We Graded Our Own Buy Signals Across Three Seasons. Here's Every One.

We graded our own buy signals across three seasons. Here's every one.

Every fantasy tool will tell you a player is undervalued. Not one of them will tell you how often they were right.

So we graded ourselves, across three seasons, and published the whole list — including the misses.

The signal

When a player's scoring falls well below what his workload should produce — he's getting the snaps and the targets, but the points haven't shown up — that gap is a buy window. The box score says he's cold. The usage says he's about to not be.

This idea isn't ours. Expected-points analysis is published by PFF, FantasyPoints, 4for4 and others, and it's a well-established way to look at fantasy players. What nobody publishes is whether their own version of it actually works.

Here's ours.

The result

We took every player Rook flagged as producing below his usage in 2023, 2024, and 2025 — using only data available at the time of the flag, no hindsight — and checked what happened to his scoring over the next four weeks.

Flagged players bounced back 78% of the time. Comparable low-scorers who weren't flagged bounced back 58.5% of the time. Average gain for the flagged group: +3.74 points per game over the following month.

The effect held in every season — 2023, 2024, and 2025 all independently positive. Pooled across three years, p ≈ 0.014.

Why the control matters more than the headline

Here's the number a careless version of this article would have led with: over all player-weeks, scoring rises about 49% of the time. Against that, 78% looks like a +29-point edge.

That number would be dishonest, and we're not using it.

Low scorers bounce back anyway — it's regression to the mean, and it's the single biggest confound in any analysis like this. A player who just had two bad weeks tends to have a better next four weeks whether or not anybody flagged him.

So we built a matched control: comparable low-scorers, same seasons, same scoring range, who were not flagged. They bounced back 58.5% of the time. That's the number to beat, and it's the only honest comparison.

78% vs 58.5%. That's the real edge, and it's the one we'll defend.

All 41 players, including the nine misses

The point of publishing this is that you can check it.

2023 — 12 of 17. Hits: Terry McLaurin (5→17), Davante Adams (6→14), Chuba Hubbard (8→16), Michael Pittman (9→16), Garrett Wilson (9→13), Jonathan Mingo (3→9), Rhamondre Stevenson (8→10), Austin Ekeler (10→11), and others. Misses: Saquon Barkley (16→14), Tony Pollard (13→11), Zack Moss (9→8), Dawson Knox (3→3), Ke'Shawn Vaughn (1→1).

2024 — 7 of 9. Hits: Calvin Ridley (2→17), Chuba Hubbard (10→21), Nick Chubb (6→12), Javonte Williams (9→12), Kyren Williams (13→15), Rome Odunze (7→10), Jonathan Taylor (10→10). Misses: Ja'Lynn Polk (3→2), Carson Steele (2→2).

2025 — 13 of 15. Hits: Tony Pollard (6→17), Wan'Dale Robinson (10→19), Emeka Egbuka (7→13), Ashton Jeanty (11→15), Saquon Barkley (12→15), Quinshon Judkins (7→11), Jordan Addison (4→8), and others. Misses: Jerry Jeudy (5→4), Alvin Kamara (11→9).

Numbers are points per game before the flag → points per game over the following four weeks.

The signal we're not publishing

We ran the same study on the inverse — players scoring above what a declining role supports, which should be a sell signal.

It doesn't work. Flagged players' scoring fell 66.2% of the time. Comparable high scorers who weren't flagged fell 69.5% of the time. Our flag added nothing. Hot players cool off whether or not anyone calls it unsustainable — that's just regression, and dressing it up as a signal would be a lie.

So we don't ship it as one, and we're telling you that instead of quietly keeping the good half.

What's fair to say against this

The sample is small. Forty-one players across three seasons. The flag is rare by design — it fires for roughly fourteen players a year — so this is as much data as three seasons can produce without changing the method. It's enough for a real result (p ≈ 0.014, and it replicated every year), but it isn't thousands.

2025 was the best year. At +26 points it was the high mark; 2023 and 2024 both came in at +17. The honest, poolable number is +19.5, which is what we've led with. If we'd quoted 2025 alone, we'd have been flattering ourselves.

One control construction. We compared against matched low-scorers within the same season, which is the right way to isolate regression — but it's the only control we tested.

One method, run three times. We did not tune the threshold, the window, or the control to improve the result. The 2023 and 2024 runs used the identical harness and replicated on the first attempt.

Why we published the misses

Because a hit rate without the misses is marketing, and a tool that won't grade itself is asking you to take its word for it.

Jerry Jeudy didn't bounce back. Neither did Alvin Kamara, or Ja'Lynn Polk, or Saquon Barkley in 2023. Twenty-two percent of the time, the usage said one thing and the season said another. That's the real number, and it's the one worth knowing before you trade for somebody.

Rook builds every player's value from in-season usage and the causes behind it, then scores trades by what they do to both starting lineups. Try it free — 30 credits, no card.