Item 7 — public accuracy reads the CLEAN ledger; BEAT CLOSE hidden until C4

Kev's call: the 30D accuracy surfaces must read TRUTH, not a cache that can't be
filtered. My earlier degraded-row exclusion only touched getModelAggregate
(Postgres); the public buckets/badge still read outcomeService (Redis outcome
log), which counts degraded projection-0 outcomes and has no field to filter on.

- /api/accuracy (AccuracyBadge) + /api/ledger/accuracy (buckets/ModelRecord)
  now source from the clean Postgres ledger aggregate via new
  ledgerService.getAccuracyView + accuracyBucketsFromAgg (model_value > 0
  excludes degraded rows). Same response shapes → no frontend change. Redis
  outcome log is now read by nothing public; it can age out or be rebuilt.

- BEAT CLOSE is a MEASURED-WRONG ZERO: captureClosing re-records the locked line
  as the "closing" line, so clv is flat on the whole sample and beat_close reads
  0% (comparing a number to itself). Full write-up: specs/audit-data/
  clv-capture-broken.md (the fix belongs to C4). Until then, beat_close_pct +
  clv_distribution are SUPPRESSED at the source (getModelAggregate, gated by
  clvCaptureReliable() / CLV_CAPTURE_RELIABLE=1). Every public surface already
  renders BEAT CLOSE only when non-null, so they all hide it now — no wrong zero
  anywhere. HIT RATE (real) is unaffected.

Suite 271/3261 green, web build exit 0.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

specs/audit-data/clv-capture-broken.md

@@ -0,0 +1,41 @@
+# CLV Capture Broken — C4 Finding (2026-07-17)
+
+Source: Truth-Everywhere Part 2, item 7. Escalated from the "0% BEAT CLOSE"
+observation on the public accuracy surfaces.
+
+## The bug
+Every settled ledger row has `closing_line == locked_line` (and
+`closing_odds == locked_odds`), so the computed `clv` is 0/flat on the ENTIRE
+sample. `beat_close_pct` therefore reads a fabricated-looking **0%** — it is
+comparing a number to itself, not measuring closing-line value.
+
+Evidence (live `/api/ledger/model`, 2026-07-17): sampled rows show
+`line == closing_line`, `clv: null`/0, `clv_result: null`/flat across the board.
+
+## Root cause (to fix in C4)
+`ledgerService.captureClosing` overwrites today's unsettled rows'
+`closing_line`/`closing_odds` on every snapshot with the CURRENT feed values.
+The intent (S58) was "the last write before game start is the close." But in
+practice the captured value equals the locked line every time — either the
+lines genuinely don't move in the captured window, or captureClosing is reading
+the same feed field the lock came from and writing it back unchanged. Net: the
+"closing" column is a copy of the lock, so CLV is structurally always 0.
+
+## The fix (C4 — not done here)
+Capture the REAL last line before game start into a SEPARATE field, never
+defaulted to the lock:
+- Record `closing_line` only from a distinct closing snapshot (the last feed
+  read before first pitch/tip), and only when it actually differs from the
+  lock — otherwise leave it null (absent, not a copy).
+- CLV = signed(locked - closing) by side, computed only when a real, distinct
+  closing line exists. No close captured ⇒ CLV null for that row (honest), not 0.
+- Verify against a few known line moves before trusting the aggregate.
+
+## Interim (shipped this pass, item 7)
+`beat_close_pct` and `clv_distribution` are SUPPRESSED at the source
+(`getModelAggregate`, gated by `clvCaptureReliable()` /
+`CLV_CAPTURE_RELIABLE=1`). Every public surface (ModelRecord, ledger MODEL tab,
+public profiles, OG images) already renders BEAT CLOSE only when non-null, so
+they all hide it now — no measured-wrong zero on any public surface. HIT RATE
+(which is real) is unaffected. Flip `CLV_CAPTURE_RELIABLE=1` once C4 lands and
+real closes are verified.

