What is DMX (Dynasty MetriX)?

Data Dictionary · the complete guide to the DMX rookie score

Why it matters

Dynasty rookie drafts are won and lost on priors. Consensus rankings drift toward film takes, draft-media narratives, and recency; DMX is a fixed, auditable prior built the same way for every prospect since 2000. It won't tell you who to love — it tells you what 25 years of comparable profiles actually did in the NFL, before the hype cycle gets a vote.

How DMX is calculated

Three components, each expressed as a z-score within the prospect's position cohort:

The composite blends the three with per-position weights fit by maximum likelihood against career outcomes on the 25-year training set, and refit roughly every three years. The specific weight values are proprietary and not published — anyone can average three z-scores; which blend best predicts NFL outcomes is the product of 25 years of refinement. Everything downstream of the weights — scores, deciles, regression coefficients, calibration — is published openly on the Model Transparency page.

What the deciles mean

Raw DMX is continuous; the familiar D1–D10 labels are decile ranks assigned within each position and draft-year cohort. Here is what each decile actually produced, across every offensive prospect from the 2001–2020 classes with a completed career window (career tiers are derived from value-over-replacement percentiles within position):

Offensive prospects, draft classes 2001–2020, n = 1,853. Star / Starter / Contributor / Bust tiers from career VBD percentiles within position cohort.

Read the gradient, not any single row: a D1 rookie is roughly 10× more likely to become a Star than a D8–D10 rookie, and four out of five D10 prospects bust. No model makes individual guarantees — the value is in stacking these probabilities across every pick you make for a decade.

A worked example

Take a real Decile 1 prospect from the 2026 board — Jeremiyah Love, RB:

Live scores from the DMX Draft Board, June 2026.

This is the shape of profile DMX is built to catch: the testing numbers alone are unremarkable, but the NFL paid premium capital and he produced at an elite, age-adjusted level. Two independent signals agreeing is what the composite rewards — a workout-warrior with no production and no draft capital can't fake his way into D1.

The dataset behind it

A model is only as good as its training history, and this is where DMX separates from one-season metrics. The database covers 3,100+ offensive prospects across 27 draft classes (2000–2026) and 2,900+ defensive prospects since 2004 — each one identity-matched across college statistics, combine results, draft records, and NFL career outcomes, then audited so the same player is the same player in every table. That curation is unglamorous and it is most of the work: public football data is riddled with duplicate identities, transferred players, and name collisions, and a scoring model built on dirty joins quietly scores the wrong careers.

The long history is also what makes the hit-rate table above meaningful. Quoting a hit rate requires waiting five NFL seasons to label each cohort, so a model launched on two classes of data has no track record by construction — ours is labeled back to 2001. The gaps are documented too: comprehensive defensive college statistics don't exist in any clean public source before 2004, so early IDP classes carry reduced-confidence scores rather than silently imputed ones.

How well does it predict?

Honestly, and in public. DMX explains roughly 10–30% of career-outcome variance depending on position and target metric. That range is strong for pre-draft data: coaching changes, scheme fit, injuries, and depth charts are invisible at draft time, and they cap what any pre-draft model can know. Draft capital alone (DPOS) runs an R² of about 5.6–13.4% standalone; the full composite improves on it at every position.

Every regression behind those claims — coefficients, R², sample sizes, cohort years — plus calibration plots by position are published on the Model Transparency page, alongside an honest list of known limits (era drift, FCS prospects, pre-2004 IDP data gaps). If you're an analyst, the regressions are reproducible from publicly readable scores. Underlying NFL data comes from sources like nflverse.

What DMX deliberately leaves out: landing spot, injury history, and character signals. It's a clean draft-day prior, meant to be combined with — not replace — your judgment about situation.

DMX vs DPX

DMX freezes at the draft; it never updates. Once a player takes NFL snaps, his dynasty value is tracked by DPX (Dynasty Performance Index) — a separate in-career model built on volume, scoring, and efficiency, updated through every NFL season. DMX answers “what should I pay on draft day?”; DPX answers “what is he worth now?”

Using DMX on the platform

• DMX Draft Board — every class since 2000, scored and ranked, with component breakdowns and career outcomes.
• Model Transparency — regressions, calibration, methodology, and known limits.
• Analytics — residual analysis, historical comps, and the Player Profiles archetypes.

Frequently asked questions