The models
-
Consensus (our call)
The average of Our model, Elo, SRS and ESPN FPI (whichever are there). This is our call.
Worked out from the four models whenever one of them changes, frozen at kickoff. Graded from Week 1.
This season: winner right 59.5 of 95 (63%), average miss 9.2, over 95 graded games.
-
Our model Experimental In our call
Our ratings from nflverse play-by-play: opponent-adjusted EPA per play, home field and the starting quarterback, frozen on the 2024 season and tested once on 2025. Experimental.
Recalculated by our NFL job after each week's games and before each kickoff. Graded from Week 1.
This season: winner right 58.5 of 95 (62%), average miss 9.1, over 95 graded games.
-
Elo In our call
Every team's Elo rating from every game since 1999 (FiveThirtyEight's method).
Updated by our NFL job after every game. Graded from Week 1.
This season: winner right 56.5 of 93 (61%), average miss 9.3, over 95 graded games.
-
SRS In our call
Simple Rating System: points scored minus allowed, adjusted for the schedule, blended with last season early on. Settings chosen by hand, not tuned.
Recalculated by our NFL job after each week's games. Graded from Week 1.
This season: winner right 58.5 of 93 (63%), average miss 9.4, over 95 graded games.
-
ESPN FPI In our call
ESPN's Football Power Index matchup predictor, read from ESPN before kickoff.
Read from ESPN in the week of the game, up to kickoff. Graded from Week 1.
This season: winner right 57.5 of 95 (61%), average miss 9.5, over 95 graded games.
Source: ESPN FPI
-
Record
The team with the better record this season; equal records go to the home team. A winner only.
Updated after every game. Graded from Week 1.
This season: winner right 57.5 of 95 (61%), over 95 graded games.
-
Market
The market's margin from nflverse's schedules; its win chance is that margin on the same curve as SRS.
From nflverse's schedules, read by our NFL job before kickoff. Graded from Week 1.
This season: winner right 60.5 of 95 (64%), average miss 9.2, over 95 graded games.
Source: nflverse
-
DRatings Outside model
DRatings' NFL predictor: a computer rating's win chance and projected score, read from dratings.com before kickoff.
Read about every 6 hours, only before kickoff. Graded from Week 1.
This season: winner right 59.5 of 95 (63%), average miss 1.9, over 95 graded games.
Source: DRatings
-
Sagarin Outside model
Jeff Sagarin's NFL PREDICTOR ratings, read from sagarin.com before kickoff: the two teams' ratings and his home edge.
Read about every 6 hours, only before kickoff. Graded from Week 7.
This season: winner right 1 of 1 (100%), average miss 2.6, over 1 graded games.
Source: Sagarin
-
The Margin Outside model
The Margin's NFL model (themarginsports.com): its projected margin and win chance, from its open weekly files frozen before kickoff.
Each week's file, frozen before kickoff, read about every 6 hours. Graded from Week 1.
This season: winner right 58.5 of 94 (62%), average miss 9.5, over 95 graded games.
Source: The Margin
-
Machine Learning Ball Outside model
Machine Learning Ball's NFL simulation (machinelearningball.com): its median margin and win chance, from its published pre-game forecast files.
Read about every 6 hours, only before kickoff. Graded from Week 1.
This season: winner right 54.5 of 95 (57%), average miss 5.5, over 95 graded games.
Source: Machine Learning Ball
-
Sonny Moore Outside model
Sonny Moore's computer power ratings, read from sonnymoorepowerratings.com before kickoff: the two ratings and his 2.3-point home edge.
Read about every 6 hours, only before kickoff. Graded from Week 7.
This season: winner right 1 of 1 (100%), average miss 1.9, over 1 graded games.
Source: Sonny Moore
Each record is from our prediction scorecard: every model graded on every game it predicted before kickoff. This week's calls: NFL predictions.