2026 Busts: Early Picks Our Model Fades

What the fade selector does

Every fade on this page starts from one gap: how far below current ADP our model ranks a player. Say the market takes someone at pick 40 and our board has him 80th overall. That's a 40-spot reach, and it qualifies for this list. We flag anyone our model ranks 15 or more spots later than his ADP, inside the top 60 picks.

Past pick 60, the gap stops mattering. Replacement-level players swap places every week of August, and calling that swap a fade is noise wearing a headline. Both cutoffs, the 15-spot gap and the pick-60 ceiling, are v1 judgment calls, not backtested constants — the same disclosure standard the values list holds itself to. If we retune either number, this page says so.

The rank side of that gap comes from the same board as the rest of the site: value over positional baseline rather than raw projected points, built by the four-step pipeline on the methodology page. That distinction matters more here than anywhere else on the site. A fade is a claim that the model is right and the market is wrong, and you should know exactly what the model means before you spend a mid-round pick believing it.

Here's this run's full reach list, recomputed against live ADP on every visit:

The two populations

Two very different situations land a player on that list, and the table above can't tell them apart. Call the first population type-A. These are the honest fades: established players with years of stable usage, whose role the model reads as clearly as it reads anyone's. When one of those players still prices below ADP, the market is paying for a name more than a role, and the gap should change how you draft him.

Terry McLaurin is the clean type-A case this run. Multiple full seasons of high-volume usage give the model a confident read on his role, and that confident read still lands well below his current ADP. Dak Prescott and Mike Evans price the same way on this run's list, established veterans the market pays partly on name recognition. One extra beat of scrutiny before you act: McLaurin and Evans are wide receivers, the one position our backtest lost to a simpler baseline — run check three twice there. The number is on the methodology page.

That's the fade worth respecting on draft day, the rare case where the model's read and years of tape agree against the market.

Call the second population type-B. These are the model's blind spots. Here's the mechanism.

Our usage numbers come from a player's last two seasons, weighted toward the more recent one. That's the same pipeline the methodology page walks step by step. A separate games-played prior, built from recent availability, turns those per-game rates into a season total. A player whose most recent season was thin on games gets hit twice by that design: once in the rate his shares are built from, again in the games prior that stretches or shrinks it into a full year.

Joe Burrow is the clearest example on this run's list. Our board prices him off a thin 2025 sample and the games prior that season creates, with no way to update either one before games are actually played this year. Rashee Rice and Malik Nabers land in the same column, for the same reason: a thin 2025 sample the model has to weight heavily, whatever the reason behind it.

Read that number as a blind spot, not advice. Our rank reflects a thin sample and nothing else; it has no separate view on where he stands right now. If you have that view, it's worth more than our rank until real 2026 snaps exist to weight.

Omarion Hampton shows a related but different blind spot. He's early enough in his career that the model has little established NFL usage to weight, so it leans on the position-average priors the methodology page describes for the situation closest to his. The market can price a depth-chart promotion or an expanded role the model has no input for yet. That gap is the model being honest about what it doesn't know, not a verdict on Hampton himself.

How to tell which is which

You don't need our internal notes to sort a fade list into its two piles. Three checks do most of the work.

First, check whether last season ended early for him or his role just changed. Either one means the model is weighting a thin sample, and the market's read outweighs ours right now.

Second, look at the projection range on his player page. A P10–P90 band spanning half the position's scoring spectrum means the model is guessing, and it knows it.

Third, ask whether he has three-plus seasons of stable, full usage behind him. If he does, that's the model's best case, and a fade that survives it deserves real weight.

The methodology page names rookies and thin-usage players as the model's weakest spots on its own terms. This section is that same limitation, aimed at one list. Sorting the two piles is the edge this page exists to hand you.

Using fades at the table

A fade only pays off if you act on it at the table, not just in your notes app.

Don't pay ADP for a type-A row. When years of stable usage and the model agree that a player is priced on reputation, let someone else spend that pick. Take the discount waiting at this run's ADP values instead, or the next name on the full rankings. Those reputation fades are the cleanest edge on this list.

Type-B rows need the opposite discipline. If a name is on this list only because of a thin, shortened, or rookie-adjacent 2025, the market's price is doing better work right now than ours is. Draft him at ADP if you'd have wanted him anyway, and don't let our rank talk you out of a player you already trust. Expect our board to look wrong on those names right through August.

That won't last. A weekly, game-level model starts grading live from Week 1, seeing the role changes and updated usage this draft-only board never will. Until then, pair this list with the draft strategy guide or how to win your league guide for how one overpriced pick fits into the rest of your draft.