Track record, 2026
How YardLine's models are doing this season, graded in public, misses included. Every number below is worked out from the predictions the site stores, against what actually happened, and adds games as they finish. For how the models are built, and how they did on the seasons held out of training, see the methodology.
Win probability
The win-probability curve on every game page, checked play by play: 79,942 plays from 455 games. Games are added as they finish: the ones the site tracks live within the hour after the final whistle, untracked ones after Tuesday's scoring run. Latest game graded: Oct 15. The model was fit on 2014-2022, so none of this season was in its training data.
Stated against what happened
Every play's home win probability, sorted into ten bins with about the same number of plays in each (tied values stay together), against how often the home team went on to win. On the dashed diagonal means a stated 70% won 70% of the time.
The ten bins
| Stated | Home won | Plays | Games |
|---|---|---|---|
| 7.9% | 4.8% | 7,995 | 145 |
| 30.1% | 22.0% | 7,994 | 169 |
| 47.6% | 44.3% | 7,994 | 189 |
| 63.1% | 62.3% | 7,994 | 194 |
| 75.4% | 83.3% | 7,994 | 208 |
| 86.5% | 94.9% | 7,994 | 222 |
| 93.7% | 99.2% | 7,994 | 265 |
| 97.6% | 99.6% | 7,994 | 264 |
| 99.4% | 100.0% | 7,994 | 237 |
| 99.9% | 100.0% | 7,995 | 184 |
Brier score by week
The average squared gap between the stated chance and the result, over every play: lower is better, and a coin flip on every play scores 0.250. The model scored 0.121 on the 2024 season it never trained on. A week of lopsided games scores lower than a week of close ones, so weeks aren't a like-for-like comparison.
| Week | Games | Plays | Brier |
|---|---|---|---|
| Week 1 | 99 | 17,196 | 0.054 |
| Week 2 | 86 | 14,989 | 0.077 |
| Week 3 | 75 | 13,217 | 0.079 |
| Week 4 | 71 | 12,551 | 0.094 |
| Week 5 | 59 | 10,438 | 0.108 |
| Week 6 | 58 | 10,288 | 0.158 |
| Week 7 | 7 | 1,263 | 0.189 |
Pregame projections
The projected score and win chance on each game page before kickoff, from a model that uses no betting line. The betting market is shown beside it for reference: it's the best public forecast there is, not a target.
7 of 336 predictions were stored before kickoff. The other 329 were scored after the games, when this version of the model was released: it was trained on earlier seasons only and sees only what was known before each game, but they aren't the numbers the site showed at the time. The table by week says which is which.
| YardLine | For reference, the market | |
|---|---|---|
| RecordStraight-up record | 258-78 (77%) | 264-72 (79%) |
| Avg missAverage miss on the margin | 12.5 pts | 11.4 pts |
| BrierBrier score320 games (320 games with moneylines) | 0.169 | 0.156 |
Straight-up record picks whoever the model gives better than even odds, or the market's favourite. The pick follows the win chance rather than the projected score, which in a near-even game can lean the other way. Average miss is the gap between the projected margin (or the spread) and the final margin, overtime included, as the scoreboard shows it. Brier score is the average squared gap between the win chance and the result, lower is better; the market's comes from its moneylines with the bookmaker's margin taken out.
By week
| Week | Games | Record | MissAvg miss | Prediction |
|---|---|---|---|---|
| Week 1 | 51 | 43-8 | 15.1 | After the factScored after the fact |
| Week 2 | 49 | 35-14 | 12.0 | After the factScored after the fact |
| Week 3 | 57 | 49-8 | 9.7 | After the factScored after the fact |
| Week 4 | 58 | 44-14 | 12.6 | After the factScored after the fact |
| Week 5 | 56 | 41-15 | 14.4 | After the factScored after the fact |
| Week 6 | 58 | 40-18 | 12.0 | After the factScored after the fact |
| Week 7 | 7 | 6-1 | 9.3 | Before kickoffStored before kickoff |
Most confident wrong calls
The games where the model gave the team that lost its best chance to win, before kickoff.
| Game | Loser's chance | Final |
|---|---|---|
| Kentucky at Texas A&M Week 3 Market TA&M -16.5 | TA&MTexas A&M 96% | UKKentucky 31-21 |
| Massachusetts at Rutgers Week 1 Market RUTG -29.5 | RUTGRutgers 96% | MASSMassachusetts 37-21 |
| Buffalo at Toledo Week 6 Market TOL -19 | TOLToledo 94% | BUFBuffalo 23-8 |
| Wisconsin at Penn State Week 4 Market PSU -10 | PSUPenn State 89% | WISWisconsin 24-20 |
| Oregon at Oklahoma State Week 2 Market ORE -24.5 | OREOregon 89% | OKSTOklahoma State 39-31 |