Files

T

builtbykev 0997334f8b Order B: retire edge_pct display. Promotion gate NOT passed — no flip.

THE PROMOTION WAS NOT PERFORMED. Champion grade path byte-identical (diff empty
across intelligence/, gradeSlateService, snapshotService). Projection, p_win and
the CLV instrument untouched.

REVIEW ZERO IS A GATE AND THREE OF FOUR PREREQUISITES FAIL:
  0.1 scores are ESTIMATED priors from the founding spec, not measured. The
      premise's cited values are not in the code either — the module holds
      nba:points .80 and mlb:total_bases .55; there is no NBA 0.72 and no WNBA
      score at all.
  0.2 VERSION-BOUNDARY TAG DID NOT LAND — config/modelEras.js has zero shading
      references. It was deliberately not applied twice (nothing had been
      promoted) and reported both times. The order's own rule says STOP.
  0.3 NO ROLLBACK FLAG EXISTS — zero occurrences of SHADING_ENABLED /
      EDGE_SHADING / shadingEnabled anywhere in src/.
  0.4 takeable tags DID land (migration 034, 1246/1254 rows). PASS.

AND THE APPROVED DELTA DOES NOT MATCH THE MEASURED ONE. Approved: 43.6% of
grades re-letter, efficient markets tighten and soft hold. Measured on all 1250
live rows: 97.4% change (1217), 79.8% move UP, 17.6% down, resulting in 79.0%
A-family (MLB 93.4%) against the champion's 0.2%. And rows_actually_shaded = 0
of 1250 — 96.5% of markets are unscored (f=1) and the one scored market present
is the anchor (f=1.0 by construction). The entire re-letter comes from switching
to edge-vs-fixed-bar grading, NOT from efficiency shading, which is inert on
this board. That is an unapproved grading-basis change riding along, which the
order's own "no new scaling changes riding along" guardrail forbids.

Flipping would re-letter 97.4% of an append-only public record, move 79.8% of
grades UP and mint A's on 79% of the board, on a letter whose measured
correlation with outcomes is r ~ 0.005 — the exact scenario the permanent
founder ruling forbids.

SHIPPED — ORDER B (independent of the promotion, and a live falsehood):
edge_pct display retired from GradeResultCard (confidence strip, EDGE stat cell
now honest-absent, alt-ladder rung) and SoccerGradeResult. DeskShowcase kept
(already honest). Computation and the board's signed-edge sort fallback SURVIVE
— deleting them would re-break the sort fixed on 2026-07-29; a test asserts all
three survive and the sort still orders agrees -> disagrees -> absent.

Fixed two build-breakers the retirement caused (orphaned edgeColor import,
orphaned edge_pct destructure; edge_pct stays on the props contract). Two
pre-existing tests superseded rather than deleted: they asserted the edge figure
is sign-coloured, and now assert the stronger property that no edge percentage
renders at all.

Floor: 314 suites / 3908 tests green (9 new), web build exit 0.

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

2026-07-31 00:57:57 -04:00

5.3 KiB

Raw Blame History

SPEC — ORDER B: edge_pct DISPLAY RETIREMENT (shipped) + PROMOTION GATE (NOT PASSED)

2026-07-31. The champion grade was NOT flipped. Projection, p_win, and the CLV instrument untouched. Push scoring untouched.

PART 1 — THE PROMOTION GATE: THREE OF FOUR PREREQUISITES FAIL. NO FLIP.

