The Stack Premium: Thirteen Seasons of QB/WR Stacking, Measured

Overview

Stacking, drafting a quarterback and his own top receiver, is one of fantasy football's most argued strategies and one of its least measured. The case for it is one sentence: when your QB hits your WR for a touchdown, you score it twice. The case against is the same sentence on a bad Sunday. Nearly every stack ranking you'll read this summer is built from highlight weeks and vibes about roles. We measured it instead: every team's top QB/WR duo, all 32 teams, for thirteen straight seasons. 416 pairs.

Three things fell out, and each one collides with how stacks actually get talked about. Stacking doesn't add points; it adds swing. The stat that stack rankings are built on barely predicts anything, while a boring one quietly does. And the new duos everyone is loudest about every July are the ones nobody, including us, can actually forecast. The 2026 Stack Board at the bottom applies all three, and we grade every row of it publicly in January.

Stacking doesn't add points. It adds swing.

Here's what a stack actually does to your team, measured across all 416 pairs: your QB's big weeks and your WR's big weeks show up together, so your combined score swings harder in both directions than the same two players would on different teams. On average, about 1.6 extra points of week-to-week swing, and it showed up for nine out of every ten pairs we measured. It's one of the most reliable effects in the whole study.

So what: a stack makes your ceiling weeks taller and your floor weeks deeper, from the exact same season average. That's the whole trade. It's the right tool when you need to win big, a tournament, a must-win week, a league where a monster ceiling takes the title, and it quietly works against you when steady weekly floors are what win your format. Every stack ranking that only talks about the upside is describing half of what you're buying.

The stat your stack rankings are built on barely works

Open any stack article this summer and the logic is target share: stack the receiver who gets all the targets, skip the offenses that spread it around. We tested that, along with eight other things you can know before the season, against how every returning duo actually performed the following year. Target share flunked. So did the pair's own chemistry, how tightly their big weeks moved together barely repeats from one season to the next.

What actually predicted next year's good stacks? Two boring things. Last year's output: good stacks stay good, and it's not close, this was the strongest signal we found. And the overall quality of the offense, measured by how efficient the team is on a normal snap. A great offense keeps feeding its duo even as roles shift underneath.

So what: when you evaluate a stack, ask two questions. Did this duo produce last year, and is the offense around them actually good? Skip the target-share deep dive and skip anyone selling you "elite QB-WR chemistry" as if it carries over. The data says it mostly doesn't. (Full test tables, thresholds, and the corrections applied are in the methodology note and the committed study.)

The duos everyone is loudest about are the ones nobody can forecast

Every July, the most hyped stacks are the brand-new ones: the traded star, the rookie QB, the reunion. So we tested 266 first-year duos across thirteen seasons: knowing everything about the pieces, how good the receiver was at his old stop, how good the destination offense was, how the previous duo produced, how well does that predict what the new pair actually does? Answer: barely at all. The best single clue we found explains about 3% of what happens. The other 97% is the part nobody knows until they play.

So what: when someone tells you a new duo is a lock, they are confidently forecasting something the last thirteen years say is close to a coin flip. That includes us: our board can describe a new pairing's ingredients, and it does below, but it cannot tell you the dish. Anyone who claims otherwise is selling confidence the data doesn't support.

Honesty check, so the Buffalo story can't cheat: 14 duos entered 2025 with a profile like Buffalo's. 8 of the 14 beat the average stack. Better than a coin flip, not a cheat code, and we publish the count so you can weigh the story instead of just enjoying it.

The 2026 Stack Board

Built on the two things that actually predicted: last year's output and the quality of the offense. The chemistry column is shown for context only, the study's own finding is that you shouldn't project it forward. Rosters verified against league pages and primary reporting as of July 20, 2026, and cross-checked against our win-total note; anything the reporting can't settle is marked OPEN instead of guessed.

The short list: where the framework's evidence is strongest

The full board below exists for completeness and for January's grading. For draft preparation, the framework's case concentrates in a handful of returning pairs that score highest on both surviving predictors, prior combined output and team EPA, with full-season samples behind them. Highest-rated is a description of the evidence, not an instruction:

• Stafford / Nacua (LA), 40.6/wk, +0.161 EPA. First on every surviving metric, with the largest gap between first and second the board has. The market knows; the framework simply agrees.
• Goff / St. Brown (DET), 34.2/wk, +0.092 EPA. Three seasons of pair history, elite output, no roster caveats. The cleanest profile on the board.
• Darnold / Smith-Njigba (SEA), 32.9/wk, +0.078 EPA. The league's highest WR1 target share on a Super Bowl offense, and a full 17-week sample.
• Allen / Shakir (BUF), 31.9/wk, +0.089 EPA. The case study. The consensus still dismisses it on the one metric that doesn't predict.
• Burrow / Chase (CIN), 33.7/wk, with the flag attached. Elite when together; the 8-shared-week sample means the per-week rate is less certain than the four above.

Tier A: Returning pairs (the study's evidence applies directly)

Tier B: New or changed pairings (Finding 3 applies: low confidence by construction)

Caveats and limitations

Single-season metrics are noisy; several board rows rest on 8 to 13 shared weeks and are flagged where the sample is thinnest. Pair correlation is displayed but should not be projected, that is Finding 2's own conclusion. The 2025 rows for teams with mid-season QB changes (SF, WAS) describe the interim reality, not the healthy pairing's baseline, and the notes say so. Half-PPR and full-PPR orderings agreed everywhere it matters. The radar and pricing side of stack value, what these pairs cost in drafts, is deliberately absent: public ADP archives for the current summer were unavailable at study time, and we would rather omit the cost column than fake it. And the projection tier's low confidence is not modesty theater; it is the measured result, and it applies to our board exactly as much as anyone's.

Methodology note

Study WS-24. Primary stacks defined per team-season as the QB with the most dropbacks and the WR with the most targets. Weekly fantasy scores computed from raw nflverse statistics; pair metrics computed over weeks both players appeared (minimum 8 shared weeks for correlation). Team EPA from neutral-situation play-by-play (win probability .20 to .80, quarters 1 through 3). Three pre-registered tests with Bonferroni correction; full tables, scripts, and verdicts are committed to our research repository, and the pipeline's key numbers were independently recomputed before publication (three team-season spot checks, all exact matches). Temporal fence throughout: every predictor is computable strictly before the season it predicts.