Who actually finished 2025 hot — and who limped in
The model weights the last six games more heavily than the first eleven. Seattle closed 81 points above its season rating; Kansas City closed 85 points below.
Fifteen percent of every 2026 opening rating comes from a form term that double-weights the closing six games of 2025. The reasoning is simple: a roster in December resembles next September's roster more than one from Week 2 does. The reasoning is also modest — fifteen percent, not fifty, because six games is a small sample and hot streaks regress.
| Team | Season-end rating | Closing form rating | Difference |
|---|---|---|---|
| Seahawks | 1643 | 1724 | +81 |
| Jaguars | 1630 | 1702 | +73 |
| Vikings | 1529 | 1586 | +57 |
| Texans | 1608 | 1661 | +52 |
| Chiefs | 1416 | 1331 | -85 |
| Cardinals | 1371 | 1308 | -63 |
| Raiders | 1374 | 1315 | -59 |
The finishers
Seattle and Jacksonville did not just win late, they won by margins that a 1700-level team produces. That is worth a handful of rating points each in the blend — enough to matter in a league where first place and fifth place are separated by 45 points, and not enough to invent a contender out of nothing.
Minnesota is the more interesting name. A 1586 closing form rating against a 1529 season-long rating is the profile of a team that improved rather than a team that got lucky, and it shows up downstream: the Vikings project to 9.17 wins with a 52% playoff chance despite ranking outside the model's top ten overall.
The stragglers
Kansas City's -85 is the largest negative gap in the league and the reason their overall rating sits at 1432 rather than somewhere in the 1470s. Arizona and Las Vegas closed badly too, but from a much lower base — for those teams the form term is confirming a bad season rather than reframing a mediocre one.
Form is a tiebreaker, not a verdict. It moves a rating by tens of points, never hundreds.
Once real 2026 games start, the form term stops mattering entirely: it is a bridge across the offseason, and results replace it week by week on the ratings history page.