Ledger takeable tagging (deferred C2); efficiency challenger BLOCKED
Champion grade UNCHANGED. Push scoring untouched. Additive tags only — nothing
deleted, nothing re-settled.
PART A — THE EFFICIENCY CHALLENGER: BLOCKED, NOT BUILT.
Review Zero came back ABSENT on all three inputs:
0.1 efficiency scores DO NOT EXIST (zero occurrences of market_efficiency /
marketEfficiency / efficiency_score in src/ or web/src/).
0.2 base thresholds DO NOT EXIST (engine1.js has zero `edge` references — the
grade is not an edge-vs-threshold comparison; grade_thresholds.json holds
PROBABILITY bands).
0.3 the +/-0.05 additive efficiency nudge DOES NOT EXIST. The only 0.05s on
the grade path are featureCache.teammate_absence_bump, a bvp_advantage
cutoff, and p*0.9+0.05 inside probabilityEstimator (the 0.5*0.1 term of
the shrink-toward-0.5). There is no additive scaling to replace.
So a challenger differing from the champion in EXACTLY ONE thing cannot be
constructed: there is no additive scaling to swap, no base threshold to
multiply, and engine1.js has zero `sport` references so market cannot reach the
grade. A threshold must exist first — that is R1 of
specs/full-output-grade-mapping.md, an explicitly held separate order. Shipping
R1+R4 together would make the Phase-3 delta report misleading: the re-letter
would be driven mostly by switching to probability grading while being
presented as the efficiency fix.
0.4 coverage: the spec names 5 scores; the live ledger has 11 markets and only
MLB total_bases maps to one. 9 of 11 have no score, so "all scored markets"
cannot be satisfied without inventing 9 numbers.
PART B — LEDGER TAKEABLE TAGGING: BUILT (the deferred C2).
New src/config/takeableStandard.js: floor on the minus side, UNCAPPED plus.
Deliberately NOT valueEngine.isTakeable (the -160..+200 PROMOTION band) — a
+400 prop is not promotable but IS takeable; a test asserts the two diverge on
the plus side and agree at the floor so they can never quietly merge. Absent
price returns null, never false (Number(null) === 0 would tag a missing price
takeable). The floor is POLICY not derived (C1 could not derive one) and is
labelled so; each row records takeable_floor so a re-derivation can re-tag.
Migration 034 (applied + tracked): ledger_entries.takeable boolean +
takeable_floor numeric, nullable, partial index. Forward tagging in
ledgerService at row build; backfill in one statement.
Result: 1254 rows, 1246 tagged (781 takeable / 465 below floor), 8 NULL with
null_despite_price = 0 (the NULLs are genuinely priceless rows). Settled 1163
and graded 1254 unchanged.
PART C — the model-version boundary tag is DELIBERATELY NOT APPLIED: no scaling
change shipped, so no boundary exists, and stamping one would mark a model
transition that never happened. modelEras.js is its home when a real one lands.
Floor: 312 suites / 3890 tests green (8 new), web build exit 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QJs13VsyiSKYQP6rj3NNmc
This commit is contained in:
@@ -0,0 +1,122 @@
|
||||
# SPEC — LEDGER TAKEABLE TAGGING (built) + MULTIPLICATIVE EFFICIENCY SCALING (BLOCKED)
|
||||
2026-07-31. Champion grade UNCHANGED. Push scoring untouched.
|
||||
|
||||
---
|
||||
|
||||
## PART A — THE EFFICIENCY CHALLENGER: **BLOCKED, NOT BUILT**
|
||||
|
||||
### Review Zero came back ABSENT on all three inputs
|
||||
|
||||
- **0.1 efficiency scores — DO NOT EXIST.** Zero occurrences of `market_efficiency`,
|
||||
`marketEfficiency`, or `efficiency_score` in `src/` or `web/src/`. `src/config/` holds
|
||||
no efficiency file (the one `0.80` there is `providers.js WARN_PCT`, a quota warning).
