QB Composite Rating Leaders
QB Composite Rating — 2023 NFL leaders, updated Sep 3, 2026. Brock Purdy (SF) led the league in QB composite rating at +1.920, ahead of 38 other qualifying players. Computed from nflverse play-by-play, NFL Next Gen Stats via nflverse and FTN Data via nflverse charting, regular season only. Through the 2023 season.
Six numbers, one score. Expected points added per game and per dropback carry most of the weight, because moving an offense is the job; completion percentage over expected is the accuracy check that does not care how far the throws travelled; success rate asks how often the play simply worked; late-and-close play and QB-fault sack rate get five percent each, small on purpose because both are small samples. Every input is measured against the same population of quarterbacks, so a score is a distance from an average qualified starter rather than a grade out of a hundred — plus one means a standard deviation better than that man, and the middle of this board is genuinely crowded. It is a way of holding six things at once, not a verdict, and the raw columns are all here to be read against each other.
| # | Player | DB | Score | EPA/DB | CPOE | Succ% | QB-fault Sk% | Listen |
|---|---|---|---|---|---|---|---|---|
| 1 | Brock Purdy | 486 | +1.920 | 0.36 | +2.5 | 54.7% | 2.9% | |
| 2 | Dak Prescott | 658 | +1.561 | 0.25 | +4.7 | 50.6% | 3.8% | |
| 3 | Josh Allen | 652 | +1.259 | 0.20 | +1.9 | 50.8% | 1.7% | |
| 4 | C.J. Stroud | 557 | +0.965 | 0.22 | +0.9 | 46.7% | 1.4% | |
| 5 | Jared Goff | 643 | +0.881 | 0.17 | +1.1 | 49.1% | 1.4% | |
| 6 | Tua Tagovailoa | 605 | +0.838 | 0.21 | +1.5 | 49.8% | 2.1% | |
| 7 | Jalen Hurts | 623 | +0.784 | 0.14 | +4.1 | 46.4% | 2.6% | |
| 8 | Jordan Love | 633 | +0.662 | 0.18 | -1.4 | 46.3% | 1.7% | |
| 9 | Jake Browning | 284 | +0.662 | 0.18 | +3.3 | 47.9% | 4.2% | |
| 10 | Lamar Jackson | 560 | +0.626 | 0.16 | +1.7 | 48.9% | 3.9% | |
| 11 | Kirk Cousins | 329 | +0.559 | 0.13 | +4.6 | 45.6% | 2.4% | |
| 12 | Derek Carr | 590 | +0.546 | 0.12 | +4.0 | 45.8% | 1.5% | |
| 13 | Geno Smith | 548 | +0.533 | 0.14 | +2.2 | 48.4% | 2.2% | |
| 14 | Baker Mayfield | 639 | +0.433 | 0.14 | +0.2 | 44.4% | 2.2% | |
| 15 | Patrick Mahomes | 672 | +0.428 | 0.14 | -1.7 | 48.4% | 1.3% | |
| 16 | Justin Herbert | 517 | +0.330 | 0.12 | +0.5 | 46.0% | 1.7% | |
| 17 | Matthew Stafford | 565 | +0.279 | 0.14 | -1.3 | 45.8% | 2.1% | |
| 18 | Joe Burrow | 402 | +0.176 | 0.09 | +1.3 | 47.3% | 2.0% | |
| 19 | Russell Wilson | 527 | +0.118 | 0.09 | +1.3 | 43.3% | 1.9% | |
| 20 | Justin Fields | 459 | +0.084 | 0.12 | +0.4 | 39.9% | 1.3% | |
| 21 | Ryan Tannehill | 268 | +0.083 | 0.11 | +0.4 | 40.3% | 0.0% | |
| 22 | Tyrod Taylor | 218 | +0.045 | 0.11 | +1.9 | 43.6% | 0.9% | |
| 23 | Trevor Lawrence | 643 | -0.010 | 0.09 | -0.5 | 46.7% | 1.2% | |
| 24 | Gardner Minshew | 548 | -0.089 | 0.10 | -1.6 | 42.3% | 1.3% | |
| 25 | Kyler Murray | 305 | -0.283 | 0.06 | -2.9 | 42.3% | 0.7% | |
| 26 | Will Levis | 291 | -0.328 | 0.11 | -3.2 | 40.9% | 1.7% | |
| 27 | Joe Flacco | 213 | -0.367 | 0.01 | -1.0 | 42.7% | 1.4% | |
| 28 | Deshaun Watson | 204 | -0.638 | 0.00 | -2.8 | 40.7% | 3.4% | |
| 29 | Aidan O'Connell | 367 | -0.639 | 0.03 | -2.3 | 41.7% | 2.2% | |
| 30 | Desmond Ridder | 439 | -0.700 | 0.01 | -1.4 | 46.7% | 3.0% | |
| 31 | Joshua Dobbs | 484 | -0.852 | -0.01 | -0.4 | 40.7% | 1.6% | |
| 32 | Kenny Pickett | 361 | -0.859 | -0.06 | +0.3 | 40.7% | 2.8% | |
| 33 | Sam Howell | 711 | -0.929 | -0.04 | +0.2 | 41.5% | 3.5% | |
| 34 | Tommy DeVito | 237 | -1.003 | -0.03 | -3.6 | 33.3% | 2.5% | |
| 35 | Daniel Jones | 211 | -1.093 | -0.10 | +0.5 | 39.3% | 4.7% | |
| 36 | Mac Jones | 380 | -1.168 | -0.10 | -0.6 | 42.6% | 1.1% | |
| 37 | Bailey Zappe | 246 | -1.567 | -0.17 | -2.2 | 33.7% | 2.0% | |
| 38 | Bryce Young | 620 | -1.595 | -0.10 | -1.6 | 37.1% | 2.9% | |
| 39 | Zach Wilson | 438 | -1.656 | -0.12 | -1.5 | 37.4% | 3.9% |
Other seasons
How this board is built
The six inputs and what they are worth