src/routes/accuracy.js

@@ -3,27 +3,31 @@
 /**
  * GET /api/accuracy (Session 55) — the system's track record.
  *
- * Public, cache-only read of the rolling accuracy record written by
+ * Public, cache-only read of the model's track record. Powers the AccuracyBadge
- * outcomeService (settled snapshot grades vs real results). Powers the
+ * "A-RATED · X% HIT · 30D" pill and the grade-card accuracy line.
- * dashboard "A-rated: 68% hit rate" pill and the grade-card accuracy line.
+ *
- * NEVER triggers settlement (that's the internal cron) → no API credits spent.
+ * Truth-Everywhere Part 2 (item 7) — sourced from the CLEAN Postgres ledger
+ * aggregate (getAccuracyView: model_value > 0 excludes the degraded
+ * projection-0 rows), NOT the Redis outcome log (which still counts them and
+ * can't be filtered). Redis is a cache; when a cache can't be filtered, read
+ * from truth. NEVER triggers settlement → no API credits spent.
  */
 
 const express = require('express');
 const { createRateLimit } = require('../middleware/rateLimit');
-const outcomeService = require('../services/outcomeService');
+const ledgerService = require('../services/ledgerService');
 
 const router = express.Router();
 router.use(createRateLimit({ windowMs: 60_000, max: 60 }));
 
 router.get('/', async (req, res) => {
   try {
-    const acc = await outcomeService.getAccuracy();
+    const acc = await ledgerService.getAccuracyView({});
     res.set('Cache-Control', 'public, max-age=300');
     return res.json(acc);
   } catch (err) {
     console.error('[accuracy]', err.message);
-    return res.status(200).json({ overall: null, sports: {}, min_sample: outcomeService.MIN_SAMPLE, updated_at: null });
+    return res.status(200).json({ overall: null, sports: {}, min_sample: 20, updated_at: null });
   }
 });
 

src/routes/ledger.js

@@ -41,11 +41,18 @@ function applyFilters(query, req) {
 }
 
 router.get('/accuracy', async (req, res) => {
+  // Truth-Everywhere Part 2 (item 7) — read the CLEAN Postgres ledger aggregate
+  // (model_value > 0 excludes degraded rows), NOT the Redis outcome log (which
+  // still counts degraded projection-0 outcomes and can't be filtered).
   try {
-    const acc = await outcomeService.getAccuracy();
+    const agg = await ledgerService.getModelAggregate({});
-    const buckets = outcomeService.accuracyBuckets(acc.overall);
+    const buckets = ledgerService.accuracyBucketsFromAgg(agg);
+    const overall = {
+      hits: agg.hits, misses: agg.misses, pushes: agg.pushes,
+      total: agg.hits + agg.misses + agg.pushes, pct: agg.hit_pct ?? null,
+    };
     res.set('Cache-Control', 'public, max-age=300');
-    return res.json({ buckets, overall: acc.overall && acc.overall.overall, updated_at: acc.updated_at });
+    return res.json({ buckets, overall, updated_at: null });
   } catch (err) {
     console.error('[ledger/accuracy]', err.message);
     return res.status(200).json({ buckets: [] });

src/services/ledgerService.js

@@ -37,6 +37,12 @@ const AGG_WINDOW_DAYS = 30;
 const AGG_FETCH_LIMIT = 5000;
 /** Below this many settled rows, callers must not render a percentage. */
 const MIN_AGG_SAMPLE = 20;
+// Truth-Everywhere Part 2 (item 7) — CLV capture is broken (closing_line ==
+// locked_line; see the C4 finding). Until C4 records a real closing line,
+// beat_close/CLV are suppressed everywhere. Read at call time (not module load)
+// so C4 can flip it via CLV_CAPTURE_RELIABLE=1 without a redeploy, and tests can
+// exercise the CLV math directly.
+function clvCaptureReliable() { return process.env.CLV_CAPTURE_RELIABLE === '1'; }
 
 /**
  * S6 (A1 board) — CLV distribution buckets (the MODEL tab strip). Signed CLV:
@@ -515,14 +521,22 @@ async function getModelAggregate(opts = {}) {
   if (agg.settled >= MIN_AGG_SAMPLE && decided > 0) {
     agg.hit_pct = Math.round((agg.hits / decided) * 100);
   }
-  if (agg.settled >= MIN_AGG_SAMPLE && agg.clv_sample > 0) {
+  // Truth-Everywhere Part 2 (item 7) — CLV is currently MEASURED WRONG:
+  // captureClosing re-records the LOCKED line as the "closing" line
+  // (closing_line == locked_line across the whole sample), so every row's CLV
+  // computes to 0/flat and beat_close reads a fabricated-looking 0%. That's
+  // comparing a number to itself. Until C4 (real closing-line capture) lands,
+  // CLV_CAPTURE_RELIABLE stays false and beat_close_pct / clv_distribution are
+  // suppressed at the SOURCE — every public surface hides BEAT CLOSE rather
+  // than showing a measured-wrong zero. Flip this to true when C4 ships.
+  if (clvCaptureReliable() && agg.settled >= MIN_AGG_SAMPLE && agg.clv_sample > 0) {
     agg.beat_close_pct = Math.round((agg.clv_beat / agg.clv_sample) * 100);
   }
   // S6 (A1 board) — clv_distribution rides the SAME n≥20 gate (this is the
   // single home of the gate — consumers never re-derive it). Null below the
   // sample floor or with zero settled clv values; the UI renders nothing.
   agg.clv_distribution = null;
-  if (agg.settled >= MIN_AGG_SAMPLE && agg.clv_sample > 0) {
+  if (clvCaptureReliable() && agg.settled >= MIN_AGG_SAMPLE && agg.clv_sample > 0) {
     const dist = CLV_BUCKETS.map((b) => ({ ...b, count: 0 }));
     let counted = 0;
     for (const r of settledRows || []) {
@@ -534,6 +548,72 @@ async function getModelAggregate(opts = {}) {
   return agg;
 }
 
