We let AI rank all 32 starting QBs from 2025. Here's what it said.
Muffed.ai · May 2026
At Muffed.ai we built an AI-driven model to rank every starting QB based on what actually happened on the field in 2025 — no narrative, no MVP-buzz adjustments, no “well, but he won the Super Bowl.” Just the data.
The model looked at every dropback by every QB this season — 19,828 of them across 272 regular-season games — and decided how to weigh six measurable things into a single composite score. Here's how it thought about it, and then the full 1-32 ranking with a one-line take on each QB.
How the model decides what's important
The starting premise: a quarterback's job is to convert dropbacks into expected points. Yards, touchdowns, completion percentage, leadership — they're all downstream of that.
So the model anchors on EPA (expected points added per play) and then adds adjustments for the things EPA alone can miss:
Pass and designed-run EPA combined, divided by games played. Mobile QBs get credit for the points they add with their legs. Injured QBs aren't punished twice for missing time. And critically: when a sack happens while the QB was pressured, we count its EPA at half its raw value — that pressure was mostly the O-line's fault, not the QB's.
Efficiency, not just total production. Same 50% pressure-sack adjustment.
This is the NFL's Next Gen Stats version. It controls for the difficulty of every throw — depth, location, defenders, separation — and tells you how much better the QB completed than the model expected. It's the single most accuracy-pure signal we have.
Percentage of dropbacks that gained adequate yardage for the down and distance. Keeps boom-bust throwers from winning purely on a few 80-yard bombs.
Q3/Q4 with the score within 8 points. Small weight because samples are small and noisy.
Not raw sack rate — pressure-to-sack rate. When the QB was already under pressure, how often did he convert it into a sack vs. escape or throw it away? That's a QB skill. Raw sack rate would punish QBs whose O-lines just give up pressure constantly (looking at you, Chargers and Browns) and we wanted this to be a QB ranking, not a passing-offense ranking.
Why z-scores
Each input is on a different scale — EPA per game is in “points,” CPOE is in “percentage points above expectation,” success rate is a percentage. To combine them, we convert each one into a z-score across the 32-QB universe: how many standard deviations above or below the league-average starter that QB is on that metric. Then we weight the z-scores by the percentages above. Final composite is in units of “standard deviations above an average starter.” Positive means above average; +1.0 means a full standard deviation above the average starter; negative means below.
This way a QB who's elite at one thing and bad at another gets a fair shake — we're not double-counting any single signal, and we're not letting unit differences accidentally dominate the math.
The full 2025 ranking
Composite score = standard deviations above an average starter
- Drake MayeNE+2.07
Led the league in CPOE (+9.14, more than four points clear of #2), in QB-only Total EPA per game (+12.88), and in success rate (55.2%). The best QB season of 2025 by a clear margin.
- Jordan LoveGB+1.35
Quiet star of 2025. Second only to Maye in EPA per dropback (+0.299), second-best late-close EPA, 15 games, only 21 sacks. Packers made the playoffs and almost nobody is talking about him.
- Brock PurdySF+1.35
Only 9 starts but elite efficiency: +0.236 EPA/dropback, +5.07 CPOE (2nd in the league), 12.2% pressure-to-sack rate (tied for best with Caleb Williams), best late-close EPA. Small sample, big numbers.
- Matthew StaffordLA+1.07
46 TDs (most in the league), 8 INTs, all 17 games at age 38. The Rams' deep-throw machine in full flight.
- Dak PrescottDAL+0.86
30/10 TD-to-INT on 656 dropbacks, +4.45 CPOE. Top-5 QB season — Dallas finished 7-9-1 on a defense and special teams that couldn't keep up. Dak wasn't the problem.
- Josh AllenBUF+0.76
14 rushing TDs, 599 rushing yards on 94 carries. Rushing alone added +48 EPA. Passing efficiency was good not great this year; the legs carried him into top-6.
- Daniel JonesIND+0.73
Career reset. +3.38 CPOE (6th), 100+ passer rating, 5 rushing TDs. Top-10 QB in only 13 starts.
- Jared GoffDET+0.67
34 TDs, 8 INTs, +0.236 QB-only EPA per dropback. Detroit went 9-8 and missed the playoffs on a tiebreaker.
- Sam DarnoldSEA+0.63
Did not regress. +4.31 CPOE (4th in the league), 25/14 TD-to-INT, played all 17 games. Seattle won the NFC West.
