CORE
Contextual On-Field Rating Estimate
expected points per 100 qualifying snaps
Independent NFL impact analytics.
Context-adjusted player impact, cumulative value above replacement, rolling form, and a six-axis evidence profile — every number carrying its unit, sample, model identity, uncertainty, and source lineage. Real football only: not fantasy, not betting.
Every position. Every role. No fake certainty. No cowardly omissions.
▸ What is real right now
This platform is under construction, and this page will not pretend otherwise. No player ratings are published yet — the models exist as registered scaffolds and must earn promotion through result-blind evidence gates before a single public value appears.
- 87 immutable source captures — every dataset hash-pinned to an exact upstream release, with license and attribution recorded per season.
- Ten seasons validated (2016–2025): ~484,000 plays through a conservation suite — official score conservation, drive reconciliation, legal scoring deltas, participation coverage — with zero fatal and zero error findings.
- Rights discipline built in: share-alike (CC BY-SA) inputs are tracked on a segregated lineage; proprietary comparators are quarantined and can never enter public artifacts.
Validated seasons
2016–2025 · all PASS — enable JavaScript for detail, or read
freshness.json
▸ The metric family
Every metric below is registry-controlled: a canonical machine ID, a display name,
a declared unit, and a status on the claim ladder
(concept → scaffold → research → provisional → validated).
Nothing renders a value until it earns one.
FORGE
Football Overall Replacement-adjusted Game Equity
expected points, then wins when calibrated
PULSE
Performance Under Latest Sample Evidence
window-dependent rate
SIGNAL
Sample · Identifiability · Granularity · Noise · Availability · Lineage
six axis scores, 0–100
▸ The trench moonshot
Offensive line, interior defensive line, off-ball linebackers, and coverage defenders who rarely see a target are where public football analytics either goes silent or goes dishonest. Gridiron Signal treats them as flagship research problems: competing model families, identifiability scores, unit-versus-player attribution, and published disagreement between models — wider uncertainty instead of cowardly omission, honest intervals instead of fake precision.
Positions publish in evidence order: quarterbacks first, then backs, receivers, tight ends, specialists, edge rushers and targeted defensive backs — and the trenches as the boss battle, attempted seriously.