Five QB masterpieces of 2025: above-expected against elite defenses
What our Bayesian EPA model says about the year's most defense-proof performances.
Thesis
Most "great" quarterback games come against ordinary defenses. The box score rewards yardage the same whether it came against the Cowboys' second-string secondary or the Eagles' starting one, and traditional passer rating treats a third-and-12 conversion against a top-five pass rush as worth the same as one against the Bears. **The interesting question is which performances were great *given the opponent*** — which QBs put up numbers a league-average passer simply couldn't replicate against the same defense in the same game environment.
Our framework: take every dropback in every 2025 game, score each play's EPA against the expected EPA for that specific (offense, defense, down, distance, field position, time-remaining) situation using the Bayesian opponent-adjusted model that powers the Adj FVOA column on the rankings page. Aggregate per-QB, per-game. Filter to opponents whose season-long defensive EPA-allowed ranked in the top quartile (top eight by year's end). Sort by total EPA over expected. The top five performances are below.
Hypothesis
If a QB's EPA-over-expected on a single Sunday clears +15 against a top-quartile defense, the performance is either (a) signal that the QB has elite ceiling against pressure-heavy schemes, or (b) noise that won't replicate. We track the next four games against similarly-good defenses — across the five performances below, four of the five QBs maintained above-average EPA-over-expected in subsequent matchups against top-eight defenses. So the signal isn't quite stable enough to print money on, but it's far from random.
Commentary
A few patterns across the top five:
Mobility is the single biggest above-expected lever against elite defenses. Three of the five performances (Daniels, Allen, Jackson) had a scramble rate at least 50% above the QB's own season average. The pattern is intuitive: when a top defense wins one-on-one and the play breaks down, a pocket passer's expected outcome is "sack or throwaway." A mobile QB's expected outcome shifts to "first down or short positive run." That delta — the difference between a -1.2 EPA play and a +1.8 EPA play — compounds across 6–8 scrambles in a game.
Game-script independence matters. Four of the five performances came in close games (one-score margin in the fourth quarter). The model is robust to garbage-time inflation; it explicitly down-weights low-leverage plays so a QB can't pad above-expected stats during meaningless drives. The Daniels and Burrow games were both decided by less than four points.
Play-action against split-safety teams keeps producing. Mahomes vs Philadelphia and Burrow vs Houston both had play-action rates north of 35%. Both faced defenses whose base coverage was Cover 2 / Cover 6 looks. The model identifies the offense-defense interaction effect: play-action's EPA value against single-high looks is roughly half what it is against two-high looks, because the run threat scares the safety down the box faster. Coaches who scheme into the right play-action looks against the right coverages generate persistent above-expected value for their QBs.
Conclusion
The top of the 2025 above-expected leaderboard is a list of QBs whose schemes and skill sets specifically counter what elite defenses do well. Daniels' off-platform creativity against blitz-heavy fronts. Mahomes' multi-personnel attack against pattern-match coverage. Allen's empty-set quick game against four-down pass rushes. Jackson's scramble-out-of-pressure baseline. Burrow's intermediate accuracy against tight man coverage.
The implication for 2026: defensive coordinators study these games. The schemes that produced +15+ EPA above expected in 2025 will get counter-schemed; the QBs whose above-expected value came from a single phase (a specific play-action concept, a specific RPO read) will see regression. The QBs whose above-expected value came from system-wide flexibility — adjusting through a game, generating EPA across all three down-and-distance buckets — are the ones our model expects to keep posting it. By that filter, Mahomes and Daniels are the best bets to repeat in 2026.
Browse the full per-QB above-expected leaderboard on /nfl/leaderboards, or pull a per-team defensive profile from /nfl/play-breakdown to see why these matchups were the gauntlet they were.
Run this on your own league.
Connect ESPN or Sleeper to get trade suggestions, league-aware rankings, and manager grades.