to do the probability calculation based on 4 factors.
1. Home/Away Status
2. A team's performance for the past 10 years
3. A team's performance for the past 5 games
4. the other team's performance
To ensure that this model provides something of value, I created a built-in backtesting system that computes the model's performance over the past 18 weeks of NFL this season, and sums up the final accuracy.
Also, I visualized the performance of the model through plotting, and compared its performacne to a coin flip and a pro target of 60%
Although I will not be putting my own money to sports bet with this model, it was a fun project that helped me gain a bit more clarity in what career path I want to pursue as an engineer.