Expected points added per game, counting designed quarterback runs as well as dropbacks, is 45 percent. Expected points added per dropback is 20. Completion percentage over expected, taken from Next Gen Stats rather than recomputed, is 15. Success rate — the share of dropbacks that gained enough to stay on schedule — is 10. Expected points per dropback in the second half within one score is 5, and QB-fault sack rate is 5 and enters negative, because taking the sack yourself is the failure. Each input is turned into a z-score across the qualifying population before the weights are applied, so the total is in standard deviations rather than in any of the six original units.
Why a small sample scores closer to average
Before the z-scores are taken, every input is pulled toward the league mean by its own sample size: a quarterback's figure and the league's are averaged, weighted by how many snaps he has and by a constant that says how much evidence it takes to trust a rate. That constant is 150 dropbacks for the per-dropback, success, accuracy and sack-rate inputs, 60 dropbacks for the late-and-close one, and 4 games for the per-game one. A passer with 600 dropbacks barely moves; one who has just cleared the floor moves most of the way. Against a version with no such pull it changes a ranking by up to three places, and it is why a quarterback with no late-and-close snaps at all sits at the league average there rather than at zero.
Sacks, and who is charged for them
FTN's charters judge every sack for whether the quarterback caused it — holding the ball, missing the pressure, running into it — or whether the protection simply broke. On a sack charged to the protection, the expected points lost are scored at zero for the quarterback in both expected-points inputs; the snap still counts as a dropback, because he still used it. A QB-fault sack keeps its whole cost. A sack with no charting row keeps its raw cost, the conservative direction, and the count of those appears on the page as part of the qualifying line.
What this does not recompute
The 2025 ranking published in Muffed's quarterback article uses version 1.1 of this model, whose pressure inputs come from participation charting that nflverse delivers only after a season ends. That ranking stands as published and is not recomputed here, which is why 2025 is absent from the seasons above. This version reads only feeds that move during the season, so it can be wrong about a quarterback in Week 4 and right about him in Week 12 — that is the point of it.
Choosing between two of these players?
Both men's 2025 production side by side, and which of them rests on inputs that historically repeat.
More leaderboards
- Rushing Yards Over Expected
- Target Share
- Running Back Receiving Efficiency
- Completion Percentage Over Expected
- Adjusted Net Yards per Attempt
- Quarterback Pressure Rate
- Defensive Blitz Rate
- Quarterbacks vs Man and Zone Coverage
- Pass Rate Over Expected
- Clean-Core Rating
- Usage vs Production
- Team Expected Points Added
- Quarterback EPA per Dropback
- Air Yards Share
- Stacked Box Rate
- Regression Flags
- Weighted Opportunity Rating
- Snap Share
- Pass-Snap Share
- Receiving Yards per Pass Snap
- Availability
- Tight End Receiving Efficiency
- Tight End Average Depth of Target
- Stat Stickiness
- Play-Action Rate
- Pre-Snap Motion Rate
- Drop Rate
- Contested Catch Rate
- Interception-Worthy Throw Rate
- Quarterback-Fault Sack Rate
- Quarterback Intended Air Yards
- Team Air Yards and Passing Volume
- Receiver Separation and Cushion
- Wide Receiver Receiving Efficiency
- Team Success Rate and Explosive Play Rate
- Goal-Line Usage
- Quarterback Deep Ball
- Team Pace
Follow the football you care about. Players, teams, stat boards — press Follow on any of them, and that is what your weekly email is made of. Nothing you didn't ask for.
Drafting this month? Follow the guys you end up with, and your first weekly email lands Week 1 — the real plays and the real numbers behind everyone on your list.
Follow what you care about — free →Use Muffed in your AI
Ask Claude or ChatGPT football questions and get these verified figures back, with sources. Add https://muffed.ai/api/mcp — free, no account, no API key.