Methods

Glossary

Every term, model output, or metric that appears across the platform, in plain language — the same definitions behind the ⓘ hover hints on every page.

Start here

70 terms

P(elite): Model probability this prospect becomes Pro-Bowl-level — top second-deal AAV or selected for a Pro Bowl in their career.
P(starter): Probability the player makes it past their rookie deal as a regular starter (≥70% snap share) without reaching elite.
P(role): Probability the player carves out a real but non-starter role — depth or rotational contributor.
Bust probabilityP(bust): Model probability the player busts — that his career finishes in the bottom tier for his own position. The four-tier model cuts tiers by percentile of career value within position, so a guard is judged against guards, not against receivers.
Career Score: Our own career-value index, and the ruler every accuracy claim on the site is measured against. Built from openly-licensed data: how much a player actually played (snap share, weighted so his best season counts most), how long he lasted, and what the market paid him on a second contract — the NFL's own verdict on the career. Replaced weighted Approximate Value in August 2026 so the yardstick is ours to publish.
Predicted tier: Dominant outcome the 4-tier model assigns: bust, role, starter, or elite.
Tier: Predicted career tier from the 4-tier hybrid model (bust / role / starter / elite).
NFL grade: Synthetic NFL-grade proxy. A calibrated probability (0-100%) that the player's NFL seasons looked like Pro Bowl / All-Pro play, built from public nflverse stats. Two numbers: recent-3-year weighted / career peak.
NFL grade (recent3): Weighted average of the last 3 NFL seasons (most recent 3x, prior 2x, two-back 1x). Captures current-form NFL play quality.
NFL grade (peak): Best single-season NFL grade across the player's career — captures upside.
NFL grade proxy (single season)pff_proxy_score: Model probability that a given season was Pro Bowl / All-Pro level, calibrated on actual award outcomes.
NFL grade proxy (recent three years)pff_proxy_recent3: Recent-3-year weighted proxy NFL grade per prospect (most recent 3x, prior 2x, two-back 1x).
NFL grade proxy (peak season)pff_proxy_max: Best single-season proxy score across a prospect's NFL career.
NFL grade proxy (career average)pff_proxy_mean: Average proxy score across all of a prospect's NFL seasons.
OVR AUC: One-vs-rest area-under-curve — how well the model separates this tier from the other three. 0.5 = coin-flip; 1.0 = perfect.
Macro OVR AUC: Average AUC across all four tiers, weighted equally — the headline accuracy number for the 4-tier model.
Bust AUC: How well the bust model separates eventual busts from non-busts, scored on draft classes it never saw during training — so the number can't be flattered by memorization. 0.5 = coin flip; higher = sharper.
Brier: Grades the model's probability calls against what actually happened (technically, the mean squared error of its predictions). Lower = the odds it quotes track reality more closely.
Confidence: Model's internal certainty about this prediction. Low confidence = the player's profile is unusual; the grade is less reliable.
Draftanomics Pick: Where the model thinks this prospect should go in this draft, given consensus boards + production + athletic profile.
Predicted pick: Model's predicted overall pick number in this draft.
Actual pick: Where the player was actually selected (NFL outcome).
Consensus rank: Average rank across the public draft boards we ingest (PFN, NMDDB, ESPN, NFL Network, etc.).
Scout grade: Average grade across team-internal scout reports loaded for this prospect (out of 10).
Scout Δ: Model grade minus scout grade. Positive = the model rates the prospect higher than scouts. Negative = scouts rate higher.
Scouts: Average team-internal scout grade for this prospect (out of 10), with the count of scouts who weighed in.
Consensus strengthf_cons: Consensus board strength — how high public boards rank this prospect, cohort-normalized.
Athletic compositef_athl: Athletic composite (RAS or position-specific equivalent), scored against the prospect's position group — 0 is average for the position, positive is better than average (a z-score).
Production compositef_prod: College production composite — yards, touchdowns, snap share, weighted by position and conference.
Program strengthf_school: Strength-of-program score (SP+ rating of the prospect's college).
Age at draftf_age: Age at draft, normalized — younger usually scores better at most positions.
Recruiting pedigreef_recruit: High-school recruit star rating / composite — prior pre-college signal.
Per-play impactf_ppa: Per-play impact (CollegeFootballData PPA) — efficiency on a per-snap basis.
Senior Bowl invitef_senior_bowl: Senior Bowl invitation flag (binary signal of pre-draft buzz).
Strength of schedulef_sos: Strength of schedule across the prospect's college career.
Breakout agef_breakout_age: Age at which the player first broke out productively. Younger breakouts predict NFL success.
Production trajectoryf_production_trajectory: Whether the player improved year-over-year or peaked early.
Workload sharef_workload_share: Share of team production the player carried (target share, carry share, etc.).
