Insights

Receipts first. Every number traces to a graded row.

The same receipts query feeds Picks, Insights and the Sunday digest — it can never disagree with itself.

Analytics · data harness

The live freshness of every feed behind the house's record.

sources become Frame bricks with lineage; lenses turn them into derived frames; surfaces and embeds only ever show the remix

sources feeds/APIs/public endpoints/tape; identified UA, robots, rate limits — NEVER impersonation
frames typed, watermarked, source_id per row; parquet lake system-of-record for ≥100k
derived consensus, dispersion, velocity
surfaces Tonight/Atlas/Blog/embeds
freshness C-18 observations + per-feed gates C-33; liveness = rows produced

Feed gates unavailable right now.

Anon

Receipts + Leaderboard public (that's the marketing); Explore = 8 house views

this viewer: signed out

Trial · paid

Explore unlimited + saved views + drill + share cards; Me

not this viewer: signed out — the tier is not in this page's payload, so no tier is claimed

Cold query

skeleton over the graded population, with the cached previous result under it

computing over 272,826 graded plays · as of Aug 28, 4:40 AM ET

Small n

cells under 50 plays grey with the n; never hidden

1 of 5 baked widget rows carry n under 50— clv-distribution n=0

Week not graded

games still pending — numbers update at final

50 upcoming games not final in the 50-game window this page reads· last graded Aug 28, 4:40 AM ET

Disagreement guard

if Receipts ≠ Picks → Record for the same filter, the page shows the diff and files an issue

checked this render: aggregate and per-market sum agree · last graded Aug 28, 4:40 AM ET