VANTAGE
← Methodology

Who makes this

One person's judgment, checked by a machine gate. Everything below is on file elsewhere on the site — this page just says who and how.

a badge and a name — labels are never paywalled, these people just helped

Everything on this site — the labels, the receipts, the refusals — stays free to read. The names below chipped in anyway. That's the whole deal: no paywall bought, none sold.

The wall starts with the first name.

Names are added on purchase and come off automatically on refund — the wall is read straight from the order ledger, like every other number here.Chip in →

one set of human eyes

Every verified number passed through one set of human eyes: Juyoung Kim. Measured against seeded gold: two-touch confirmed 10/10 on the holloway sweep — measured undercount ~8.8% (global estimate 11–15%). Single-annotator truth is a limit — it's disclosed on every page it touches.

byte-identical regression + SHA-pinned packs

Every pack is rebuilt behind a byte-identical regression gate: Topuria's frozen sheet is replayed and asserted equal to the last accepted build, and the canonical event stream is checked byte-for-byte against its reference — any drift blocks the build (PASS). Each pack is SHA-pinned (pack-v3, engine hash in the manifest), so the numbers you read trace to an exact, reproducible compile — see the methodology for the full convention set.

and the measured reason why
  • comparisons — v1 limitation — single-fighter queries only
  • predictions matchmaking — out of scope by design — we report measured tendencies, never outcomes
  • clinch ground — capture-blocked — corpus is standup-scoped (single broadcast feed)
  • landed machine claims — landed/missed carries ~no machine signal at current capture; human labels only
  • fine subtype beyond vocab — vocab v1.1 boundary (e.g. no spinning/switch variants)