+// Truth-Everywhere Part 2 (item 7) — the public 30D accuracy VIEW, built from
+// the CLEAN ledger aggregate (model_value > 0), NOT the Redis outcome log
+// (which still counts degraded projection-0 rows and can't be filtered). Same
+// shape the AccuracyBadge / buckets consumed from outcomeService, so no
+// frontend change. Redis is a cache; when a cache can't be filtered, read truth.
+const ACCURACY_VIEW_SPORTS = ['mlb', 'wnba', 'nba', 'soccer'];
+function _aggToRecord(agg, sport) {
+  const byGrade = {};
+  for (const [tier, b] of Object.entries(agg.by_tier || {})) {
+    byGrade[tier] = {
+      hits: b.hits, misses: b.misses, pushes: b.pushes,
+      total: b.hits + b.misses + b.pushes, pct: b.hit_pct ?? null,
+    };
+  }
+  return {
+    sport,
+    updated_at: null,
+    window_days: agg.window_days,
+    sample: agg.settled,
+    min_sample: agg.min_sample,
+    overall: {
+      hits: agg.hits, misses: agg.misses, pushes: agg.pushes,
+      total: agg.hits + agg.misses + agg.pushes, pct: agg.hit_pct ?? null,
+    },
+    byGrade,
+  };
+}
+async function getAccuracyView(opts = {}) {
+  const base = { sb: opts.sb, nowMs: opts.nowMs };
+  const overallAgg = await getModelAggregate(base);
+  const sports = {};
+  for (const s of ACCURACY_VIEW_SPORTS) {
+    const a = await getModelAggregate({ ...base, sport: s });
+    if (a.settled > 0) sports[s] = _aggToRecord(a, s);
+  }
+  return {
+    overall: _aggToRecord(overallAgg, 'overall'),
+    sports,
+    min_sample: overallAgg.min_sample,
+    updated_at: null,
+  };
+}
+
+// Grade-tier buckets for the ledger accuracy strip, from the clean aggregate.
+function accuracyBucketsFromAgg(agg) {
+  // First-letter buckets (A+ folds into A for the public strip, matching the
+  // old outcomeService.accuracyBuckets contract), n≥20 gate per bucket.
+  const order = ['A', 'B', 'C', 'D', 'F'];
+  const rolled = {};
+  for (const [tier, b] of Object.entries(agg.by_tier || {})) {
+    const k = tier === 'A+' ? 'A' : tier[0];
+    rolled[k] = rolled[k] || { hits: 0, misses: 0, total: 0 };
+    rolled[k].hits += b.hits;
+    rolled[k].misses += b.misses;
+    rolled[k].total += b.hits + b.misses + b.pushes;
+  }
+  return order
+    .filter((k) => rolled[k] && rolled[k].total > 0)
+    .map((k) => {
+      const r = rolled[k];
+      const decided = r.hits + r.misses;
+      const pct = r.total >= MIN_AGG_SAMPLE && decided > 0 ? Math.round((r.hits / decided) * 100) : null;
+      return { grade: k, hits: r.hits, total: r.total, pct };
+    });
+}
+
 module.exports = {
   recordPipelineGrades,
   captureClosing,
@@ -542,6 +622,8 @@ module.exports = {
   applyRevision,
   countRowsForDate,
   getModelAggregate,
+  getAccuracyView,
+  accuracyBucketsFromAgg,
   MIN_AGG_SAMPLE,
   __internals: {
     rowsFromSnapshot, computeClv, clvResultOf, indexProps, gameIdFor,

tests/integration/accuracy.test.js

@@ -1,7 +1,8 @@
 'use strict';
 
