Through Week 2, including Monday Night Football — updated Sept. 22, 2026. Every team has two games in the book.

Rating is how many points per game a team has outplayed its opponents by, counting only the snaps that mattered — anything after a game was already decided is thrown out, so a late touchdown in a blowout helps nobody. Zero is league average. San Francisco at +16.0 has been about sixteen points a game better than the teams it has played; Indianapolis at −20.0 has been twenty points worse.

Off and Def split that same performance by side of the ball, shown per play instead of per game, which is why those numbers look small — a tenth of a point a snap is an enormous edge across sixty snaps. On defense, negative is good: those are points the defense kept off the board. Opp is how tough a team's slate has been so far, shown as context only — it is not baked into the rating. SOS ranks the schedule still to come from 1 to 32, hardest first, with the difficulty figure beside it.

Exactly how the rating is built is ours. What it is not: there is no poll in it, no preseason expectation, and no special teams number, because the underlying data does not carry one. Two games is also not much — read the top and the bottom of this board with more confidence than the middle, where a single quarter still moves a team ten places.

These rankings reflect what the numbers say about how each QB has actually played in the games measured so far this season — not career value, draft stock, or who's the better long-term quarterback. Small sample size, will shift as the season goes on.

Through Week 2, including Monday Night Football — corrected September 22, 2026. Jaxson Dart has been removed from this board: he is out for the season, so there is no week-to-week ranking left to give him. Week 3 availability reflects injury and depth-chart news as of that morning.

The grade is built from this season's play-by-play. Every dropback outside garbage time is measured three ways: whether the play actually helped the team score, how accurate the throw was for its difficulty, and whether it moved the chains rather than just occasionally going deep. Efficiency carries most of the weight, accuracy and consistency the rest. Garbage time is thrown out, so checkdowns in a blowout pad nobody. Quarterbacks with few attempts are pulled toward league average in proportion to how little they have thrown, which is why one good half does not make a number one.

A single game of five or fewer dropbacks that ends with the quarterback leaving hurt is dropped entirely, because that sample is too small to mean anything and often includes the play the injury happened on. That is why Kyler Murray does not appear at all — his only appearance this season was five plays ended by a concussion. He has cleared protocol and starts in Week 3; there is simply nothing to grade him on yet.

Zero is league average. Tiers are buckets of the grade, nothing more: S is 35 and up, A is 15 to 34, B is 0 to 14, C is −1 to −15, and D is below −15. The grade is not a real unit like yards or points, so the gap between two quarterbacks says more than either number on its own.

Turnover luck check, through Week 2. This does not change any ranking above — a turnover's real cost is already inside every grade. This only separates the part that was a decision from the part that was a bounce.

Every interception thrown in Week 1 was charted as a genuine bad decision, not a fluke: 17 interceptions, 17 different quarterbacks, none flagged as tipped or freak bounces. Fumble recovery, by contrast, is close to a coin flip league-wide — offenses recovered their own fumble 54 percent of the time this season, 25 of 46.

• Justin Herbert fumbled three times and recovered all three himself. Lucky. His ball security is shakier than his numbers show.
• Baker Mayfield fumbled four times and lost three. Unlucky, on top of a rough stretch.
• Jordan Love fumbled four times and lost one. Some of his season number is quietly propped up by recovery luck.

Week 3, 2026 season (games through Sept 19, 2026) — corrected composite build.

One number from three ingredients: an opponent-adjusted efficiency rating, raw per-play production, and the betting market. The adjusted rating does nearly all of the work, because it is the only one of the three that accounts for who a team has actually played. The other two are there to catch what it misses, and an earlier version of this board leaned on them far too hard — a single blowout of a bad opponent, or an unremarkable preseason win total, was enough to carry a team thirty spots higher than its play deserved. That is fixed. How the three are weighted is ours, and it is a judgment call rather than a standard.

The number is roughly how many points per game a team projects to beat an average FBS opponent by, so +29.2 is a four-touchdown favorite against average and −27.3 is the same margin the other way. Off and Def are national ranks out of 138 in the on-field offense and defense scores, where 1 is best and the market plays no part, since odds do not split by side of the ball. No AP or Coaches poll touches any of it.

Through Week 3, 2026 season. New this week: the grade now adjusts for how good the defense actually was. Beat a bad defense and the game counts for a bit less; beat a good one and it counts for more. That is a real change to the number, not a relabelling — several quarterbacks moved ten places or more, and the three who fell furthest had all just played weak defenses.

The grade blends two things: how many points a quarterback's plays were worth per snap, and how often those plays gained enough to keep the drive on schedule. Efficiency carries the larger share, staying on schedule the rest. Both are measured against the national average, so zero is average and higher is better. Early-season numbers are pulled toward that average in proportion to how little a quarterback has thrown, which is why a hot fifty snaps does not automatically win the board.

The defense listed beside each last game — Tough, Average or Weak — is the same strength measure the adjustment uses, shown so you can see what a number is built on. Two quarterbacks faced opponents with no rating, shown as unrated.

Completion percentage appears in the full table for color only. It is raw accuracy with no adjustment for how hard the throws were, and it does not feed the grade at all. Garbage time is stripped out. Passer identity comes from the official play-by-play text checked against the roster, which covers about nine in ten pass plays nationally; the rest were left out rather than guessed at. Tiers are recalculated from scratch every week, and three games is still a small sample.