Week 1 is a 16 percent signal. You're trading like it's ninety.

What this means for your roster

Buy Drake Maye at our QB2 and Bo Nix at QB7 from anybody who just watched them score nine and five points and decided the sky fell. Sell Bryce Young into his QB3 week, our board has him QB21 and that gap is the whole trade. Sell Jacoby Brissett at our QB27 to the manager who saw him finish QB15. Hold Jaxson Dart at QB5 and Caleb Williams at QB6 and refuse the sell-high haul, because our board had both of them inside the top six before Sunday, on what they had already done, and one week of confirmation is not a reason to take a discount on your own read.

One more, the uncomfortable one: Patrick Mahomes is our QB16 in dynasty and he finished week 1 as QB9. That is the week arguing against us, in public, and we are not moving him. Read on for why that is the rule and not an exception to it.

What the consensus says

The genre arrived on schedule. DraftKings Network ran top quarterback sells to trade away and named Jaxson Dart at QB9 and Caleb Williams at QB4, the argument being that "the price tag will likely never be higher." CBS was rebuilding waiver priorities off one Sunday. RotoWire filed week-1 reactions. Bleacher Report and Yahoo both ran overreaction columns.

And here is the tell. ESPN's own week-1 piece is titled "Resist the urge to overreact to Week 1". The same industry that says resist the urge spent the week publishing the urge. Nobody ever puts a number on how much resisting you should do, which is convenient, because a number would end the genre.

The claim, in plain English

The claim is that a quarterback's week-1 result is news you should trade on. Turn that into something you can check: rank every quarterback who actually started week 1 by what he scored, rank the same quarterbacks by what they averaged from week 2 onward, and ask how much the first list knows about the second. That is it. No archetypes, no narrative, one number per season.

The control is the whole game. I am not asking whether week 1 correlates with anything, I am asking how much of the ordering it recovers, because "week 1 matters a little" and "week 1 should reorder your quarterback room" are very different bets and the columns are selling the second one.

How I beat on it

Two reads, both on the scoring our boards use, half-PPR with a tight-end premium. The historical control: eight seasons, 2018 through 2025, every quarterback who genuinely started week 1 and then played at least eight more games, 258 week-1 starter seasons. The live read: our four served quarterback boards, dynasty and redraft crossed with one-quarterback and superflex, graded against what the top 24 did in week 1 of 2026.

That second read is the first out-of-sample week our quarterback model has ever faced. 2026 is a locked holdout here, so no 2026 row may touch a calibration, ever. The boards were built on seasons through 2025 and never saw a 2026 snap, and the script asserts that in code before it prints a thing. Full framework: the autonomous research spec.

What the data actually said

First swing, the control. Across eight seasons the week-1 quarterback order lines up with the rest-of-season order at a rank correlation of 0.40. Square it and you get the number that matters: week 1 explains sixteen percent of what follows. Eighty-four percent of the season is still sealed on Monday morning.

And it is getting worse, not better. The two most recent seasons are the two weakest in the sample.

Now the part I did not expect. Our preseason board, blind to 2026, predicted week 1 at 0.34 in dynasty superflex and 0.39 in redraft. Sleeper ADP over the same quarterbacks landed at 0.38 and 0.40. Nobody won. A board built from seasons of production and a crowd of drafters agreed with one Sunday to the same weak degree, and that degree is the same size as week 1's own grip on the season.

So when you reprice a quarterback on Sunday's result, you are not upgrading from ignorance to information. You are swapping one 16 percent instrument for another and paying a trade tax to do it.

The misses are where it gets useful. Our board was too low on Bryce Young, QB21 into a QB3 week on 31.44 points, and on Patrick Mahomes, QB16 into QB9. Too high on Drake Maye, QB2 into a QB27 week on 9.82 points, and on Bo Nix, QB7 into QB31 on 5.44. Average rank error across the top 24, about five and a half spots. That is a board roughly right about the shape of the position and unable to tell you which Sunday is which, which is exactly what a 16 percent world looks like.

What the engine already figured out

This is the rare case where the answer was already baked in and the receipt just proves it out loud. Our quarterback projection is built on multi-season weighted production with recency weighting, and a single game cannot move it, by construction. No week-1 override, no hot-start multiplier, no knob for "he looked great on Sunday." Sixteen percent says that was the right build.

So there is no engine change in this post, and that is not modesty. What I measured is our own board's rank against an outcome, and a board rank is just our production math sorted. Grading it teaches us how well we did, never what new input to add. It stays a receipt.

Which brings back Mahomes, our QB16 at 31 years old, week-1 QB9. If I moved him up on that, I would be doing the exact thing this post says costs you money, on the exact size of evidence this post prices at sixteen percent. He moves when six or eight games say he moves. A rule you abandon the first time it points at your own board was never a rule.

What to do about it

Trade against the repricing. The people selling Bryce Young's QB3 week and buying Drake Maye's nine points are handing you both sides of a good trade, and the only edge required is refusing to learn much from one Sunday.

The 2026 names

Current board ranks, dynasty superflex first, redraft one-quarterback second where it differs meaningfully.

One honest caveat: sixteen percent is not zero, and the effect was positive in all eight seasons I measured, so a week-1 result genuinely is weak evidence rather than no evidence. Let it nudge a ranking by a spot or two. Do not let it price a trade.

← Back to all posts