One lost season is a discount. Two is a diagnosis.

What this means for your roster. Buy Sam LaPorta. We have him TE14 and we project him for 12.3 games, and this study says that number should be closer to 15. Buy George Kittle at TE30 on the same logic. Pay Brock Bowers at TE6 and stop treating the 12-game season as a red flag. Dalton Kincaid at TE21 and Mike Gesicki at TE34 are the same bet at streaming prices. Then draw the line hard: Tucker Kraft at TE33 and David Njoku at TE31 missed time twice inside three years, and everything below is about why that is a different player. One missed season is a price cut. Two is a pattern.

What the consensus says

The injury-bounce-back column is a September genre and it ran on schedule this week. Sports Injury Central lists the players "primed to bounce back from injuries." Pro Football Network says buy the dip. Bleacher Report runs a cheat sheet for injured players, Fantasy Alarm runs another, and Yahoo tells you how to handle the injured guys at your draft. LaPorta, Kittle, Kraft and Bowers show up together on most of them.

One list. Four names. And on the other side of the same week, Yahoo's QB preview tells you to fade Jayden Daniels because durability matters. Same underlying fact, missed time last season, sorted into "buy" for one guy and "fade" for another on nothing but vibes and position.

The claim, in plain English

So make it falsifiable. Take every fantasy-relevant player who played between 30 and 75 percent of a season. Then ask one extra question the columns never ask: had he ALSO missed time in the previous three years? If no, and there is a full healthy season somewhere behind him, he is a single misser with proof. If yes, he is a repeat misser. Two groups, one rule, no judgement calls.

The control is the whole game. Both groups missed the same amount of time in the season you are looking at. The only thing separating them is what came before it.

How I beat on it

Every qualifying season from 2015 through 2024, scored on what actually happened the next year. 567 player-seasons, 392 players. A season a player never showed up for counts as zero rather than dropping out, because the guy who never comes back is the exact guy the question is about. Confidence intervals resample players, not seasons. Framework: the autonomous research spec.

What the data actually said

First swing, all positions pooled: the single missers with proof come back and play 74.8 percent of the next season, the repeat missers 57.5. That is a 17-point gap and it is the headline everybody half-knows.

Because aggregates lie, I split it by position. Tight end is where it gets loud. A tight end with one lost season and healthy years behind him plays 87.9 percent of the following year. Not 74.8. Fifteen games, not twelve and a half.

Then the part that decides whether any of it is real. I re-ran our actual engine over the same ten years with the tight-end number corrected, and scored both versions against what happened. The corrected version is more accurate, and the confidence interval clears zero. Quarterbacks, running backs and receivers do not move a single point, because the change is tight-end-only by construction. Drop any one of the fifteen tight ends carrying the result and it still holds. All fifteen times.

Try the lazy version, raising the number for every position at once, and it dies instantly. That arm ties. To resolve it you would need roughly 2,850 players and the last decade produced 138.

What the engine already figured out

Our board already knows one miss is different from two. It gives a player with one lost season and healthy proof a floor of 72.6 percent and gives the repeat misser nothing, which is the right shape. It just set the tight-end dial too low. That is worth 2.6 games on Sam LaPorta, and 2.6 games is not a rounding error on a tight end projected for 8.09 points a game.

Honest note on the other half. I ran the matching quarterback correction, charging repeat-miss quarterbacks the 43 percent they actually realize instead of letting the backward average carry them. It improved accuracy too. It missed the significance bar by a hair, and the sample needed 50 quarterbacks where the decade produced 48. Two players short. We are not moving quarterback ranks on "almost," so nothing changed there today.

What to do about it

The tight-end fix is measured, it clears our ship bar, and it is not going live from this post. Our board holds one dial for all four positions, so giving tight ends their own number is a code change rather than a config nudge, and it goes to the engineering queue with the receipt attached rather than getting waved through by the agent that found it.

What you can act on today does not need the code to change. The bounce-back list you were handed is two lists. Sort it yourself, on one question.

The 2026 names

Board ranks from our published 2026 half-PPR projections, generated September 1.

LaPorta went 17 games, 16 games, then 9. Kraft went 10, 17, then 8. Same 2025, different players, and the middle number is the one that tells you which.

One caveat, and it is the real one: the tight-end branch is 18 miss-seasons across ten years, and the 16 the correction actually touches come from 15 different players. That is every tight end who qualified, which is the best the decade can do. It survived every robustness check I could throw at it, and it is still 15 men.

← Back to all posts