docs: grade-board sort fingerprinted (deploy boundary captured; ladder induced both directions)

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QJs13VsyiSKYQP6rj3NNmc

outputs/VYNDR-COMPLETION-MATRIX.md

@@ -350,6 +350,12 @@ grade-tier-first "top GRADES"; hero = p_win-first "top read"). They agree exactl
 leading tier (verified); across tiers the board may lead with an A the hero doesn't pick. Not a
 contradiction — do not "fix" it by making the board ignore grade.
 
-**Floor:** 310 suites / 3864 tests green, web build exit 0. Dashboard + Desk visuals are auth/feed-gated
+**Floor:** 310 suites / 3864 tests green, web build exit 0. **Post-deploy fingerprint (`b85b351`),
-→ tagged for the Chrome audit, no visual faked. **Held:** edge_pct rescale/display retirement, building
+both halves verified:** the deployed `/dashboard` chunk was polled across the deploy boundary — attempts
+1-3 carried the OLD `Math.abs(...-1/0)` key, attempt 4 flipped to the new comparator with the old one
+GONE (before/after observed, not inferred); API health 200 on snapshot/mlb, snapshot/wnba, accuracy. The
+backend ladder was **induced on demand** on live MLB game logs rather than waiting for a cron slot:
+OVER @0.5 → `0.5•(C) 1(C) 1.5(F)` (line-ASC PASS), UNDER @1.5 → `2.5(C) 2(C) 1.5•(C) 1(C) 0.5(F)`
+(line-DESC PASS). Only the *rendered* Desk card remains unverified anonymously → tagged for the Chrome
+audit, no visual faked. **Held:** edge_pct rescale/display retirement, building
 the missing `/api/props/top-graded` selector, exposing p_win to unentitled tiers.

specs/STATE.md

@@ -245,6 +245,20 @@
 > locks: disagreement never outranks agreement, missing sorts LAST and stays present, null/'' edge is
 > absent-not-zero, takeable p_win beats untakeable higher-p_win chalk, scales never mixed, the two
 > takeable bands match, grade tier still dominates, and the ladder order for BOTH directions.
+> ### ✅ POST-DEPLOY FINGERPRINT (commit `b85b351`, both halves VERIFIED)
+> **Client half — before/after captured, not inferred.** Polled the deployed `/dashboard` chunks every
+> 15s: attempts 1-3 served the OLD key (`Math.abs(...-1/0)` present, new comparator absent); **attempt
+> 4 flipped — new comparator present, OLD abs-key GONE.** Both states observed across the deploy
+> boundary. API health after deploy: `/api/snapshot/mlb` 200, `/api/snapshot/wnba` 200, `/api/accuracy` 200.
+> **Backend half — REAL LADDER PATH INDUCED ON DEMAND** (not waiting on a cron slot), `analyzeViaEngine1`
+> with live MLB game logs: **OVER** Junior Caminero hits @0.5 → grade C, p_win 0.644, ladder
+> `0.5•(C) 1(C) 1.5(F)` = **line-ASC PASS**; **UNDER** @1.5 → grade C, p_win 0.662, ladder
+> `2.5(C) 2(C) 1.5•(C) 1(C) 0.5(F)` = **line-DESC PASS**. Both directions correct on the real engine,
+> `base` rung marked in both. (The Supabase-backed side-signals logged lookup failures locally — the
+> known WSL2 `*.supabase.co` DNS blocker, BLOCKER-003 — the grade + ladder still computed from real
+> stats.) `model_snapshots.code_sha` will independently stamp `b85b351` on the next tick (latest is
+> `9b5235c`); the on-demand induction above is the verification, not that stamp.
+
 > **HELD:** edge_pct rescale/display retirement (Order B) · building the missing
 > `/api/props/top-graded` server selector (that is what would make the board render at all) ·
 > exposing p_win to unentitled tiers. **Dashboard + Desk visuals are auth/feed-gated → TAGGED FOR