2026 NFL DRAFT

METHODOLOGY

// Scoring rules · limitations · multi-year aggregates

· /methodology.md (same content, plain text for tools).

Summary for citation

  • Scope: Round 1, each analyst’s labeled final mock vs our actual board for that draft (pick count follows that board, usually 32).
  • Metric: 0–100 trust from weighted exact / player / position+team / proximity, averaged across n picks, scaled by 100÷3, plus optional chaos adjustment. Read trust and exacts first. S–D = fixed trust cutoffs (DEFAULT_SEASON_TIER_THRESHOLDS). Writers aggregate: same idea on mean trust via DEFAULT_CAREER_TIER_THRESHOLDS — not an NFL or outlet grade.
  • Multi-year: Stable writerId; primary aggregate needs ≥2 published years (mean trust).
  • Corpus: Ships over time — archive, /llms.txt.

What we compare

Published final Round 1 mock vs actual pick order on our board — descriptive only; no NFL, team, or outlet affiliation.

Final mock window

Last full Round 1 projection before the draft, as tagged in data. A newer file from the same outlet wins only when we replace the snapshot. The year hub / rankings view shows what we scored.

Trust index (single year)

At each board index i, mock vs actual. Exact (player + team at slot; real trade order on our board), player (right player, wrong slot), position + team, proximity (mocked player on the actual board within three indices of this slot).

Weights: exact ×3.0, player ×1.5, position+team ×1.0, proximity ×0.5; small chaos adjustment when documented. Average across n slots, multiply by 100 ÷ 3 → 0–100 trust (all-exact = 100). S–D uses fixed cutoffs in DEFAULT_SEASON_TIER_THRESHOLDS (S is rare). Optional sorts reorder rows; tier follows that season’s trust.

Edge cases

Trades, slides, and reaches move exacts — that is why softer match types exist. Mocked in Round 1 but not taken in Round 1 → miss for this scope.

Writers / multi-year

Stable writerId. Aggregate = mean trust across years (median shown); primary table needs ≥2 published years. Career tier uses DEFAULT_CAREER_TIER_THRESHOLDS on that mean; the single-year block uses the same thresholds on one year’s mean (matches the year leaderboard). Ties: more seasons → higher latest-year trust → writerId.

Limitations

  • Round 1 only unless a future season documents otherwise in data.
  • “Final mock” follows our labels; outlets differ.
  • Trust, exacts, and S–D are our rubric — not universal or official grades.
  • Corpus grows over time — see archive and /llms.txt.
Reproducible

Scoring in TypeScript with tests — rubric changes do not silently reshuffle ranks.

Stable URLs

Per-year routes stay linkable for fans, search, and citation.

Honest scope

Thin multi-year data is labeled plainly on aggregate pages.

Independent analysis. Trust, exact counts, and S–D tiers are MockGrader metrics only (not league or outlet grades).