An experiment to build a play by play win probability model.
Inspired by a win probability model article at Medium.
Uses logistic regression to model wins at each point of the game based on the score, field position, home field advantage, distance to the first down, time remaining, etc.
The chart displays the output of the model (black), the built-in nflfastR win probability model (blue), scoring events (lines at bottom), and other reference lines (50% probability, times of quarters, etc.).
The nflfastR is more accurate because it also uses:
- Expected points
- Other attributes
My accuracy:
