Jul 24, 2026

Regime & Verdict#

• Regime: PULLBACK (unchanged tier, deepened) · VIX 18.58 (−0.12 d/d, contango intact VIX9D 17.62 < VIX 18.58 < VIX3M 20.51 — front-end eased, vol de-escalating, not stress) · short-gamma amplification (SPY GEX −1.556B, QQQ GEX −1.753B, both FULLY_NEGATIVE) · breadth 72.17% green · directional_tradable=true · s1_standdown=**FALSE`
• SPY 738.93 (+0.10%), ret5 −0.587% / ret10 −2.122%, below 20-SMA (746.15) and 50-SMA (745.07). ret5 stabilized (−1.67%→−0.59%) while ret10 deepened (−1.80%→−2.12%). A grind-down, not a crash.
• s1_standdown FALSE is affirmatively supported, not defaulted: ret10 −2.12% is past the −2% confirmed-downtrend boundary, so this is explicitly not the unconfirmed-V-bounce setup the guard exists to catch. Neither the up-thrust leg (ret5 > +2.5%) nor the unconfirmed-dip leg (ret5 < −2% AND ret10 > −2%) fires.
• Breadth is bifurcated — the day's most informative reading. 72.17% of the S&P was green (363 adv / 140 decl, median +0.97%) while SPY itself managed +0.10%: the median stock rallied, mega-cap tech dragged the cap-weighted index. But options-flow breadth reads only 33% bullish (2,074 vs 4,215 tickers). Price breadth broad-bullish, positioning broadly hedged. Logged as a distribution/hedging tell — two-sided, not a confirmation of the short book.
• Tier-1 event cluster INSIDE h10: FOMC 07-29 (+ MSFT/META), Core PCE 07-30 (+ AAPL/AMZN), NFP 08-07 (h10 boundary). Three macro prints and four Mag7 reports on a fully-negative-GEX tape.
• Bottom line: 0 new directional starters. The only name to clear the full OI_FADE gate stack (ALLE) was VETOED on fundamentals. Every other lane returned watch/advisory. The live decisions tonight are de-risking the held book: ORCL TRIM 50% + invalidation tightened; PLNT HOLD with a hard pre-ER exit date. This is the modal, correct output.

Directional Book (excess-scored)#

New starters tonight: 0.

ALLE — the one name that cleared the lane, and why it still didn't trade#

OI_FADE ranked ALLE as the single clean fresh short: 5d net call-OI +3,901 (call +3,998 / put +97), positive every one of five days, persistence_ratio 0.455 (organic, not a block artifact), relative build 2.30× its 30-day average call OI, PCR 0.05. Liquidity $224M ADV, ER 10-22 clear of h10, mid-range at 48.8% of its 52w band (no breakout tail), no squeeze. Independently reproduced from the raw panel.

The lane's discriminator vs SSNC held up on inspection — SSNC's build stopped dead the moment its catalyst landed (+335/+3,070/+731/0/+1) while ALLE's continued through and after its print (+880/+653). But two things the lane understated:

1. ALLE's build is 61% pre-print (+2,368 over 07-20→22) vs 39% post-print (+1,533) — real chasing, but not the purely-post-print pattern described.
2. ALLE gapped +10.45% on 07-23 on earnings and held the gap (−0.78% on 07-24). Shorting that is fighting PEAD, and the OI_FADE prior (n=640) was measured on a general population — not conditioned on "one day after a +10% earnings surprise." Out of sampled support.

Fundamentals gate returned VETO, and the evidence is decisive: Q2 EPS $2.40 vs $2.256 consensus (+6.39%), revenue $1.15B +13% YoY, management RAISED FY26 guidance (adj. EPS $8.85–9.00), and two same-day post-print PT raises — JPMorgan $150→$170, Baird $175→$190, both 11–24% above the $153.36 close. The call-OI build is not an unwarranted crowd; it has a live, dated, sell-side-endorsed justification underneath it. Risk-sizer concurred. Watch-only.

