Stop persisting missed_window refusals (halt the bleed at source)

missed_window means 'this capture pass ran after first pitch'. It is a fact
about our cron cadence, not the market: once a game starts the same prop emits
a fresh refusal every ~20 minutes for the rest of the night, per book, per
side. Measured over 7 days of production that is 332,608 rows/day - 83.1% of
all closing_captures writes, ~66 MB/day - and since B1 filtered both readers,
nothing reads them.

The filter lives in persist(), not buildCaptureRows(), and that is the whole
trick: the caller computes the capture-rate alarm from the full in-memory
array, so filtering at build time would have blinded the ops alarm to the exact
condition it exists to catch. captureRateAlarm is pure; a test asserts the
caller still passes the full array, and that a 10-priced/90-late pass still
fires at 0.10.

Narrow by design: one_sided_price still persists (liquidity signal), priced
captures unchanged, and the rare fault refusals still persist because each
names a pipeline fault worth seeing. A missed_window row carrying a price is
kept.

Growth drops 400,469 -> 67,862 rows/day. The 4.2M historical rows are now
static, so the cleanup is a calm decision rather than a race.

Nothing deleted. Grade untouched.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

BUILD-STATE.md

@@ -3,6 +3,22 @@
 ## Last Updated
 2026-08-12
 
+## Fix B2 (2026-08-12) — halt the closing_captures bleed at source ✅
+4,793 tests / 373 suites, web build exit 0. **No rows deleted. Grade untouched.
+Readers untouched. No R2 involved.**
+- `closingCapture.persist()` no longer writes `missed_window` rows (both odds
+  null). Priced + `one_sided_price` + the rare fault refusals persist unchanged.
+  Revert with `CLOSING_PERSIST_MISSED_WINDOW=1`.
+- **Write reduction, measured on 7 days of production: 400,469 → 67,862
+  rows/day (−83.1%); ~80 MB/day → ~13.5 MB/day at 199.6 bytes/row.**
+- **Ops alarm intact** — `captureRateAlarm` is pure and the caller still passes
+  the full in-memory array, so a late pass still pages even though the refusals
+  are no longer stored. 13 tests, incl. a 10-priced/90-late pass asserting the
+  alarm still fires at 0.10.
+- **The 4,236,398 historical rows are now STATIC** (975 MB of a 1,314 MB DB).
+  The cleanup is now a calm scheduled decision, not a race — and a delete will
+  no longer be undone by regrowth.
+
 ## Fix B1 (2026-08-12) — dCLV close-leg filter ✅
 4,780 tests / 372 suites, web build exit 0. **Grade untouched (0 grade-related
 diff lines); no rows deleted or archived.**

CLAUDE.md

@@ -2116,6 +2116,31 @@ phased plan in the Session-57 conversation / BUILD-STATE Next section).
 - **This makes `missed_window` genuinely unread**, which is the precondition the
   reclaim order stopped on. The reclaim is still downstream and still needs R2.
 
+## B2 — stop writing the refusals nothing reads (non-obvious)
+- **`missed_window` is a fact about OUR CRON CADENCE, not the market.** Once a
+  game starts, `buildCaptureRows` emits a refusal for that prop on EVERY ~20-min
+  cycle for the rest of the night, x every book, x BOTH sides. Measured
+  **332,608 rows/day — 83.1% of all closing_captures writes — ~66 MB/day**, read
+  by nothing since B1 filtered both readers.
+- **THE FILTER LIVES IN `persist()`, NOT `buildCaptureRows()`, AND THAT IS THE
+  WHOLE TRICK.** The caller computes `captureRateAlarm({eligible: rows.length,
+  captured})` from the FULL in-memory array. Filtering at build time would have
+  silently blinded the ops alarm to exactly the condition it exists to catch —
+  a capture pass running late. `captureRateAlarm` is pure (two numbers, no DB),
+  so it never touched the stored rows; a test asserts the caller still passes the
+  full array.
+- **NARROW BY DESIGN.** `one_sided_price` still persists (real liquidity signal,
+  standing feature candidate). Priced captures unchanged. `unbound_game_time` /
+  `doubleheader_ambiguous` / `fetch_failed` still persist — 0 rows on record, and
+  each names a pipeline fault worth seeing, which `missed_window` does not.
+  A `missed_window` row that somehow carries a price is KEPT (`isUnreadRefusal`
+  requires both odds null) — do not over-cut.
+- `CLOSING_PERSIST_MISSED_WINDOW=1` restores the old behaviour.
+- **The bleed was the fire; the 4.2M backlog is the scar.** Growth drops
+  **400,469 → 67,862 rows/day (−83.1%)**, ~80 MB/day → ~13.5 MB/day. The
+  historical 4,236,398 rows are now STATIC — the cleanup can be scheduled
+  calmly instead of raced, and it no longer regrows after a delete.
+
 ## Active Skills
 - vyndr-voice (all user-facing output)
 - prop-analysis (grading methodology)

src/services/closingCapture.js

@@ -37,6 +37,37 @@ const WINDOW_MINUTES_BEFORE = Number(process.env.CLOSING_WINDOW_MIN || 45);
 const SHARP_BOOKS = new Set(['pinnacle']);
 const CAPTURE_RATE_FLOOR = Number(process.env.CLOSING_CAPTURE_FLOOR || 0.6);
 
+/**
+ * FIX B2 — REFUSALS THAT NOTHING READS ARE NOT PERSISTED.
+ *
+ * `missed_window` means "this capture pass ran after first pitch". It is a fact
+ * about OUR CRON CADENCE, not about the market: the same prop generates a fresh
+ * refusal on every ~20-minute cycle for the rest of the night, x every book,
+ * x both sides. Measured: 4,236,398 rows — 82.7% of `closing_captures` and ~75%
+ * of the entire database — accruing ~465k rows and ~92 MB per day.
+ *
+ * Both readers now filter them out (`attachClosingProb` always did;
+ * `computeDirectionalForRow` since B1), so they are read by nothing. Writing a
+ * row no query will ever select is pure cost.
+ *
+ * NARROW BY DESIGN. `one_sided_price` still persists — a book quoting one side
+ * only is real market-structure information and a standing feature candidate.
+ * Priced captures persist unchanged. `unbound_game_time` /
+ * `doubleheader_ambiguous` / `fetch_failed` also still persist: they are
+ * genuinely rare (0 rows on record) and each names a pipeline fault worth
+ * seeing, which `missed_window` does not.
+ *
+ * Set CLOSING_PERSIST_MISSED_WINDOW=1 to restore the old behaviour.
+ */
+const UNREAD_REFUSALS = new Set(['missed_window']);
+const persistMissedWindow = () => process.env.CLOSING_PERSIST_MISSED_WINDOW === '1';
+
+/** Is this row one nothing will ever read back? */
+function isUnreadRefusal(row) {
+  return !!row && UNREAD_REFUSALS.has(row.missed_reason)
+    && row.over_odds == null && row.under_odds == null;
+}
+
 function etDate(iso) {
   if (!iso) return null;
   const t = new Date(iso);
@@ -222,15 +253,29 @@ function captureRateAlarm({ eligible = 0, captured = 0 } = {}, opts = {}) {
  *  historical fact, and rewriting it would destroy the provenance that C4
  *  lacked in the first place. */
 async function persist(rows, deps = {}) {
-  const out = { attempted: rows ? rows.length : 0, written: 0, skipped: false, error: null };
-  if (!out.attempted) return out;
+  const all = Array.isArray(rows) ? rows : [];
+  // FIX B2 — drop the rows nothing reads BEFORE the insert. Filtering here (not
+  // in buildCaptureRows) is deliberate: the caller computes the capture-rate
+  // alarm from the FULL in-memory array, so the ops signal must keep seeing
+  // every refusal even though we no longer store them.
+  const keep = (deps.persistMissedWindow ?? persistMissedWindow())
+    ? all : all.filter((r) => !isUnreadRefusal(r));
+  const out = {
+    attempted: all.length,
+    to_write: keep.length,
+    skipped_unread: all.length - keep.length,
+    written: 0,
+    skipped: false,
+    error: null,
+  };
+  if (!keep.length) return out;
   try {
     const getClient = deps.getClient || require('../utils/supabase').getSupabaseServiceClient;
     const sb = getClient();
     if (!sb) { out.skipped = true; return out; }
     const CHUNK = 250;
-    for (let i = 0; i < rows.length; i += CHUNK) {
-      const chunk = rows.slice(i, i + CHUNK);
+    for (let i = 0; i < keep.length; i += CHUNK) {
+      const chunk = keep.slice(i, i + CHUNK);
       const { error } = await sb.from('closing_captures').insert(chunk);
       if (error) { out.error = error.message; break; }
       out.written += chunk.length;
@@ -243,5 +288,6 @@ async function persist(rows, deps = {}) {
 
 module.exports = {
   buildCaptureRows, captureWithRetry, captureRateAlarm, persist,
+  isUnreadRefusal, UNREAD_REFUSALS,
   WINDOW_MINUTES_BEFORE, SHARP_BOOKS,
 };

src/services/intradayRefreshService.js

@@ -232,7 +232,11 @@ async function runIntradayRefresh(sport, opts = {}) {
       const res = await closing.persist(rows);
       const captured = rows.filter((r) => !r.missed_reason).length;
       const al = closing.captureRateAlarm({ eligible: rows.length, captured });
-      console.log(`[intraday] closing capture ${sp}: ${captured}/${rows.length} priced, ${res.written} stored${al.alarm ? ' — RATE ALARM' : ''}`);
+      // B2 — `stored` is now less than `eligible` on purpose: post-first-pitch
+      // refusals are no longer persisted. The alarm still sees every one of them.
+      console.log(`[intraday] closing capture ${sp}: ${captured}/${rows.length} priced, `
+        + `${res.written} stored, ${res.skipped_unread || 0} unread-refusals dropped`
+        + `${al.alarm ? ' — RATE ALARM' : ''}`);
       if (al.alarm && deps.notify) {
         await deps.notify(`Closing capture degraded for ${sp.toUpperCase()} — ${al.reason}`,
           { title: 'VYNDR closing capture', priority: 'high', tags: ['rotating_light'] });

