Aug 11, 2026

Regime & Verdict#

• Regime: UPTREND · ret5 −0.100% · ret10 +4.010% · dd15 −1.540% (scripts/regime_check.py, authoritative)
• Vol-state: VIX 15.28 (LOW tercile), normal contango (VIX3M 18.91). SPY dealer gamma SHORT — spot 770.60 vs zero-gamma flip 789.87, 2.44% below (last night: pinned at the flip). QQQ spot 718.26 vs flip 769.03, 6.6% below, short gamma — both sources now agree, the first agreement on the short side this stretch.
• Breadth: price breadth 54.67% green vs options-flow bullish 36.1% — an 18.6pt gap that persisted even as the tape strengthened. Rotation OUT of Technology + both Consumer sectors, INTO Energy / Financials / Industrials (Tech outflow now two-session-persistent).
• directional_tradable = TRUE · s1_standdown = FALSE
• Bottom line: no new starters. Zero directional calls sized for a TENTH consecutive session (07-29 → 08-11); the book stays FLAT for a SEVENTH. Two OI_FADE shorts reached Phase D and both landed at WATCH on two independently sufficient gates. Vol book empty.

What actually changed tonight#

The crash/rebound guard expired on 08-10, and tonight is the session that shows what its lifting meant: the thrust has fully flattened. ret5 has decayed +5.53% (08-05 peak) → +3.51% → +2.03% (08-10) → −0.10% tonight. The UPTREND label is now trailing a decayed number rather than describing live acceleration — ret10 (+4.01%) is carrying the label on its own.

Simultaneously SPY dealer gamma went from pinned at the flip to 2.44% clear below it, and QQQ's two gamma sources resolved their 08-10 disagreement — to SHORT, not to a tradeable long-gamma read. Aggregate gamma sign (wing-width) and spot-vs-flip (regime/sell decision) are reported separately below and were not merged.

Read together — decayed thrust, persistent flow/price breadth divergence, short gamma on both index books, and three consecutive Tier-1 macro prints — this is a transitional tape wearing an uptrend label, not a confidently in-regime one.

---

Directional Book (excess-scored)#

EMPTY — no name sized. Both Phase D candidates are shown for the record at their final tier.

Tier arithmetic — why nothing sized#

OI_FADE's pooled prior (+0.59% mean) sits in the MEDIUM (+0.3–1%) starter band on its own. Two separate gates each apply −1 tier, and either one alone takes starter → WATCH:

1. Fundamentals CAUTION (−1) on both names.
2. Event risk (−1): CPI 08-12, PPI 08-13, Advance Retail Sales 08-14 — three Tier-1 prints on days 1–3 of the h10 window an 08-11 entry opens (closes ~08-26).

This is overdetermined, not a close call, and explicitly not an additive-confluence downgrade — two independent gates reaching the same floor. Half-cap/no-auto-full never came into play; neither name approached the ≥1% HIGH band.

The out-of-regime half-cap (P0.6) would have been applied on tonight's transitional read, but is non-binding — the starter band already sits below the half ceiling.

Per-name notes#

COO — Cooper Companies ($76.82, Healthcare/medical devices, $150.5M ADV, Common Stock) Build accelerated: rel_build 1.029 → 1.317, net_5d +4,619 → +6,113, persistence 0.659 (organic, <0.85). But 66% of the 5-day build landed in a single session (today, +4,028). Per the standing M&A-rumor gate gap, the news tape was checked directly — nothing COO-specific in the 08-05→08-11 window; no deal, rumor, or activist catalyst. The spike is positioning, not fundamentals. That is a build-quality concern layered on top of the formal gates, not itself a gate. Fundamentals CAUTION, explicitly unchanged from 08-10 — no new fact on either side. Contradicting the short: 4/4 EPS beat streak, insider MSPR +40.33 (net buying), D/E 0.30. Confirming it: EPS growth −42.61% YoY, revenue +6.09%, ~63–65x trailing P/E, UBS Neutral PT $75, below spot. Short float 4.30%, days-to-cover 2.96 — low squeeze risk. RSI 67.49. Worth carrying: COO has closed higher in 6 of the last 7 sessions ($72.32 → $76.82, +6.2%) and is ~4.6% below the $80.50 reference. This is a call-OI build into strength, on a name approaching overbought.

NRG — NRG Energy ($119.79, Utilities/IPP, $383.2M ADV, Common Stock) — new, no prior adjudication Mechanically the cleaner build of the two: persistence 0.315 (well-distributed across five sessions) vs COO's 0.659, on 2.5× the liquidity. net_5d +66,359 (call +74,076 / put +7,717), rel_build 0.835. Catalyst boundary re-verified by hand. NRG gapped −15.5% on its 08-04 Q2 print ($138.47 → $117.04, 13.8M sh vs 2.6M normal). The shipped catalyst_split >= reads 60.4% post-catalyst; the strict > recompute reads 66.3% — because the print day itself was net −5,937 (calls closed into the print). Excluding it raises the share. Both boundaries agree LIVE_BUILD: the known print-day defect strengthens this name rather than flipping it — the opposite of the EWTX case that killed that name on 08-10. The build is a genuine post-crash dip-buying crowd rebuilding call exposure well past the print. Fundamentals CAUTION. Confirming the short: second consecutive EPS miss (Q1 −17.63%, Q2 −13.08%), insiders did not buy the dip (3mo MSPR −60.91; July −100, Aug −69.63), leverage high (D/E 5.55–9.78). Contradicting it: a dated, post-crash bullish catalyst stack — SeekingAlpha upgrade to Buy (08-06) citing the LS Power acquisition and a new 1.2-GW Texas hyperscaler BYOP project ($500M incremental annual EBITDA); Scotiabank (08-05) and Evercore ISI (08-05) both reiterate Outperform at $211 / $195 — roughly 63–76% above spot. Forward P/E 11.10, PEG 0.69, EPS Next Y +22.42%. The risk here is a fundamentals re-rating, not a covering squeeze (short float 3.06%, days-to-cover 2.12). Since the crash NRG has held a tight $116.56–$124.71 band, closing mid-band.

---

Vol Book (non-directional)#

EMPTY — 0DTE-VRP stands aside on both SPY and QQQ, for cleaner reasons than last night's source conflict.

The pooled headline net_expectancy is positive (SPY +0.127%/day, QQQ +0.255%/day) — but it pools all VIX terciles. Tonight is unambiguously LOW tercile (VIX 15.28), and the conditional net-of-cost is negative for both:

Acting on the conditional, not the pooled figure. Independently, both books are dealer-short-gamma (spot below flip) — a structural stand-aside on its own. QQQ additionally shows front-end backwardation (0DTE IV 1.35× VIX). And CPI/PPI land inside the next two 0DTE sessions — event vol this sleeve's validation sample never priced.

Earnings IV-crush — ADVISORY ONLY, no backtested harness exists for this stack. Four candidates in confirmed backwardation: TRMB (08-12, IV-rank 100, ~8.0% implied vs 2.42% typical, ~3.3×, best liquidity) · ALMS (08-12, ~25.0% implied vs 4.44%, ~5.6×, thin OI) · RRGB (08-12, ~15.9%, no realized-vol cross-check available — unconfirmed) · BSP (08-13, ~13.5%, liquidity-marginal). All three 08-12 names print into CPI — the implied move embeds single-name risk, not a macro-print tape, so wings need sizing well past it. Not sized.

---

Watch / Stood-down#

• MOM_SHORT — 12 names (APP, WING, RBA, ONON, VICI, ALHC, ACM, PEG, CMS, GME, ROL, GPI), up from 7 on 08-10. Watch-only, never sizes — knowingly negative baseline (−0.0128, n=803), the one recorded exception to "no lane negative-excess". s1_standdown is FALSE, so the standing invariant, not tonight's regime, is the binding cap.
  • Data-quality: 4 of 12 (APP, WING, RBA, ONON) carried negative UW pct_52w_range (stale w52l above close). All four Yahoo-verified as genuine near-lows (0.1% / 0.4% / 0.3% / 3.6% of range) — the screener filter works, the reported field is corrupt.
  • GPI re-emerged after being dropped 08-10 on a split/adjustment artifact — flagged, not promoted; re-verify before it is ever scored.
• MOM_LONG — 20-name basket top slice. Basket/watch only, never sizes; baseline −0.0108 (n=1066), DURABLE-N bar unmet. All 20 carried UW pct_52w_range >1.0 (out of range); all 20 Yahoo-verified as genuine near-highs (96–100%).
• S2 — 80 names post-gate (41 BUY-side). Advisory-only, never sizes. The lane did not complete its news gate — the full list is carried as news-unverified and was deliberately withheld from the structured write-back.
• S4 — 17 names, all gates pass. Advisory-only, never sizes. Plus a 4-name ivrank_chg_5d h3 tilt (LYG, LBRT, IMAX, INCY) reported separately, never stacked.

---

Risk#

• Correlation clusters: COO (Healthcare) and NRG (Utilities/IPP) share no sector, factor, or catalyst thread — two singleton clusters. No pairwise coefficient was computed; sector distinctness was treated as sufficient for two unrelated names. A numeric check should gate any future night that pairs two OI_FADE names inside the same sector.
• Event calendar: CPI 08-12 · PPI 08-13 · Advance Retail Sales 08-14 (all Tier-1) · UMich prelim 08-14 (Tier-2). FOMC 09-15/16, outside the window.
• Tail caps: N/A — OI_FADE is not one of the gate-6 tail-capped lanes (MOM_SHORT / S2-long / vol-short), and none of those sized.
• Hedge note: book is flat; no hedge required.
• Short gamma + three Tier-1 prints means realized vol tomorrow is likely understated by VIX 15.28. Anything opened here should assume wider realized ranges than the LOW tercile implies.

Tooling defects found tonight (worth fixing, not blocking)#

1. finnhub_enrich.py leverage_flag is wrong — labelled NRG's D/E of 5.55–9.78 as "low". Both Finnhub's own metric and fz disagree with the script's own label. A threshold bug in the classifier; it would soften a genuinely levered name's risk read.
2. fz CLI does not expose Recom / Target Price in any view checked (overview / valuation / technical / financial / ownership), and bare fz news <T> / fz insider --ticker return malformed output. Analyst PTs had to come from Finnhub news; insider-cluster data was unavailable for both names.
3. Phase D hand-rolled jsonschema instead of scripts/validate_decision.py. Re-validated with the prescribed script — VALID. (Envelope also patched: NRG's invalidation was an unevidenced "~$125"; now pinned to the actual $124.71 post-crash swing high from chart.py.)

---

Truth-set parquets (prices / returns / features) were stale by one session at preflight and were rebuilt before any lane read — prices 2,449 tickers / 342,980 rows, features 85 dates / 2,478 tickers. Re-verified clear. Prior book (conviction_2026-08-10.json) confirmed genuinely flat — 0 open, 0 closed — checked against the file directly, not on the bare held_book.py empty result.