Auto-updated daily
Loading...

This Week's Games

0 games
Loading predictions...

Model Controls

Real-time

Adjust model parameters. Changes update the Matchup Predictor live. Full model retraining runs daily via GitHub Actions.

ELO Parameters

K-Factor 20
Home Field Advantage (ELO pts) 65
MOV Multiplier Weight 1.0
Recent Form Blend 0.30

Game Adjustments

SOS Weight 0.50
Head-to-Head Modifier 0.50
Turnover Impact 1.0
Rest/Travel Scaling 1.0
Time Decay λ 0.10

Ensemble Weights

Logistic Regression 0.30
Linear model trained on historical win/loss matchup data. Good baseline; fast to update.
XGBoost 0.25
Gradient-boosted decision trees trained on 17 features (ELO, efficiency, rest days, travel distance, pace, passing/rushing splits, etc.). Learns non-linear interactions that a logistic model misses — e.g. a rested team with a big ELO edge matters more than either factor alone.
ELO Model 0.20
Points-based rating updated after every result. Simple, proven, and self-correcting — a team on a hot streak earns a higher rating quickly.
Pythagorean 0.15
Expected win rate from points scored vs. allowed (Bill James formula). Captures scoring dominance independent of actual W/L record — teams that outshoot opponents tend to be better than their record shows.
Efficiency 0.10
Offensive and defensive efficiency margins per 100 possessions. Pace-adjusted, so fast-paced teams aren't over-rated just for scoring volume.
Total weight 1.00
Weights auto-nudge after 50+ resolved games toward whichever models have been most accurate recently. Use the button below to apply that auto-learned adjustment now.

Matchup Predictor

Any Teams
🏈

Select Two Teams

Choose a home and away team, then click Run Prediction.

ELO Leaderboard

All 32 Teams
Click column header to sort
Rank Team ELO ▲ ▼ ±Band Record Pythagorean Net Eff Off Eff Def Eff Playoff % SB % Trend

Model Performance

Loading...
Calibration Curve
Historical Accuracy by Season
Loading prediction log...