Entity Atlas -- Per-Entity Analytics Coverage at Scale, Every Count Auditable

The atlas family renders one small analytics card per entity, at scale, across every sport this system holds -- and every card is traceable to a manifest count you can re-derive. Six builder modules wrote seven manifests holding 1,549 cards total: 1,523 per-entity cards across six sport/entity-type manifests, plus 26 calibration-checkpoint cards built on the same machinery. Each count below is the manifest's own n_entries field, reported verbatim, and each was cross-checked against the rendered PNGs on disk -- all seven match exactly.

The only superlative on this page is the one the whole design earns: not the best entity analytics in the world -- the most auditable. Every card names its inclusion floor on its own face, every count is a JSON field you can open, and the whole family is served through the same fail-closed answer engine as every other resolver, so an entity we never built returns an honest no_data, never a guess.

### What it is (and is not)

Each card is a 1-row strip of 2-4 mini-panels rendered from recorded data -- box scores, Statcast pitch logs, results history -- reshaped into a compact, provenance-stamped picture. Every card carries a red `DESCRIPTIVE_ONLY` badge and a source | floor | as_of footer stamped by the factory; every manifest header sets descriptive_only: true; the resolver returns edge_claimed: false on every hit. There is no forecast on these cards. They show what an entity has done, gated by a declared minimum-sample floor, and nothing more. Edge- or ROI-shaped questions are refused upstream by the same answer engine, by design.

Breadth in an analytics system is usually asserted with a round headline number. This family proves it instead: 1,549 cards across seven manifests, each count a JSON field you can open and a PNG you can count, cross-checked to match on disk, every card carrying its own inclusion floor and a DESCRIPTIVE_ONLY badge on its face. "At scale" here does not mean a big number to impress -- it means the coverage is wide and every unit of it is individually auditable, down to the entity, the floor that let it in, and the manifest line that records it. The transferable thing is not the count. It is that a system can render analytics for 1,523 entities and still refuse -- no_data, "refusing, not guessing" -- the moment you ask about one it never built. The truth-source for every figure is docs/JOB_EVIDENCE_PACKET.md.