|
||||
- **0.2 base thresholds — DO NOT EXIST.** There is no "MLB A: 5% edge / NBA A: 7% edge"
|
||||
table. `engine1.js` contains **zero `edge` references**; the grade is not an
|
||||
edge-vs-threshold comparison. `grade_thresholds.json` holds PROBABILITY bands.
|
||||
- **0.3 the ±0.05 additive nudge — DOES NOT EXIST.** Searched the whole grade path. The
|
||||
only `0.05` values are `featureCache.teammate_absence_bump` (a feature), a
|
||||
`bvp_advantage` cutoff for a matchup letter, and `p*0.9 + 0.05` inside
|
||||
`probabilityEstimator` (the `0.5 × 0.1` term of the shrink-toward-0.5, not efficiency).
|
||||
**There is no additive efficiency application to replace.** No market-conditional logic
|
||||
exists in `engine1.js` or `analyzeViaEngine1.js` at all.
|
||||
|
||||
### Why the challenger cannot be built as specified
|
||||
|
||||
The order requires a challenger differing from the champion in **EXACTLY ONE thing —
|
||||
multiplicative vs additive scaling**. That construction is impossible here:
|
||||
|
||||
- there is no additive scaling to swap,
|
||||
- there is no base threshold to multiply,
|
||||
- and `engine1.js` has **zero `sport` references**, so market/sport cannot reach the grade.
|
||||
|
||||
To have a threshold to scale, the challenger must FIRST grade on probability-vs-threshold —
|
||||
which is rule **R1 of `specs/full-output-grade-mapping.md`**, an explicitly held separate
|
||||
order. Shipping R1+R4 together would (a) violate the isolate-one-variable requirement and
|
||||
(b) make the Phase-3 delta report **misleading**: the re-letter would be driven mostly by
|
||||
switching to probability grading, presented as the efficiency fix. Kev would be approving
|
||||
the wrong attribution. **That is why this half stopped rather than guessed.**
|
||||
|
||||
### 0.4 — coverage gap that would sink the build anyway
|
||||
|
||||
The spec names 5 scores (NFL passing .85, NBA points .80, MLB K .60, MLB TB .55, NBA role
|
||||
.55). The live public ledger contains **11 markets**: MLB hits (376 rows), doubles (247),
|
||||
total_bases (44), stolen_bases (17), earned_runs (6), home_runs (4), outs (2); WNBA points
|
||||
(246), rebounds (156), assists (142), threes (10). **Only MLB total_bases maps to a specced
|
||||
score.** MLB K's and NFL passing do not appear at all; WNBA is not in the spec list. So
|
||||
**9 of 11 live markets have NO score** — and the guardrail ("all scored markets, or it's
|
||||
incoherent") cannot be satisfied without inventing 9 numbers. Inventing them would be
|
||||
fabricated calibration.
|
||||
|
||||
**What Part A needs before it can run:** (1) R1 probability grading, so a threshold exists;
|
||||
(2) efficiency scores for the markets we actually grade — ideally **FIT from each sport's
|
||||
own accrued record**, not hand-set (that is R4's stated design).
|
||||
|
||||
---
|
||||
|
||||
## PART B — LEDGER TAKEABLE TAGGING: **BUILT** (the deferred C2)
|
||||
|
||||
Independent of Part A: it needs no threshold, no scaling, no grade change.
|
||||
|
||||
### The standard (`src/config/takeableStandard.js`)
|
||||
|
||||
isLedgerTakeable(american) = american >= LEDGER_TAKEABLE_FLOOR // default -160
|
||||
// FLOOR on the minus side, UNCAPPED on the plus side.
