Methodology

GRID stands for Gridiron Résumé Index. It follows the design of the NCAA's basketball NET, a résumé metric that combines results with efficiency and sorts games into quadrants, adapted for football. It isn't affiliated with the NCAA.

The basketball NET

The NCAA Evaluation Tool is built from two parts. The Team Value Index looks only at results: who you played, where, and whether you won. Adjusted net efficiency measures points scored minus points allowed per 100 possessions, adjusted for opponent and location. Games are then sorted into quadrants by the opponent's NET rank and where the game was played. GRID keeps that structure.

Results Index → wins and losses only

This is GRID's version of the Team Value Index. The NCAA doesn't publish its formula, so here it's a Bradley–Terry model: each team gets a rating so that the chance of beating an opponent depends on the gap between their ratings, plus home field (0.3 logits, about a 57% win rate for the home team when two teams are even). Margin doesn't matter at all. Winning on the road against a strong team raises a rating the most. Losing at home to a weak team lowers it the most.

Adjusted efficiency → points per drive

Basketball's unit is the possession; football's is the drive. Using each game's drive chart, a weighted ridge regression estimates every team's offensive and defensive points per drive against an average opponent at a neutral site. Home field is worth 0.2 points per drive (about 2.5 points a game). Ratings are pulled toward average by the equivalent of 60 average drives, so a team's efficiency counts for less until it has played about five games. Adj. PPD is offense minus defense.

Blowouts get diminishing credit: a winning margin up to 21 points (three scores) counts in full, and each point beyond that counts 50%. A 56-point win counts as a 38½-point margin, with the reduction taken half off the winner's score and half added to the loser's. Displayed scores aren't changed. The original 2018 NET capped margin at 10 points, then dropped the cap in 2020. A hard cap throws away real information: on 2022–2025 games, every hard cap tested made rankings worse at picking the next week's winners. Half credit past three scores cost almost nothing.

Combining them

Both components are converted to z-scores across FBS teams and blended 50% results and 50%efficiency. The NCAA doesn't disclose its weights, so this is a judgment call.

Early-season carryover

This part is not in the NCAA formula, which uses only current-season games. The NCAA works around early noise by not publishing the NET until teams have played 6–8 games, and until then sites like bballnet keep showing last season's final NET. Football seasons are too short to wait that long, so early on each team's GRID score is blended with its final GRID score from last season.

Last season's share of the score shrinks each week:

On 2022–2025 data, last season mainly helps predict results in the first four weeks. From week 5 on, rankings with and without it pick later games equally well. So it fades quickly: enough to steady September, while this season's results, like a bad early loss, take over fast. The results and efficiency columns always show current-season values only.

FCS opponents

FCS teams are Division I, so they're included, just as every D-I team counts in basketball. They're rated from all their games, including FCS vs. FCS, but never listed in the rankings. For quadrants, each FCS team gets an equivalent rank, meaning where its GRID score would place it among FBS teams, and goes through the same cutoffs as any other opponent. Most FCS teams land in Q4, but an elite one can reach Q3 on the road. Games against non-D-I teams are ignored.

Quadrants

Basketball's cutoffs assume about 365 D-I teams. Football has about 136 FBS teams, so the cutoffs are scaled by the same proportions and rounded.

Other columns

• +/−: change from the GRID rank computed through the previous week.
• SOS: the average GRID score of the opponents a team has played.