Skip to content
MBStats

The Ledger

Sealed predictions, graded in public

Every claim below was frozen at commit time and is never edited after sealing. Each carries a grading date fixed in advance, and misses publish alongside hits.

Sealed claims
6
Pending
6
Due for grading
0
Graded
0

Sealed 2026-07

Frozen at commit time; never edited after sealing. Each claim carries its grading date and metric. Grades append to a separate results file; misses publish alongside hits.

ClaimSealedGrades afterStatusResult

SCV 2026 Retention

For the 2026 signings below, two-season retention of platform-year portable production follows the SCV v2/v3 forecasts: the below-80 rate lands within 10pp of each anchorRate cell, and the mean per-player forecast error (expectedRetention vs realized) beats the no-model baseline of predicting 1.0 for everyone.

below-80 rate by anchor cell vs anchorRate; MAE(expectedRetention, realized retention) vs MAE(1.0, realized)

63 signings sealed

2026-07-252028-02-15PENDINGawaiting data

Kicker Mean Reversion

Kicker FG% over expected is ~noise (measured yoy r = .034). 2026 FGOE for the kickers below will regress essentially to zero: predicted 2026 FGOE = 0.034 x 2025 FGOE for every kicker. The 2025 top-quintile group will NOT outkick the 2025 bottom-quintile group by more than 1.5pp of FGOE in 2026.

corr(fgoe_2025, fgoe_2026); group gap top-vs-bottom quintile; MAE of the 0.034-slope forecast vs a persistence (slope 1.0) forecast

30 kickers sealed

2026-07-252027-02-15PENDINGawaiting data

Age Cliff 2026

Among the claim-1 players, the age-cliff-flagged group goes below 80% retention at a rate at least 10pp higher than unflagged players on the same rung (incumbent or external).

below-80 rate, flagged vs unflagged, within rung

2026-07-252028-02-15PENDINGawaiting data

R1 Need Reach 2026

2026 first-round picks that landed in the drafting team top-3 need groups (team_need_season 2026) will trail 2026 R1 BPA picks in three-year-starter rate by at least 5pp, despite inheriting more snaps (measured historical gap -12.9pp with controls).

seasons_started >= 3 by 2029, need-top3 vs rest of R1, from draft_picks

2026-07-252029-06-01PENDINGawaiting data

Scheme Sorting 2027

External CB/S/LB signings in the 2027 offseason will correlate at r >= .10 (within-season z-grain) between the player prior man exposure and the signing team prior man rate. Sourcing stays scheme-sorted; it will NOT anticipate 2027 scheme changes better than it matches 2026 schemes.

corr(player man_share_z t-1, signer man_z t-1) on 2027 externals 1%+

2026-07-252027-04-01PENDINGawaiting data

Punter Over Kicker 2026

Punter EPA-per-punt persistence (2025 to 2026, 30+ punts) exceeds kicker FGOE persistence (20+ kicks) by at least .10 of correlation.

r_yoy(punter epa) - r_yoy(kicker fgoe)

2026-07-252027-02-15PENDINGawaiting data

Several claims price the models behind the free-agency class. The ledger files live in the repo under packages/stats/predictions, so the git history is the timestamp.