🏈 NFL Prop Predictor

A machine learning tool that predicts the probability of NFL players hitting their prop lines using XGBoost and normal distribution modeling.

Features

• Supports RB, WR, TE, and QB positions
• Pulls 4 seasons of real NFL data (2022-2025)
• Engineers 7-12 predictive features per position including:
  • Weighted recent averages (L5 games)
  • Usage rates (carries/targets/attempts)
  • Opponent defensive ratings
  • Home/away splits
  • EPA efficiency metrics
  • Target share and WOPR (WR/TE)
• Outputs over/under probability using normal distribution
• Interactive web UI built with Streamlit

Model Performance

Position	MAE	R²
RB	19.58 yards	0.434
WR	21.71 yards	0.311
QB	63.66 yards	0.317

How to Run

Install dependencies

pip install -r requirements.txt

Run the web app

streamlit run app.py

Run the CLI version

python nfl_prop_predictor.py

Example Output

Enter a player, opponent, and prop line to get:

• Predicted yards
• Over/under probability
• Normal distribution visualization
• Defensive context

Tech Stack

• Python 3.11
• XGBoost
• Scikit-learn
• Pandas / NumPy
• SciPy
• Streamlit
• nflreadpy