- Patrick MahomesKC+0.60
His CPOE of -2.90 is the lowest of his career. Total production held up (+8.56 EPA/G, 56 carries for 429 yards) because he's still Patrick Mahomes. Missed the last 3 games with an ACL tear.
- Joe BurrowCIN+0.23
Only 8 games before injury. 17/5 TD-to-INT, +2.06 CPOE. Hard to read the small sample; the talent says higher than the production showed.
- C.J. StroudHOU+0.21
14 games, +0.152 QB-only EPA per dropback, more rushing involvement than past seasons. Bounce-back year after a noisy 2024.
- Justin HerbertLAC+0.15
Pressured on 36.3% of dropbacks — second-highest in the league. When pressured, escapes/throws at a league-average rate. The 54 sacks are the LAC O-line, not him. 8th in CPOE, second-most QB rushing yards (510). Top-10 talent stuck behind a sieve.
- Jalen HurtsPHI+0.12
8 rushing TDs, 436 rushing yards on 94 carries, +3.09 CPOE. EPA per dropback and success rate are middling — value is concentrated in red-zone runs and explosive throws.
- Bo NixDEN+0.11
12.3% pressure-to-sack rate, 3rd-best in the league (behind only Purdy and Caleb Williams at ~12.2%). Most dropbacks in the league at 674. CPOE of -2.07 is what's keeping him out of the top 10.
- Trevor LawrenceJAX-0.02
29 passing TDs, 9 rushing TDs, all 17 games. CPOE of -2.73 says the accuracy isn't there yet. Coen's scheme helped him with designed runs.
- Lamar JacksonBAL-0.05
13 games. 29.0% pressure-to-sack rate (4th-highest). -2.23 CPOE. +0.150 QB-only EPA per dropback. He was great late-and-close (+0.204) and still rushed for 359 yards. But 2024 MVP Lamar is not 2025 Lamar.
- Jacoby BrissettARI-0.13
+3.37 CPOE (7th in the league) on 552 dropbacks behind a chaotic Cardinals OL that pressured him 32.8% of the time. He completes throws above expectation; they just don't generate enough value because Arizona's offense around him was stuck. Murray missed half the year.
- Marcus MariotaWAS-0.17
Stepped in for Jayden Daniels and was fine. +3.29 CPOE on 227 attempts. Career backup who didn't sink the season.
- Baker MayfieldTB-0.18
26/11 TD-to-INT, but per-dropback efficiency cooled from 2024. 42 of his 47 carries were scrambles; rush EPA of +41 (3rd among QBs) is doing real work.
- Caleb WilliamsCHI-0.20
Threw for 27/7. CPOE of -6.88 is the worst of any 200+ attempt QB in the league. His pressure-to-sack rate (12.2%, tied for best with Purdy) is hiding the accuracy hole — when he was pressured he escaped almost every time, so the sack count looked fine. The throws are the problem.
- Jaxson DartNYG-0.32
Rookie. 9 rushing TDs on 41 designed runs (3rd-most), +41 rush EPA. The throwing isn't there yet (-2.03 CPOE, 28.2% pressure-to-sack rate, 5th-worst).
- Michael Penix Jr.ATL-0.39
Took over from Cousins mid-season. 9/3 TD-to-INT, -3.20 CPOE. Reasonable rookie growth profile.
- Tyler ShoughNO-0.40
Rookie. 10/6 TD-to-INT in 11 games. Saints are rebuilding.
- Tua TagovailoaMIA-0.53
20/15 TD-to-INT. CPOE of -0.32 — actually below NFL average for the first time as a starter. The YAC machine broke down this year.
- Aaron RodgersPIT-0.53
24/7 TD-to-INT but +0.079 QB-only EPA per dropback. Pressured only 21% of the time — the lowest in the league — and still couldn't generate value. His 5.95 intended air yards is the shortest of any starter.
- Bryce YoungCAR-0.61
Year 3 was supposed to be the leap. +0.041 QB-only EPA per dropback. The leap is still pending.
