Draft Capital (DPOS): The Strongest Single Signal
Data Dictionary · what the NFL's own behavior tells you about a prospect
DPOS is the draft-capital piece of DMX. It takes a prospect's pick number and turns it into a z-score within his position, with undrafted guys parked at pick 300. Basically, it boils the NFL's entire scouting operation — 32 front offices spending real money — down to one number. And it's the strongest single predictor of career outcomes there is, at every position.
The numbers
On its own, draft capital explains somewhere between 16% and 22% of career outcomes depending on the position (it's on the Model Transparency page) — more than any other single input, testing and college production included. Put it in hit-rate terms. Split the 2001–2020 cohort into draft-capital quartiles within each position and here's what you get:
| Draft capital | N | Star or Starter % | Bust % |
|---|---|---|---|
| Top quartile | 508 | 68.3% | 12.2% |
| 2nd quartile | 507 | 36.5% | 36.7% |
| 3rd quartile | 506 | 12.1% | 68.2% |
| Bottom quartile | 505 | 14.5% | 68.3% |
Offensive prospects, 2001–2020 classes, DPOS quartiles within position. Live query of the public database.
The top quartile of draft capital hits at about five times the rate of the bottom half. That's why “fade the combine, trust the draft” is half-right — collectively, the NFL is a really good scout.
Why within-position z-scores
Raw pick numbers don't travel across positions, because the league prices positions differently — quarterbacks get drafted ahead of their talent on scarcity, running backs get shoved down because they're easy to replace. Normalizing within position asks the question that actually matters: how much capital did this guy command against others at his own job? A pick-40 RB and a pick-40 QB are telling you two different things, and DPOS captures that.
Why not only draft capital
If DPOS were enough on its own, the rest of DMX would be pointless. It isn't — the full blend beats capital alone at every position (the side-by-side R² is on the Model Transparency page). Here's why: draft position is a market price, and markets get things wrong. Teams reach for need, overpay for workout numbers, and let productive-but-unsexy guys slide. Testing and age-adjusted production are the two independent checks that catch those mistakes, in both directions. And honestly, the most useful signal isn't the capital or the model by itself. It's where the two of them disagree.
Where capital decays
Draft capital is a draft-day read, and the halo wears off as real games pile up — a first-round pick gets two or three years of benefit of the doubt, not ten. I model that fade on the in-career side: DPX runs a draft-capital decay term in its dynasty modifier, so a former first-rounder's pedigree counts for less every season his production doesn't back it up.
Frequently asked questions
Is draft capital all that matters for rookies?
It's the strongest single signal, but it's not the whole story — the full DMX blend outpredicts draft capital alone at every position. The market misprices guys all the time (reaches, workout overdrafts, productive players sliding), and testing plus age-adjusted production are what catch those misses.
How big is the hit-rate gap by draft slot?
Big. In the 2001-2020 cohort, top-quartile draft capital within a position hit (Star or Starter) at 49.5%, versus about 12% for the bottom two quartiles — where nearly three in four guys busted.
How are undrafted free agents handled?
I give UDFAs pick 300 before the within-position z-score, which keeps them in the dataset at the very bottom of the capital scale instead of tossing them out.
Why is DPOS computed within position?
Because the league prices positions differently — QBs go ahead of their talent on scarcity, RBs get pushed down because they're replaceable. A within-position z-score isolates how much capital a guy commanded against his own job market.