Accountability

Every prediction, graded

Every 2026 game the model would have called, using only games played before it. Nothing here is stored — the whole season is recomputed on every request, so it cannot be edited after the fact and a bad week cannot quietly go missing.

228

Games graded

70.6%

Winner called

13.5 pts

Average miss

Tennessee's games

4 of 4 called correctly. Newest first.

Wk 6 Tennessee Volunteers at Arkansas Razorbacks
called it
Said Tennessee Volunteers by 12.4 (79%) · actual Tennessee Volunteers by 7
Wk 5 Auburn Tigers at Tennessee Volunteers
called it
Said Tennessee Volunteers by 5.1 (63%) · actual Tennessee Volunteers by 10
Wk 4 Texas Longhorns at Tennessee Volunteers
called it
Said Texas Longhorns by 3.4 (59%) · actual Texas Longhorns by 3
Wk 3 Kennesaw State Owls at Tennessee Volunteers
called it
Said Tennessee Volunteers by 10.5 (75%) · actual Tennessee Volunteers by 33

Week by week

Wk 343/57 · 75% · 12.3 pts
Wk 443/58 · 74% · 14.8 pts
Wk 541/56 · 73% · 14.0 pts
Wk 634/57 · 60% · 12.9 pts

Where it was most wrong

The ten games whose final margin was furthest from the prediction.

Wk 5 Middle Tennessee Blue Raiders at Kansas Jayhawks
called it
Said Kansas Jayhawks by 4.3 (61%) · actual Kansas Jayhawks by 55
Wk 4 UCLA Bruins at Maryland Terrapins
called it
Said UCLA Bruins by 1.0 (53%) · actual UCLA Bruins by 51
Wk 4 Tulsa Golden Hurricane at Arkansas Razorbacks
missed
Said Tulsa Golden Hurricane by 15.1 (84%) · actual Arkansas Razorbacks by 28
Wk 5 Stanford Cardinal at Wake Forest Demon Deacons
called it
Said Wake Forest Demon Deacons by 12.5 (79%) · actual Wake Forest Demon Deacons by 54
Wk 6 San Diego State Aztecs at Oregon State Beavers
called it
Said Oregon State Beavers by 11.0 (76%) · actual Oregon State Beavers by 52
Wk 3 Eastern Michigan Eagles at Wisconsin Badgers
called it
Said Wisconsin Badgers by 3.6 (59%) · actual Wisconsin Badgers by 44
Wk 5 Eastern Michigan Eagles at Massachusetts Minutemen
missed
Said Massachusetts Minutemen by 15.6 (85%) · actual Eastern Michigan Eagles by 24
Wk 5 Memphis Tigers at Charlotte 49ers
called it
Said Memphis Tigers by 12.2 (79%) · actual Memphis Tigers by 51
Wk 3 Colorado Buffaloes at Northwestern Wildcats
missed
Said Colorado Buffaloes by 2.8 (57%) · actual Northwestern Wildcats by 34
Wk 6 Buffalo Bulls at Toledo Rockets
missed
Said Toledo Rockets by 21.5 (92%) · actual Buffalo Bulls by 15

Road teams it took, and got right

Picking against home field is the call the model actually had to make.

Wk 4 UCLA Bruins at Maryland Terrapins
called it
Said UCLA Bruins by 1.0 (53%) · actual UCLA Bruins by 51
Wk 5 Memphis Tigers at Charlotte 49ers
called it
Said Memphis Tigers by 12.2 (79%) · actual Memphis Tigers by 51
Wk 4 Notre Dame Fighting Irish at Purdue Boilermakers
called it
Said Notre Dame Fighting Irish by 17.8 (88%) · actual Notre Dame Fighting Irish by 39
Wk 5 Baylor Bears at Arizona State Sun Devils
called it
Said Baylor Bears by 3.4 (59%) · actual Baylor Bears by 36
Wk 6 Ole Miss Rebels at Vanderbilt Commodores
called it
Said Ole Miss Rebels by 4.6 (62%) · actual Ole Miss Rebels by 35
Wk 5 Alabama Crimson Tide at Mississippi State Bulldogs
called it
Said Alabama Crimson Tide by 3.6 (59%) · actual Alabama Crimson Tide by 33
Wk 5 Indiana Hoosiers at Rutgers Scarlet Knights
called it
Said Indiana Hoosiers by 12.6 (80%) · actual Indiana Hoosiers by 32
Wk 5 Oregon State Beavers at Colorado State Rams
called it
Said Oregon State Beavers by 0.7 (52%) · actual Oregon State Beavers by 30
Wk 4 Massachusetts Minutemen at Sacramento State Hornets
called it
Said Massachusetts Minutemen by 6.8 (67%) · actual Massachusetts Minutemen by 29
Wk 3 Georgia Bulldogs at Arkansas Razorbacks
called it
Said Georgia Bulldogs by 13.3 (81%) · actual Georgia Bulldogs by 28

How it is graded

For every week, the ratings are recomputed from the games played before it, and the prediction that follows is exactly the one the model would have made at the time. No game contributes to its own prediction.

Early weeks are not graded at all — until roughly a hundred and fifty games have been played, the ratings are too thin to be worth marking. Games involving a team outside the ratings pool are excluded, because there was no rating to predict with. Ties grade nobody.

Probabilities use a spread fitted on these same out-of-sample predictions, which is the same conversion the rest of the site uses. See the accuracy page for whether that confidence is honest.