Offense or defense? Splitting every contender's rating in two
The Rams have the league's best offensive rating at 1686 and the Seahawks the best defensive rating at 1646. Only one team is top five in both.
A single team rating is a summary, and summaries hide things. Two clubs can sit within five points of each other and be built in completely opposite ways. So alongside the overall number, the model keeps a separate offensive and defensive rating for every team, derived from points scored and allowed against the strength of the opponents faced.
| Team | Overall | Offense | Defense | Shape |
|---|---|---|---|---|
| Seahawks | 1619 | 1635 | 1646 | Balanced elite |
| Rams | 1555 | 1686 | 1567 | Offense-led |
| Texans | 1588 | 1519 | 1642 | Defense-led |
| Patriots | 1587 | 1645 | 1605 | Balanced |
| Broncos | 1574 | 1514 | 1618 | Defense-led |
| Bills | 1564 | 1632 | 1539 | Offense-led |
Seattle is the only genuinely two-sided team at the top
The Seahawks are third in offensive rating and first in defensive rating — the only club in the top five of both lists. That balance is a large part of why they hold the No. 1 overall rating despite the Rams outscoring them by 35 points in 2025. Los Angeles put up 518 points, the most in the league, but gave up 346; Seattle scored 483 and allowed 292.
Why the splits matter for matchups, not just rankings
- Two defense-led teams tend to produce lower totals — Houston hosting Buffalo carries a 44.5 total while the offense-heavy Bengals game sits at 51.5.
- An offense-led team is usually more volatile week to week, which widens the range of outcomes even when the mean projection is unchanged.
- The overall rating, not the split, is what drives the win probability. The splits are context for reading the number, not a second model bolted on top.
The strangest split in the league
Kansas City went 6-11 and yet carries a defensive rating of 1593, which would be top five. Their offensive rating of 1457 is what dragged the season down: 362 points scored against 328 allowed is a point differential that belongs to a .500 team, not a 6-11 one. That mismatch is the single clearest example of why the model keeps both halves visible instead of only reporting the summary.
Every team's split is on its own page next to the rating trail, so the shape of a roster is visible before you click into a matchup.