How do you spot a college football breakout before your league does?

Sportsbooks now post weekly numbers on college players. When a player's role jumps, that number lags — and the size of the gap is the whole signal.

GWTTKB Research · 22 August 2026
25,357 graded prop lines · 14,016 projected player-weeks · 2024–2025

In the NFL, everyone has the same information at the same time. Snap counts are published Monday, injury reports are mandatory, and by Wednesday every projection system in the industry has repriced. College football has none of that. There are no injury reports. There are no snap counts. And until recently, there was no market on individual players at all.

That changed. Sportsbooks now post player props on most college games — receiving yards, rushing yards, passing yards, anytime touchdown — and those numbers are projections made by firms with money at risk. We bought two full seasons of them: every closing line, every book, 2024 and 2025.

Which finally lets us ask the question that matters for a college fantasy manager: when a player's role changes, how long does it take the market to notice — and can you get there first?

The signal, in one number

We built a projection for every propped player-week by converting his posted lines into expected fantasy points, then compared that to what his own trailing five games said. The interesting cases are the disagreements — specifically, when a player's usage is rising and his recent form projects well above the market's number.

And it isn't just that they outscore. In that same bucket, the trailing data is more accurate than the market — mean absolute error of 6.90 versus the market's 8.24. That is a 1.34-point accuracy edge over a firm that prices these for a living.

Read the top row carefully, because it is the honest half of this finding. When the market is the one projecting higher, the market wins. This is not "we beat Vegas." It is narrower and more useful than that: we beat Vegas in one specific direction — on players whose roles are expanding.

What it looks like on a real player

Carnell Tate, WR, Ohio State

Tate is the cleanest example in the dataset because his role expanded gradually and the market chased it the entire way without ever catching up.

Follow the line: 57.6 → 60.6 → 53.7 → 62.5 → 65.5 → 67.5. Over eight weeks the market moved him up about 17%. Over those same weeks he posted 101, 183, 111 and 124 receiving yards. The market was directionally right and magnitudinally wrong the entire season — it kept nudging while he kept doubling.

Note week 6 in particular. He went for 183 yards on a 53.7 line with a 37% target share, and the following week the book posted him at 62.5. A 16% bump after a 240% game.

Elijah Sarratt, WR, Indiana

Sarratt's 2025 is the target-share version of the same story. His line sat in the 71–86 band all year while he was running target shares of .435, .333 and .433.

Noah Whittington, RB, Oregon

The most extreme case is the one where the market did not post a number at all. Through the back half of 2024, Whittington had no receiving-yards market in most weeks while his trailing average climbed from 4.58 to 12.20 fantasy points. He scored 14.6, 20.6 and 15.6 in that stretch. The book's implied view of him never updated because the book never priced him.

Why college lags and the NFL doesn't

Three structural reasons, all specific to the sport.

Books price off season-long usage. A prop model needs a usage prior, and the natural one is the player's share of his team's targets or carries across the season. In the NFL, roles are stable enough that this works. In college, a true freshman can go from an 8% target share to 30% in three weeks — and a season-long average will keep dragging the number toward a role the player has already outgrown.

Coverage is thin and uneven. Books post props on marquee games and skip the rest. In our sample only about 750 players per week get a line at all, and many get a single market from a single book. Less coverage means less repricing pressure.

There is no injury report. Absence has to be inferred from usage gaps, by us and by the books. That creates a window where a role has changed for reasons nobody has published.

The other triggers we measured

Working from the same 14,016 projected player-weeks, we tested a battery of contextual triggers against the blended projection. Effect is in fantasy points relative to expectation — negative means the player underperforms his projection when the trigger fires. Every row below replicated independently in both 2024 and 2025 except where noted.

Two of these are worth dwelling on because they invert the intuition.

Alpha target share is a trap. A receiver commanding 28% or more of his team's targets underperforms his projection by 1.20 points. The concentration is real — it is simply already priced, and then some. High target share is a description of what happened, not a prediction of what will.

Vacancy does not help the backup. When a starting running back is out, the instinct is to start his replacement. But the whole offense degrades — worse protection, defenses keying differently, fewer scoring opportunities — and the teammates left behind underperform by 0.37. The unit gets worse faster than the replacement gets better.

Quarterback change is the most stable effect we found. −0.46 in 2024, −0.46 in 2025, across 1,152 player-weeks. When the quarterback changes, every skill player in that offense loses roughly half a point against his projection. It is also the trigger the market handles worst.

What this means for a lineup

Concretely, the checklist that falls out of the data:

Buy: rising usage over the last two games versus the two before it; a posted line that has not kept pace with recent production; a bellcow carry share; a high team implied total; and — counter-intuitively — the absence of a market on a player whose role is clearly expanding.

Fade: a quarterback change anywhere in the offense; a teammate's absence being treated as a reason to start the backup; an already-elite target share; and a low team implied total.

How far this generalizes

Roughly 750 players a week get a prop. College fantasy pools are much larger than that, so most of your roster will never have a market. We handled this by training a model to reproduce the market's projection from context alone — trailing usage, PPA, implied team total, spread — on the players who do have lines, then applying it to the ones who don't.

It reconstructs the market's number closely: mean absolute error of 1.43 at tight end, 1.94 at running back, 2.15 at receiver. Applied to 15,504 unpropped player-weeks, it projects more accurately than trailing average alone (4.94 versus 5.08, and 4.78 blended). So the same breakout test can run on a Group of Five receiver no sportsbook has ever priced.

The short version

College betting markets are informative but slow. They price players off season-long usage in a sport where usage changes in a fortnight. That lag is not a flaw you can bet profitably against very often — but it is a gap you can exploit every single week in a fantasy lineup, because fantasy charges no vig on being right.

Watch the role. Check whether the number moved. If the role is growing and the number is not, you have found the player your league is still waiting on.