NFL player prop projection model. One of the edge-models family.
Give it a player's recent game log and a betting line. It returns a calibrated probability that the player goes over, plus a quarter Kelly stake. Trained on 5.0M synthesized lines pulled from real NFL game logs.
Held out test split (1.04M rows that the model never trained on):
| accuracy | brier | log loss | baseline | lift |
|---|---|---|---|---|
| 94.7% | 0.046 | 0.184 | 0.252 | +26.9% |
Lift is how much lower the calibrated log loss is than a no-skill baseline (0.068 absolute, 26.9% relative). Strongest of the lot. NFL prop lines move slowly and the books are soft, so the calibrated model has the most room here. One game a week also means recent form is a strong signal.
git clone https://github.com/LeSingh1/edge-nfl
cd edge-nfl
node src/predict.js --stat "Pass Yards" --line 248.5 --log 270,233,251,288,240
Output is the projection, the raw over probability, the calibrated probability, and a suggested stake.
- Weight the recent games up and fit a mean and standard deviation.
- Turn the line into a raw over probability with a normal model.
- Correct that probability with the trained isotonic calibrator in
models/calibration.json. The correction is clamped to +/-0.20 so a thin bucket cannot fake confidence. - Size the bet with quarter Kelly.
The calibrator is keyed by stat type, so "Pass Yards" gets a different correction than other stats.
The scraping pipeline and the raw training rows live in the private app this came out of. This repo ships the trained model and the code to run it, which is the useful part.
MIT. Not financial advice. The house edge is real, bet responsibly.