College Football
Methodology
Each competition has its own independent model, weights, parameters, versions and performance history. Nothing is shared or copied between College Football and the NFL.
Active model version: 1.3.0 published — engine revision gda-engine-1. Predictions are enabled.
Rating construction
A team's base GDA Rating is a weighted combination of category grades. Weights must total exactly 100%.
| Category | Weight |
|---|---|
| coaching | 7.0% |
| secondary | 10.0% |
| linebackers | 7.0% |
| quarterback | 15.0% |
| receiver te | 8.0% |
| recent form | 4.0% |
| running back | 5.0% |
| injury health | 3.0% |
| special teams | 4.0% |
| defensive line | 10.0% |
| offensive line | 10.0% |
| talent recruiting | 10.0% |
| returning production | 7.0% |
Weight total: 100.00% — valid.
From rating to model line
Predicted home margin = points-per-rating-unit × (home GDA Rating − away GDA Rating) + home-field points + home adjustment − away adjustment. Home-field points are excluded at neutral sites. A positive number favours the home team.
- Points per rating unit: 1.0301846185
- Home-field points: 2.5
- Residual cap: 28
- Learning gain: 0.25
- Maximum single move: 4
- Update cadence: weekly
Updates, freezing and honesty rules
- Predictions are frozen at kickoff along with the market line used, so published history can never be quietly rewritten.
- Rating changes are capped, and the winner's gain equals the loser's loss so the league stays balanced.
- Market lines are stored with their source and capture time. Where no line exists, the model edge is reported as unavailable rather than assumed.
- When a required input is missing, GameDay Advantage states exactly what is missing instead of estimating a number.