- Justin FieldsNYJ-0.87
7 TDs and only 1 INT (the only good news) in 9 games. 29.3% pressure-to-sack rate (3rd-worst). Brutal -0.229 late-close EPA per dropback.
- Geno SmithLV-1.14
Steepest year-over-year decline in the league. 19/17 TD-to-INT. -0.032 QB-only EPA per dropback. 31.8% pressure-to-sack rate is the worst in the league. CPOE still positive (+0.77) — accuracy into hopeless situations.
- J.J. McCarthyMIN-1.57
Rookie after a lost 2024. -5.16 CPOE, 11/12 TD-to-INT, 27.6% pressure-to-sack rate. The keys didn't fit.
- Cameron WardTEN-1.64
Hardest rookie year for any #1 overall pick in recent memory. 15/7 TD-to-INT but -2.42 QB-only Total EPA per game across all 17 starts. 55 sacks.
- Shedeur SandersCLE-2.16
Worst per-dropback efficiency of any 200+ dropback QB in 2025 at -0.157. Pressured on 45.5% of dropbacks (the highest rate in the league by 9 percentage points). 3.23s time to throw was the slowest in the league.
A note on pressure and the O-line
One methodology choice deserves calling out, because it changes a few rankings noticeably. We do not punish QBs for raw sack rate. Instead we use pressure-to-sack rate: when the QB was already under pressure, how often did he convert it into a sack vs. escape or throw it away? That's a QB skill. Raw sack rate, by contrast, is dominated by O-line quality and isn't fair to a QB whose line just gives up pressure constantly.
Justin Herbert is the case study: 54 sacks this year on a 36.3% pressure rate (2nd-highest in the league), but when pressured he escaped/threw away at a league-average rate — meaning the sack volume was overwhelmingly an OL problem. We also count the EPA of pressure-related sacks at half their raw value, on the theory that the OL and the QB share the blame when a QB is already under pressure and goes down.
Bo Nix's 12.3% pressure-to-sack rate (3rd-best in the league) is still rewarded; Geno Smith's 31.8% (worst) is still punished. The methodology is meant to be a QB ranking, not an overall passing-game ranking.
What about strength of schedule?
Fair pushback: not every QB threw against the same defenses. So we ran the SOS adjustment too — weighting each opponent defense by how often each QB faced them, then subtracting that “expected” EPA from each QB's actual.
The interesting findings:
- Drake Maye had the 5th-easiest schedule in the league (opponents allowed +0.084 EPA per dropback, vs the +0.043 league average). His SOS-adjusted EPA per dropback drops. He's still #1 by a clear margin.
- Patrick Mahomes had the 3rd-hardest schedule. Adjusting for it moves him from #10 to #6 in the composite — a four-spot jump. His “down year” looks materially better once you account for who he was throwing against.
- Cameron Ward had the single hardest schedule in the league. He still finishes 30th composite even after the adjustment — there wasn't enough on-field production to climb meaningfully.
Top 4 and bottom 5 don't change with SOS adjustment. The biggest middle-of-the-ranking moves besides Mahomes: Mariota +3 (19 → 16), Lamar +2 (17 → 15), Bo Nix -3 (15 → 18), Lawrence -3 (16 → 19).
We kept the headline ranking unadjusted for schedule in this v1.1 — partly because a one-pass SOS calculation is itself biased (a defense's rating is influenced by the offenses it faced) and partly because we want readers to be able to disagree with us about whether and how much SOS should matter. The SOS-adjusted EPA per dropback for every QB is in our verified stats file if you want to pull it and re-rank.
A note on how we work
Every number above was looked up in a versioned stats file before it was written — there are no estimates and no “from memory” claims. The full 32-QB metric bundle, composite scores, z-scores, SOS adjustments, and pressure metrics live in a JSON file we regenerate from raw nflverse play-by-play and NFL participation data. If a number here turns out to be wrong, it's a bug in our stats library and we'd rather know — that's the whole point of building it this way.
— Muffed.ai
Listen
We do this every week. For every team. For every top fantasy player.
Muffed is the AI-narrated NFL podcast network — same-depth coverage of every team, every week. Pre-season, we've seeded every team page with a 2025 season-in-review episode. Once the 2026 season kicks off, weekly recaps drop every Tuesday.