-// Session 55 — the self-learning loop's public read endpoints. Redis is mocked
+// Truth-Everywhere Part 2 (item 7) — the public accuracy endpoints now read the
-// so these run offline; the store is seeded per-test via cacheGet.
+// CLEAN Postgres ledger aggregate (getModelAggregate / getAccuracyView), NOT the
+// Redis outcome log. We partial-mock ledgerService so these run offline.
 
 const request = require('supertest');
 
@@ -14,12 +15,37 @@ jest.mock('../../src/utils/redis', () => ({
   isDegraded: () => false,
 }));
 
+jest.mock('../../src/services/ledgerService', () => {
+  const actual = jest.requireActual('../../src/services/ledgerService');
+  return {
+    ...actual,
+    getAccuracyView: jest.fn(),
+    getModelAggregate: jest.fn(),
+    accuracyBucketsFromAgg: actual.accuracyBucketsFromAgg, // keep the real bucketer
+  };
+});
+
+const ledgerService = require('../../src/services/ledgerService');
 const app = require('../../src/app');
 
-beforeEach(() => { mockStore = {}; });
+const EMPTY_AGG = {
+  window_days: 30, min_sample: 20, settled: 0, hits: 0, misses: 0, pushes: 0,
+  hit_pct: null, beat_close_pct: null, clv_distribution: null, by_tier: {}, pending: 0,
+};
+
+beforeEach(() => {
+  mockStore = {};
+  ledgerService.getAccuracyView.mockReset();
+  ledgerService.getModelAggregate.mockReset();
+});
 
 describe('GET /api/accuracy', () => {
-  test('cold cache → valid empty-safe shape', async () => {
+  test('cold (no data) → valid empty-safe shape from the ledger view', async () => {
+    ledgerService.getAccuracyView.mockResolvedValue({
+      overall: { sport: 'overall', window_days: 30, sample: 0, min_sample: 20,
+        overall: { hits: 0, misses: 0, pushes: 0, total: 0, pct: null }, byGrade: {} },
+      sports: {}, min_sample: 20, updated_at: null,
+    });
     const res = await request(app).get('/api/accuracy');
     expect(res.status).toBe(200);
     expect(res.body).toHaveProperty('overall');
@@ -27,13 +53,14 @@ describe('GET /api/accuracy', () => {
     expect(res.body.min_sample).toBeGreaterThan(0);
   });
 
-  test('returns the persisted record when present', async () => {
+  test('returns the clean ledger record when present', async () => {
-    mockStore['accuracy:overall'] = {
+    ledgerService.getAccuracyView.mockResolvedValue({
-      sport: 'overall', updated_at: '2026-07-10T00:00:00Z', window_days: 30, sample: 20,
+      overall: { sport: 'overall', window_days: 30, sample: 20, min_sample: 20,
-      overall: { hits: 14, misses: 6, pushes: 0, total: 20, pct: 70 },
+        overall: { hits: 14, misses: 6, pushes: 0, total: 20, pct: 70 },
-      byGrade: { 'A': { hits: 8, misses: 2, pushes: 0, total: 10, pct: 80 } },
+        byGrade: { A: { hits: 8, misses: 2, pushes: 0, total: 10, pct: 80 } } },
-    };
+      sports: { mlb: { sport: 'mlb', overall: { hits: 14, misses: 6, pushes: 0, total: 20, pct: 70 }, byGrade: {} } },
-    mockStore['accuracy:mlb'] = mockStore['accuracy:overall'];
+      min_sample: 20, updated_at: null,
+    });
     const res = await request(app).get('/api/accuracy');
     expect(res.body.overall.overall.pct).toBe(70);
     expect(res.body.sports.mlb).toBeTruthy();
@@ -41,25 +68,25 @@ describe('GET /api/accuracy', () => {
 });
 
 describe('GET /api/ledger/accuracy', () => {
-  test('returns grade-tier buckets from the accuracy record', async () => {
+  test('returns grade-tier buckets from the clean ledger aggregate', async () => {
-    mockStore['accuracy:overall'] = {
+    ledgerService.getModelAggregate.mockResolvedValue({
-      sport: 'overall', updated_at: '2026-07-10T00:00:00Z', window_days: 30, sample: 15,
+      ...EMPTY_AGG, settled: 30, hits: 20, misses: 10, hit_pct: 67,
-      overall: { hits: 10, misses: 5, pushes: 0, total: 15, pct: 67 },
+      by_tier: {
-      byGrade: {
+        'A+': { settled: 3, hits: 3, misses: 0, pushes: 0, hit_pct: null },
-        'A+': { hits: 3, misses: 0, pushes: 0, total: 3, pct: 100 },
+        'A': { settled: 22, hits: 15, misses: 7, pushes: 0, hit_pct: 68 },
-        'A': { hits: 5, misses: 2, pushes: 0, total: 7, pct: 71 },
+        'B': { settled: 5, hits: 2, misses: 3, pushes: 0, hit_pct: null },
-        'B': { hits: 2, misses: 3, pushes: 0, total: 5, pct: 40 },
       },
-    };
+    });
     const res = await request(app).get('/api/ledger/accuracy');
     expect(res.status).toBe(200);
     expect(Array.isArray(res.body.buckets)).toBe(true);
     const grades = res.body.buckets.map((b) => b.grade);
-    expect(grades).toContain('A+');
+    expect(grades).toContain('A'); // A+ folds into A in the public strip
-    expect(grades).toContain('A');
+    expect(res.body.overall.pct).toBe(67);
   });
 
