You can find all the previous editions of the Bayesian Quarterback Rankings here, and the backlog for Adjusted Quarterback Efficiency is here.
GO-FORWARD PROJECTED ADJUSTED EFFICIENCY
This update includes a pretty big re-work of the methodology, which projects more useful EPA per game numbers, along with details on exactly how those projections have moved from the prior week and through the season.
Additions to this update include:
Removing the YAC and sack adjustments from the adjusted EPA data going into the model, with testing confirming those are less “luck” based than contextual information that enhances projection accuracy.
Giving additional weight to “breakout” seasons (those occurring in years 2-4 +0.1 adjusted EPA higher than any previous season) to mitigate poor prior performance that’s less predictive.
Establishing a “starter” designation that raises their draft prior to match franchise quarterbacks taken in that range, not backups filling in. The starter designation is phased in over the first 1,000 career snaps.
“Stretching” (or shrinking) the projections to match the current EPA efficiency environment. We’re currently in a phase of enhanced efficiency since roughly the quarter-point of the 2025 season.
Capping older season results at +/- 1.5 standard deviations to smooth outlier effects that are less meaningful to today’s projections.
Lowering the weight of prior-team results to roughly 50% of current team.
These changes were all made through the iterative process of extensively testing ideas and measures at different levels and choosing the parameters that ended up with the most predictive power.
These rankings are for the go-forward projections of adjusted quarterback efficiency as of this week. They do not account for opponent strength or injuries to supporting cast, both of which go into the weekly Power Rankings.
Speaking of power rankings, this update to the Bayesian rankings and projections has been highly useful in determining quarterback adjustments when a team’s starter changes due to injury or benching.
In the tables below, I also included the adjusted EPA per play rankings for each quarterback over 2026 (min 100 play involvements) and the prior nine seasons (minimum 300 plays) so you can see the evidence going into the projections.
Older data is decayed over time, so the 2026 and 2025 EPA per play data matters more than those from prior years. That said, older data can’t be fully discounted, or else you miss bounce-back performers of great quarterbacks returning to form, like Aaron Rodgers returning to form after a few poor statistical seasons and winning his third and fourth MVPs in 2020 and 2021.
TIER 1
Well, we finally did it. The last update of the Bayesian rankings a couple weeks ago had only the tightest gap between Josh Allen and Patrick Mahomes, so it didn’t require a major rework to get Allen into the No. 1 spot. While some of the methodology changes did help Allen relative to Mahomes, he still would have been No. 2 to start the season either way, as you can see that his year-to-date projections lead is more than his current gap with Mahomes.
There’s probably less to contemplate with what could be off with Allen’s top ranking than Mahomes’ second-place position. Mahomes is only 14th in adjusted EPA this year, following multiple seasons that he didn’t crack elite status (top-8).
He’s still held up by the insanity of his earlier-career run, which combined having the top rankings with off-the-chart levels of efficiency versus the rest of the NFL. Has he been figured out? I’m not willing to say that yet, and the projections still see a material gap from Mahomes to the rest of the NFL.
