Research notes

The injury-prone label, measured.

July 12, 2026 · 6 min read

Our first research note ended on a cliffhanger: games played is one of the least sticky stats in football, and that finding is the reason our board projects established starters at a healthy season. This note is the full argument. We gave the “injury prone” label its fairest possible test, including the strongest version skeptics can ask for, and measured what it’s actually worth on draft day.

First, a correction in the label’s favor

The stickiness study reported games played correlating at .07 for RBs and .12 for WRs year over year. Honest disclosure: that sample required a full season’s worth of fantasy points, which quietly excludes the very seasons wrecked by injury. For this note we rebuilt the sample the fair way, qualifying players by their per-game role (20+ pass attempts for QBs, 10+ touches for RBs, 5+ targets for WRs) so a star who played six games still counts. Under that lens the correlation rises to about .19 at RB and .20 at WR. The label is not pure superstition. It contains a real signal, and the right question is how big that signal is.

Everyone misses games; history barely says who

Two things are true in that chart at once. The gradient is real: players coming off a heavily injured season (8 games or fewer) average about 4 fewer games the next year than players coming off a full one, and their odds of another lost season (10 games or fewer) are 52% against 19%. If your league-mate says the label means nothing, that row disagrees.

But look at the top row. The most durable cohort in football, players who just answered the bell 16 or 17 times, still averages only 13.6 games the following season, and misses a month or more 19% of the time. Missing games is not a trait some players have; it’s the baseline condition of the sport, and last season’s health moves your expectation around that baseline by a couple of games at most. At RB and WR specifically, the gap between the extreme buckets is about two to three games, on a stat whose season-to-season swing is enormous. Explaining 4% of next season is signal, but it is not the 20-spot ADP discount the label routinely inflicts.

The strongest version: chronically hurt players

One bad season could be luck, so we ran the skeptic’s best case: players who missed five-plus games in two CONSECUTIVE seasons while holding a real role. There were 69 of them. Their third season averaged 10.5 games against 13.7 for players who played 15+ in both years, and only 30% of them got back to a full season against 60%. So yes, chronic is worse than a one-off, by about three games. Even here there’s a catch the raw number hides: a two-year injury run also captures aging players being phased out, so part of that gap is role loss wearing an injury costume. Three games is the ceiling of what the label is worth, not the floor.

Quarterbacks are the real exception

One position breaks the pattern. QB games played correlates at about .50 under the fair sample, more than double any other position, and chronically hurt QBs are the one cohort with a genuinely alarming tail: quarterbacks with two straight injury-shortened seasons averaged just 8.2 games in year three. Quarterbacks are also the position where the floor under a lost season is different in kind; a top RB’s handcuff exists, but a top QB’s replacement is the waiver wire’s QB25. If durability belongs anywhere in your draft process, it belongs at quarterback.

We know because we tried to use it. Pricing each QB’s own expected games into his draft ranking measurably improved our backtested QB ordering, and we shipped exactly that for one day in July before pulling it back out. It quietly moved a healthy-lined Joe Burrow from QB10 to QB20, hedging the very upside a projection is supposed to show you. The ordering gain is real and stays in our ablation records; the product decision was that boards show healthy-pace stat lines and you price durability yourself, with this note as the price guide.

Why the board projects health

Our projections show established starters at essentially a full healthy season. That stance was re-tested over all four backtest seasons before we adopted it, and projecting proven starters healthy reduced points error at every position; the residual durability signal this note measures is small enough that hedging against it costs more accuracy than it saves. The deeper reason is the asymmetry of draft-day mistakes. When a starter gets hurt, the waiver wire replaces his missed weeks; when you let an injury discount talk you out of a healthy elite season, nothing replaces that. A model that haircuts every previously hurt star systematically buries the exact players who win leagues, to protect you from a risk worth two or three games.

The honest cost, stated plainly: when the injury does recur, our line misses high, and our methodology page names those misses (2025 Joe Burrow and Kyler Murray among them) rather than hiding them. We take that miss on purpose, because the alternative costs more.

Fine print: seasons 2015 to 2025; a game counts when a player logs any stat line that week, so the outcome measure blends injury with benching, suspension, and late-career phase-outs, which inflates the apparent injury signal rather than hiding it. Pairs require the player to appear in the following season at all, so retirements drop out. Role floors apply to the qualifying season only. Buckets pool QB, RB, WR, and TE; per-position numbers are quoted in the text.