Athleticism × productionf_ras_x_production: Interaction term: how rare is the combination of this athleticism AND this production.
Schedule-weighted productionf_sos_weighted_production: Production weighted by strength of schedule — separates stat-padders from real producers.
Age-adjusted productionf_age_adj_production: Production adjusted for age — younger players doing the same numbers score higher.
Conference strengthf_conference_strength: Conference-level quality adjustment (Power 5 vs G5).
Position purityf_position_role_purity: How consistently the prospect played their listed position vs. moonlighting.
Board agreementf_board_consistency: How much public boards agree on this prospect (low = scout-divided, high = clean signal).
Coverage: Number of public draft boards that rank this prospect.
Board coverageconsensus_coverage_count: Number of public draft boards that rank this prospect.
Predicted AAV: Model-predicted second-deal $/yr, fit on 1,333 historical second contracts. On players the model never trained on, its typical miss is about $2.7M and 75% of predictions land within $3M (root-mean-square error $5.3M, for the statistically inclined).
Actual AAV: Real signed second-deal AAV from OverTheCap / nflverse. Only present for players who've signed their second deal.
AAV: Average annual value of a contract, in $M/yr. Standard contract size unit.
Predicted contract valuepredicted_aav_M: Model-predicted second-deal AAV ($M/yr) — what the model thinks this player will sign for on their second deal.
Board: The default Draftanomics board. Blends the model's signal with public board rankings (PFN, NMDDB, ESPN, NFLN) — most accurate for normal draft prep.
Sober: Same blend as the default Board, but caps how much public-board hype can move a prospect. Anthony Richardson 2023 falls from #13 → #41 here.
Pure model: What the model says in a vacuum. Strips out every public board and ranks on college production, athletic testing, age, and school strength only.
Consensus-aware: Renamed to 'Board' in the May 2026 fan-tier pivot. The default Draftanomics board — model signal blended with the analyst boards.
Measurables only: Renamed to 'Pure model' in the May 2026 fan-tier pivot. What the model says with no analyst board input.
Risk scoreRisk Score: Composite 0–100. Combines bust probability, injury flags, character flags, and scheme-fit risk. >=50 = material risk to flag in meetings. >=20 = worth noting. <20 = clean profile.
Feature contribution: Pre-rank score added to the prospect by this feature, in standard-deviation units. +1.0 ≈ pushes the player ~25 spots earlier in the draft order; −1.0 ≈ ~25 spots later. Magnitude over sign — small contributions ride along; only the top 3–5 are decision-relevant.
Raw scoreraw_score: Draftanomics's overall prospect score, on the same 0-to-20 scale as the Rookie score: 0 is the weakest profile we've ever graded, ~20 is an elite prospect's expected NFL production in years 1-3. Appears on /class-strength and /movers.
Rookie scoredraftiq_rookie_score: Predicted NFL years-1-3 production grade on a 0-20 scale. Built mostly from projected playing time — six parts snap share, two parts share of games started — plus a charting-grade component for positions the box score can't see, or a box-score component for positions it can.
Elite probabilityp_elite: Probability the player becomes a Pro Bowl-caliber NFL player — top second-deal AAV at his position or named to a Pro Bowl in his career. The Y-axis on the risk-reward quadrant.
STEAL: Pick verdict: the model's predicted overall pick for this player was 10+ spots LATER than where they were actually drafted. Value pick.
FAIR: Pick verdict: the actual pick was within 10 spots of the model's predicted overall pick.
REACH: Pick verdict: the actual pick was 10+ spots EARLIER than where the model said the player should have gone.
Capital: Total draft-capital remaining or spent, expressed as a percent of the team's pre-draft capital. Capital is the Jimmy Johnson chart value of all picks owned. 100% = full slate, 0% = exhausted.
Score: Position-cohort-normalized version of raw_score on the /movers page. Compares each prospect to the median prospect at their position in the same draft class.
Biggest fall: Distance between the player's pre-draft consensus rank and the pick where they ended up — when they slid by 10+ spots. Surfaces during live draft as picks come off the board.
Consensus dispersion: Standard deviation of the prospect's rank across analyst boards. High dispersion = analysts disagree, which historically correlates with bust risk (boom/bust profiles).
Snap-weighted charting grade: Player's career charting grade weighted by snaps played each season. Punishes high-grade-low-snap one-game wonders relative to a player who graded out at the same level on triple the volume.
RAS: Relative Athletic Score (Kent Lee Platte). A 0-10 score that grades an NFL Combine performance against every other prospect at that position dating back to 1987. RAS sub-scores (size / speed / explosion / agility) decompose it.
Scout synthesis: Draftanomics's AI-assembled read of every public scouting report and consensus board row for a prospect. Cites sources, surfaces analyst disagreement, and produces a verdict, case-for, and case-against. Refreshes when new sources come in.