The order states Review Zero is a gate ("all must pass before flip") and that 0.2 failing means STOP. Checked honestly, against code, not memory:

gate	result
0.1 scores measured or estimated?	ESTIMATED priors, transcribed from the founding spec. And the premise's cited values are not in the code: the module holds `nfl:passing_yards .85 · nba:points .80 · mlb:strikeouts .60 · mlb:total_bases .55 · nba:role_player .55`. There is no NBA 0.72 and no WNBA 0.68 — no WNBA score exists at all.
0.2 version-boundary tag landed?	❌ NO. `config/modelEras.js` contains zero references to shading. It was deliberately not applied (twice), because nothing had been promoted, and that was reported both times. The order's own rule: STOP.
0.3 rollback flag exists?	❌ NO. Zero occurrences of `SHADING_ENABLED` / `EDGE_SHADING` / `shadingEnabled` anywhere in `src/`. There is no flag to revert with.
0.4 takeable tags landed?	✅ YES — migration 034, 1,246 of 1,254 rows tagged, floor recorded per row.

The approved delta does not match the measured delta

The premise approves "re-letters 43.6% of grades in the correct direction (efficient markets tighten, soft hold)." Measured on all 1,250 live ledger rows:

measure	approved	actual
grades changed	43.6%	97.4% (1,217 of 1,250)
direction	efficient tighten / soft hold	79.8% move UP, 17.6% down
resulting A-family	—	79.0% A (MLB 93.4%) vs champion 0.2%
rows actually shaded by efficiency	—	0 of 1,250

The entire re-letter comes from switching to edge-vs-fixed-bar grading — not from efficiency shading, which touches zero live rows (96.5% of markets are unscored → f=1, and the one scored market present, mlb:total_bases, is the anchor where f=1.0 by construction). So the mechanism being promoted is inert, while the thing that would actually change the board is an unapproved grading-basis switch riding along — which the order's own "no new scaling changes riding along" guardrail forbids.

Why this is a refusal to flip, not a delay

Promoting would re-letter 97.4% of an append-only public record, moving 79.8% of grades UP and minting A's on 79% of the board — on a letter whose measured correlation with outcomes is r ≈ 0.005. That is the precise scenario CLAUDE.md's permanent founder ruling forbids: "Minting A's without new information is a relabelled B sold as an A and it corrupts an append-only ledger." No new information is present here — 0 rows are shaded.

To unblock: land the version-boundary tag, add a rollback flag, replace the input (p_win vs fair_prob — a real price-aware edge) so the fixed bar is meaningful, and fit scores for markets we actually grade. Then re-measure and re-approve against the true delta.

PART 2 — ORDER B: edge_pct DISPLAY RETIRED (SHIPPED)

Independent of the promotion, and a live falsehood, so it shipped in full.

Retired (display only):

GradeResultCard confidence strip — the +X% edge figure.
GradeResultCard stat row — EDGE is now honest-absent —.
GradeResultCard alt-ladder rung — the same price-free number; the rung's line and grade (real model output) still render.
SoccerGradeResult — the · +X.X% edge clause.

Deliberately KEPT: DeskShowcase (already honest — nulls anything past the sane cap).

Computation and sort survive, by design. Backend edgePctFor, web computeEdge, and the board's signed-edge sort fallback (descNullsLast(a.edge, b.edge)) are all intact — deleting them would re-break the grade-board sort fixed on 2026-07-29. A test asserts all three survive and that the sort still orders agrees → disagrees → absent.

Two build-breakers found and fixed: retiring the display orphaned the edgeColor import in GradeResultCard and the edge_pct destructure in SoccerGradeResult. edge_pct stays on the props contract (callers still pass it, the backend still computes it) — only the destructure for display was dropped.

Two pre-existing tests were SUPERSEDED, not deleted (ds4Billboards, colorContract): they asserted the edge figure is coloured by sign, which was correct while a figure was shown. They now assert the stronger property — no edge percentage renders at all. edgeColor itself is untouched and still enforced for its other consumers (MarketBreadth).

VERIFY

Champion grade path byte-identical (diff empty across intelligence/, gradeSlateService, snapshotService) — no flip performed. Projection, p_win, CLV instrument untouched. 314 suites / 3,908 tests green (9 new), web build exit 0.

5.3 KiB Raw Blame History Unescape Escape