tests/unit/closingCaptureB2.test.js

@@ -0,0 +1,154 @@
+'use strict';
+
+/**
+ * B2 — stop persisting refusals nothing reads.
+ *
+ * `missed_window` is a fact about our cron cadence, not the market: the same
+ * prop generates a fresh refusal every ~20 minutes for the rest of the night,
+ * per book, per side. 4.2M rows, ~465k/day, read by nothing since B1.
+ *
+ * The cut must be NARROW — priced captures and `one_sided_price` are read, and
+ * the capture-rate alarm must keep seeing every refusal even though we no longer
+ * store them.
+ */
+
+const cap = require('../../src/services/closingCapture');
+
+const capture = (calls) => ({
+  getClient: () => ({
+    from() {
+      return { insert: async (chunk) => { calls.push(...chunk); return { error: null }; } };
+    },
+  }),
+});
+
+const priced = (p = {}) => ({
+  sport: 'mlb', player_key: 'a', stat: 'hits', side: 'over', game_date: '2026-08-12',
+  book: 'draftkings', line: 0.5, over_odds: -120, under_odds: 100, missed_reason: null, ...p,
+});
+const refusal = (reason, p = {}) => ({
+  sport: 'mlb', player_key: 'a', stat: 'hits', side: 'over', game_date: '2026-08-12',
+  book: 'draftkings', line: null, over_odds: null, under_odds: null, missed_reason: reason, ...p,
+});
+
+describe('what persists and what does not', () => {
+  it('a PRICED capture persists', async () => {
+    const calls = [];
+    const out = await cap.persist([priced()], capture(calls));
+    expect(out.written).toBe(1);
+    expect(calls).toHaveLength(1);
+    expect(out.skipped_unread).toBe(0);
+  });
+
+  it('a ONE_SIDED_PRICE refusal persists — it is a real liquidity signal', async () => {
+    const calls = [];
+    const out = await cap.persist([refusal('one_sided_price')], capture(calls));
+    expect(out.written).toBe(1);
+    expect(calls[0].missed_reason).toBe('one_sided_price');
+  });
+
+  it('a MISSED_WINDOW refusal is NOT persisted', async () => {
+    const calls = [];
+    const out = await cap.persist([refusal('missed_window')], capture(calls));
+    expect(out.written).toBe(0);
+    expect(out.skipped_unread).toBe(1);
+    expect(calls).toHaveLength(0);
+  });
+
+  it('a mixed pass writes exactly the readable rows', async () => {
+    const calls = [];
+    const rows = [
+      priced(), priced({ player_key: 'b' }),
+      refusal('one_sided_price', { player_key: 'c' }),
+      refusal('missed_window', { player_key: 'd' }),
+      refusal('missed_window', { player_key: 'e' }),
+    ];
+    const out = await cap.persist(rows, capture(calls));
+    expect(out.attempted).toBe(5);
+    expect(out.written).toBe(3);
+    expect(out.skipped_unread).toBe(2);
+    expect(calls.map((r) => r.player_key).sort()).toEqual(['a', 'b', 'c']);
+  });
+
+  it('OTHER refusals still persist — each names a pipeline fault worth seeing', async () => {
+    for (const reason of ['unbound_game_time', 'doubleheader_ambiguous', 'fetch_failed']) {
+      const calls = [];
+      const out = await cap.persist([refusal(reason)], capture(calls));
+      expect(out.written).toBe(1);
+    }
+  });
+
+  it('a missed_window row that somehow CARRIES a price is kept — do not over-cut', async () => {
+    const calls = [];
+    const out = await cap.persist([refusal('missed_window', { over_odds: -110, under_odds: -110 })], capture(calls));
+    expect(out.written).toBe(1);
+  });
+
+  it('the flag restores the old behaviour', async () => {
+    const calls = [];
+    const out = await cap.persist([refusal('missed_window')], { ...capture(calls), persistMissedWindow: true });
+    expect(out.written).toBe(1);
+  });
+
+  it('a pass that is ALL unread refusals never opens a client', async () => {
+    let opened = false;
+    const out = await cap.persist([refusal('missed_window'), refusal('missed_window')], {
+      getClient: () => { opened = true; return null; },
+    });
+    expect(opened).toBe(false);
+    expect(out.skipped_unread).toBe(2);
+    expect(out.written).toBe(0);
+  });
+});
+
+describe('THE OPS SIGNAL SURVIVES — the alarm never read the stored rows', () => {
+  it('captureRateAlarm is pure: it takes counts, not a database', () => {
+    const late = cap.captureRateAlarm({ eligible: 100, captured: 10 });
+    expect(late.alarm).toBe(true);
+    expect(late.reason).toMatch(/unrecoverable/);
+    const fine = cap.captureRateAlarm({ eligible: 100, captured: 90 });
+    expect(fine.alarm).toBe(false);
+  });
+
+  it('still fires on a late pass whose refusals are now dropped', async () => {
+    // 100 props, 10 priced, 90 post-first-pitch: nothing but the 10 persists,
+    // and the alarm must still see 10/100.
+    const rows = [
+      ...Array.from({ length: 10 }, (_, i) => priced({ player_key: `p${i}` })),
+      ...Array.from({ length: 90 }, (_, i) => refusal('missed_window', { player_key: `m${i}` })),
+    ];
+    const calls = [];
+    const out = await cap.persist(rows, capture(calls));
+    expect(out.written).toBe(10);
+    expect(out.skipped_unread).toBe(90);
+
+    // The caller computes the alarm from the FULL array, not from what persisted.
+    const captured = rows.filter((r) => !r.missed_reason).length;
+    const al = cap.captureRateAlarm({ eligible: rows.length, captured });
+    expect(al.alarm).toBe(true);
+    expect(al.rate).toBeCloseTo(0.10, 5);
+  });
+
+  it('the caller still passes the FULL array to the alarm', () => {
+    const src = require('fs').readFileSync(
+      require('path').join(__dirname, '..', '..', 'src/services/intradayRefreshService.js'), 'utf8');
+    expect(src).toMatch(/captureRateAlarm\(\{ eligible: rows\.length, captured \}\)/);
+    expect(src).toMatch(/const captured = rows\.filter\(\(r\) => !r\.missed_reason\)\.length/);
+  });
+});
+
+describe('buildCaptureRows is UNCHANGED — the alarm depends on it', () => {
+  it('still emits missed_window rows in memory', () => {
+    const started = { player: 'A', stat_type: 'hits', game_time: '2026-08-12T00:00:00Z', book: 'dk' };
+    const rows = cap.buildCaptureRows('mlb', [started], { now: new Date('2026-08-12T02:00:00Z') });
+    expect(rows).toHaveLength(2);                       // both sides
+    expect(rows.every((r) => r.missed_reason === 'missed_window')).toBe(true);
+  });
+
+  it('isUnreadRefusal identifies exactly those rows', () => {
+    expect(cap.isUnreadRefusal(refusal('missed_window'))).toBe(true);
+    expect(cap.isUnreadRefusal(refusal('one_sided_price'))).toBe(false);
+    expect(cap.isUnreadRefusal(priced())).toBe(false);
+    expect(cap.isUnreadRefusal(null)).toBe(false);
+  });
+});