|
||||
|
||||
**🔴 It is deliberately NOT `valueEngine.isTakeable`.** Those answer different questions:
|
||||
|
||||
| | shape | question |
|
||||
|---|---|---|
|
||||
| `valueEngine.isTakeable` | −160..+200, **bounded both ends** | "may we PROMOTE this to the hero / top of board?" |
|
||||
| `takeableStandard.isLedgerTakeable` | ≥ −160, **uncapped plus** | "was this a price a bettor could actually have TAKEN?" |
|
||||
|
||||
A **+400** prop is **NOT promotable but IS takeable** — both true at once. A unit test asserts
|
||||
they diverge on the plus side and agree at the minus floor, so the two can never quietly merge.
|
||||
|
||||
**Absent price → `null`, never `false`.** Without the strict guard `Number(null) === 0` is
|
||||
≥ −160 and a MISSING price would be tagged takeable — fabricated data in the record.
|
||||
|
||||
**The floor is POLICY, not derived, and is labelled so everywhere.** C1 (2026-07-30) could not
|
||||
derive one: every price bucket's 95% ROI interval contained zero, the curve was non-monotone
|
||||
(deepest buckets positive), and resolving a 5-point edge needs 661–2,285 rows/bucket against
|
||||
actual sizes of 8–71. −160 is held because it is directionally the better half of the only
|
||||
available comparison (in-band ROI +3.84% vs out-of-band −0.04%, both spanning zero).
|
||||
**Re-derivation trigger:** any negative bucket reaching n ≥ 300, or the end of the MLB regular
|
||||
season — and adopt a derived floor only when a bucket's 95% CI excludes zero.
|
||||
|
||||
### Schema (migration 034, applied + tracked)
|
||||
`ledger_entries.takeable boolean` + `takeable_floor numeric`, both nullable, plus a partial
|
||||
index on `(sport, takeable) WHERE user_id IS NULL`. **`takeable_floor` records the floor each
|
||||
row was judged under**, so a future re-derivation can re-tag safely instead of guessing which
|
||||
standard a row was tagged by. Additive only — nothing deleted, nothing re-settled.
|
||||
|
||||
### Forward + backfill
|
||||
Forward: `ledgerService` stamps `takeable`/`takeable_floor` at row build from the locked odds.
|
||||
Backfill: every row with a parseable locked price, one statement, no outcome/grade/price touched.
|
||||
|
||||
**Result — 1,254 rows: 1,246 tagged (781 takeable / 465 below floor), 8 NULL, and
|
||||
`null_despite_price = 0`** — i.e. every priced row is tagged and the 8 NULLs are genuinely
|
||||
priceless rows (honest absence). One distinct floor (−160). Settled 1,163 and graded 1,254
|
||||
both unchanged.
|
||||
|
||||
---
|
||||
|
||||
## PART C — MODEL-VERSION BOUNDARY TAG: **DELIBERATELY NOT APPLIED**
|
||||
|
||||
Phase 2.4 asks for a boundary separating pre-fix (additive) from post-fix (multiplicative)
|
||||
grades. **No scaling change shipped, so no such boundary exists.** Stamping one would mark a
|
||||
model transition that never happened — a fabricated era in the permanent record, which is the
|
||||
exact class of thing the Data Semantics Rule forbids. `src/config/modelEras.js` (Session 64)
|
||||
is the correct home for it **on the day a real scaling change promotes**.
|
||||
|
||||
---
|
||||
|
||||
## VERIFY
|
||||
Champion grade path untouched (`engine1`, `analyzeViaEngine1`, `probabilityEstimator`,
|
||||
`gradeSlateService`, `grade_thresholds.json` — no diff). **312 suites / 3,890 tests green**
|
||||
(8 new), **web build exit 0**. Grade visuals are auth-gated → Chrome audit; this order changed
|
||||
nothing visual.
|
||||
|
||||
## TAGS
|
||||
VERIFIED: efficiency scores/base thresholds/±0.05 nudge all absent; 11-market coverage gap;
|
||||
tagging built + backfilled + integrity-checked. CANNOT DETERMINE: none. **BLOCKED: Part A
|
||||
(the efficiency challenger) — needs R1 probability grading plus fitted per-market scores.**
|
||||
Reference in New Issue
Block a user