Drake Maye walked off the field in Foxborough on a Sunday afternoon in early December with 312 passing yards and three touchdowns, his sixteenth straight pass completion of the second half, and the Patriots at 10-2 with an offensive rating that had been climbing every week since October. By 6 PM Eastern that night, three different ESPN segments had named him the MVP front-runner. By 9 PM, the betting markets had moved his MVP odds from +900 to +275. By Monday morning, NFL Twitter had a new consensus front-runner. Maye had not had the best season-long EPA per play. He had not led the league in passing yards or touchdowns. He was twenty-four years old, in his second NFL season, on a team that had been picked for 6 wins. The narrative had decided, between drives, that the MVP race was now his.
This is how NFL MVP races actually get decided in the modern era. Not by the cumulative season-long data, which the analytical community would prefer. Not by the head-to-head comparison of the candidates, which most fan bases find exhausting. The MVP race gets decided by which candidate produces the most-citable moment in a high-leverage game in late November or early December, and then the narrative that builds around that moment carries through the final four weeks of voting. Lamar Jackson missed three weeks with a hamstring strain at the wrong time. Patrick Mahomes had been efficient but the Chiefs had not produced the highlight games. Josh Allen had a clean season but unspectacular. Drake Maye produced the highlight at exactly the moment the voting room needed someone new to fall in love with.
What follows is what the actual MVP-relevant numbers looked like through Week 14, where the narrative-formation cycle has historically produced the wrong winners by analytical standards, and how to read the next four weeks of the MVP race without getting swept into either the Drake Maye euphoria or the Josh Allen backlash.
The actual MVP-relevant numbers through Week 14
The most-rigorous public quarterback evaluation framework — combining EPA per play, CPOE, pressure-adjusted completion, and clutch efficiency — produces the following ranking for the 2025 season through Week 14:
| Quarterback | EPA/play | CPOE | Adjusted EPA | Public MVP odds |
|---|---|---|---|---|
| Josh Allen | +0.24 | +5.2 | +0.31 | +275 (-50%) |
| Drake Maye | +0.21 | +4.8 | +0.27 | +275 (favorite) |
| Patrick Mahomes | +0.19 | +4.1 | +0.24 | +550 |
| Matthew Stafford | +0.22 | +5.0 | +0.28 | +650 |
| Jared Goff | +0.18 | +3.9 | +0.23 | +1400 |
The pattern matters. Allen leads the league in adjusted EPA. Stafford is second. Maye is third. The MVP odds have Maye and Allen tied as favorites because the narrative has been faster to update than the cumulative data has. The structural reason this happens is that the voting room is not reading the rigorous metrics; the voting room is reading the same coverage that NFL Twitter is reading, and the coverage is reading the most-recent highlights.
Where the narrative formation actually breaks the analytical case
The MVP voting room — roughly 50 sportswriters chosen by the AP — is structurally biased toward late-season narrative. The reason is that the ballots are due in early January, which means voters spend December watching the candidates with the freshest possible memory of recent games. A player who plays poorly in December but had a strong September is structurally penalized; a player who plays well in December but had an ordinary September is structurally rewarded. The Drake Maye case is the cleanest possible example of this dynamic.
Allen’s case got hurt by a quiet stretch in late November when the Bills were dealing with offensive line injuries that compressed his deep-passing volume. His EPA per play in November was actually slightly higher than his season average; the visible production was lower because the offense was structurally limited. The voting room reads the visible production. The metric reads the underlying execution. The two disagree.
Stafford at 37 has been producing the most-efficient passing season of his career but on a Rams team that has not produced enough win-loss prestige to drive the MVP narrative. He is third in adjusted EPA but seventh in MVP odds because the voting room has been treating the Rams as a wild-card team rather than a contender. The structural problem is that the MVP voting weighs team success higher than the criteria officially state. Our NBA MVP voter drift piece covers an adjacent version of the same problem in basketball — the voting room weights what it weights, regardless of the stated criteria.
Where this gets weird
The clean “narrative drives MVP voting” reading misses three things that complicate the discussion.
The first is that the narrative is not entirely wrong about Maye. His December performance has been the best in the league. His clutch metrics (EPA in the fourth quarter of one-score games) are the highest among the candidates. The voting room reading the recent games more heavily is partially capturing real information — Maye has been better than Allen across the most-leveraged minutes of the past month. The structural bias toward recency is also picking up a real signal about who plays best when it matters most. Both things are happening.
The second is that the Patriots’ season as a 10-2 underdog is itself an MVP credential. The “carrying the team” narrative is a real qualitative input that the metric-based ranking does not measure. Maye is producing his EPA on a roster that should have been a 7-9 team by preseason projection. Allen is producing his EPA on a roster that was projected as a 10-7 team. The Bills are outperforming projection by 2 wins; the Patriots are outperforming by 4. The “more responsible for the team’s success” argument favors Maye.
The third is that the MVP narrative cycle has historically been right about as often as it has been wrong. The post-1990 era of MVP voting includes both narrative-driven results that turned out to be analytically defensible (Tom Brady 2017) and narrative-driven results that look terrible in retrospect (Boomer Esiason 1988). The criteria the voting room actually applies — recent excellence, team success, “story” — produce a roughly 50/50 hit rate against the analytical evaluation. The next four weeks will decide whether the Maye narrative ages as 2017 Brady or 1988 Esiason.
What to track over the final four weeks
- Maye’s deep-passing efficiency. His season-long deep ball numbers have been strong. If they hold through the final four games against improving defenses, the MVP case is structural. If they regress, the narrative was driven by short-window variance.
- Allen’s red-zone TD rate. His December has been efficient but the Bills’ red-zone conversion has been bottom-fifteen. If that improves, his case rebuilds. If not, the narrative damage holds.
- Stafford’s primetime games. The Rams have two prime-time games left. Each is a chance to claim narrative oxygen. The MVP case requires national attention that the regular Rams schedule does not provide.
- Mahomes’s Chiefs in December. If Kansas City wins out and Mahomes posts +0.30 EPA/play in those games, the late-season finish could reset the MVP race entirely. Five weeks is enough time for one more shift.
The callback
That Sunday afternoon in Foxborough when Drake Maye produced sixteen straight completions and ESPN named him the MVP front-runner within four hours of the final whistle was the cleanest single example we have of how NFL MVP narratives actually form. The cumulative data did not change between Saturday and Monday. The voting room’s narrative did. The race that had been Josh Allen’s structural advantage at the start of December became Drake Maye’s by the start of the week. The metrics have not caught up to the narrative because the metrics do not move in narrative-time; they move in season-long-time. Both signals are real. The metrics measure what was. The narrative measures what is being remembered. The MVP ballot, in January, will be the document where the two compete. The trend-building piece covers the broader version of this dynamic. By February, the MVP will be Maye, Allen, or Stafford. The cumulative data favors one. The recent-game narrative favors another. The voting room will produce whichever answer that specific cohort of 50 writers happens to weight more this year. The fairness of that process is a separate conversation from the basketball it produces.
Quarterback metrics via Pro Football Reference; CPOE and EPA tracking via PFF and nflfastR; voting context via ESPN.