-  test('cold cache → empty buckets, never 500', async () => {
+  test('empty aggregate → empty buckets, never 500', async () => {
+    ledgerService.getModelAggregate.mockResolvedValue({ ...EMPTY_AGG });
     const res = await request(app).get('/api/ledger/accuracy');
     expect(res.status).toBe(200);
     expect(res.body.buckets).toEqual([]);

tests/unit/ledgerAggregateScope.test.js

@@ -65,6 +65,7 @@ describe('getModelAggregate scoping', () => {
   });
 
   test('user scope renders percentages at n≥20 like the public record', async () => {
+    process.env.CLV_CAPTURE_RELIABLE = '1'; // exercise the CLV math (suppressed by default, item 7)
     const rows = [
       ...Array.from({ length: 14 }, () => ({ outcome: 'hit', clv_result: 'beat', grade: 'A' })),
       ...Array.from({ length: 7 }, () => ({ outcome: 'miss', clv_result: 'faded', grade: 'B' })),
@@ -73,5 +74,6 @@ describe('getModelAggregate scoping', () => {
     expect(agg.settled).toBe(21);
     expect(agg.hit_pct).toBe(Math.round((14 / 21) * 100));
     expect(agg.beat_close_pct).toBe(Math.round((14 / 21) * 100));
+    delete process.env.CLV_CAPTURE_RELIABLE;
   });
 });

tests/unit/ledgerService.test.js

@@ -206,6 +206,27 @@ describe('captureClosing — real feed values only', () => {
 });
 
 describe('getModelAggregate — never a % under min sample', () => {
+  // beat_close/CLV are suppressed by default (item 7 — CLV capture broken until
+  // C4). These tests exercise the CLV MATH, so enable the reliable flag; a
+  // separate test below locks the default-suppressed behavior.
+  beforeAll(() => { process.env.CLV_CAPTURE_RELIABLE = '1'; });
+  afterAll(() => { delete process.env.CLV_CAPTURE_RELIABLE; });
+
+  test('beat_close is SUPPRESSED by default until C4 (CLV capture broken)', async () => {
+    delete process.env.CLV_CAPTURE_RELIABLE; // default state
+    const sb = fakeSb();
+    sb._state.selectResults = [[
+      ...Array.from({ length: 13 }, () => ({ outcome: 'hit', clv_result: 'beat' })),
+      ...Array.from({ length: 7 }, () => ({ outcome: 'miss', clv_result: 'faded' })),
+    ]];
+    sb._state.countResult = 0;
+    const agg = await ledger.getModelAggregate({ sb });
+    expect(agg.hit_pct).toBe(65);            // hit rate still renders (it's real)
+    expect(agg.beat_close_pct).toBeNull();   // BEAT CLOSE hidden — measured-wrong
+    expect(agg.clv_distribution).toBeNull();
+    process.env.CLV_CAPTURE_RELIABLE = '1';  // restore for the rest of the block
+  });
+
   test('below 20 settles → hit_pct/beat_close_pct null, counts real', async () => {
     const sb = fakeSb();
     sb._state.selectResults = [
@@ -296,6 +317,9 @@ describe('indexProps — prefers a book row with both sides priced', () => {
 // centralized HERE (getModelAggregate) — consumers never re-derive it.
 describe('getModelAggregate — clv_distribution (n>=20 gate lives in the service)', () => {
   const { clvBucketIndex, CLV_BUCKETS } = ledger.__internals;
+  // CLV suppressed by default (item 7); enable to test the distribution math.
+  beforeAll(() => { process.env.CLV_CAPTURE_RELIABLE = '1'; });
+  afterAll(() => { delete process.env.CLV_CAPTURE_RELIABLE; });
 
   test('below 20 settles → clv_distribution is null (never a small-sample chart)', async () => {
     const sb = fakeSb();