Held-book reconciliation (path-aware Yahoo chart-API closes + fresh 07-24 UW panel)#

• ORCL (short @ $131.54, 07-13): close $114.99 → +12.58% gross. Fell −4.21% vs SPY +0.10% = +4.31% short-excess today. Printed a fresh 52-week low ($114.75 intraday); pct_52w_range 0.1% — the literal bottom of its range, −66.7% off the $345.72 52w high (Sept-2025 OpenAI-cloud spike; verified genuine, not a data artifact). RSI-14 27.4 (Wilder, independently computed).
  • ACTION: TRIM 50% at $114.99; tighten the remaining half to close > $123.50 with 2-day confirmation.
  • Why: the standing invalidation (RSI>45 AND close>$144.22) had become vestigial — an AND-conjunction set at entry, now 25.4% away on the price leg. ORCL could rally 25% without triggering a documented exit, and $144.22 sits above the $131.54 entry, so holding to the stop would surrender the entire +12.58% gain and turn the trade into a loss before firing. A maximally-extended short (pct_52w 0.1%, RSI 27) with no practical stop, heading into FOMC + PCE + NFP on short gamma where a dovish surprise gets amplified by dealer hedging, is not a risk-managed position. The trim caps giveback near half the open gain. Lane remains historically positive-excess, which is why this is a trim and not a close.
  • No adds — MOM_SHORT tail cap forbids new sizing until a DURABLE-N≥30 forward re-validation clears.
• PLNT (short @ $55.20, 07-20): close $53.48 → +3.12% gross; −1.38% vs SPY +0.10% = +1.48% short-excess. 5d net call−put OI: +24,376 → +750 → +395 → −4 → +180. Today call_net +173 / put_net −7 — calls still gross-OPENING, so the tightened invalidation (requires a genuinely negative call_net, not a put-driven wash) is NOT triggered. Fundamentals CONFIRM (0 post-print PT raises; DB cut $61→$55 supports the short). HOLD, no add. Hard exit: close by 2026-08-05, the last full session before the 08-06 pre-market ER (which sits 1 trading day inside h10).

Vol Book (non-directional, net-of-cost, delta-neutral — advisory)#

• 0DTE-VRP: PARTICIPATE both SPY & QQQ, capped at 1.0× (NOT the lane's proposed 1.5×). Both gates that forced stand-aside on 07-23 released: VIX Δ −0.12 (no spike trip) and QQQ's front-IV/VIX ratio normalized 1.44 → 0.94 (no backwardation trip). Net expectancy SPY +0.175%/day, QQQ +0.305%/day after ~0.10% round-trip cost. Down-weighted deliberately: the 60-day sample contains no vol-shock day (left tail unsampled), a 5-day lookback reclassifies both to MID/1.0×, and this is the densest Tier-1 macro week of the quarter on short gamma — the wrong week to lean into extra short-vol size. Re-run nightly; tonight's scalar is not durable through the cluster.
• Earnings IV-crush SELL-VOL — defined-risk, wide wings, small size:
  • Tier A (clean, implied move = normal basis): CDNS (07-27, IV-rank 93.9, backwardation confirmed 83%→58%, implied ~8.9%), WHR (07-27, IV-rank 93.0, 14.3%), SWKS (07-28, IV-rank 96.5, 14.0%), ECL (07-28, IV-rank 95.4, 5.9%).
  • Tier B (macro overlap — implied move is a FLOOR for wings, not a ceiling; size smaller): SIMO (07-29, FOMC, IV-rank 95.5, 22.9%, PCR 11.4 — largest binary in the set), FSLR (07-30, PCE, IV-rank 100, backwardation 94%→80%, implied ~10.3%), ALGM (07-30, PCE, IV-rank 86.7, 20.0%), EME (07-30, PCE, IV-rank 82.2, 9.9%). Do not stack full size across more than one Tier-B name — correlated macro exposure.
  • DATA-QUALITY CATCH: uw insights earnings-play returned a near-zero implied_move_perc (0.41–0.66%) for CDNS and FSLR — a 0DTE contract contaminating the field — while both names' own term structure independently confirms real backwardation. Manual straddle re-derivation gives 8.9% and 10.3%. Taken at face value that field would have sized wings ~20× too tight. Do not trust it for these names without a cross-check.

Watch / Stood-down#

• OI_FADE: ALLE watch-only (fundamentals VETO). SSNC excluded — mechanically top-ranked but its build was 99.98% pre-print and died on the catalyst (+10.35% earnings breakout); correct pre-emptive real-catalyst invalidation. FHN — the 07-17 block confirmed rolled off; the residual build is now genuinely organic (persistence 0.556, positive all 5 days) but oi_rel_build is only 0.061 vs ALLE's 2.30, and it sits at pct_52w 0.848. Downgrade reason shifts from "block artifact" to "organic but sub-threshold + near-52w-high." Still not sized. FIG — most persistent build in the panel (net_5d 71,681, persistence 0.499) but doubly blocked: squeeze (SF 33.16% / DTC 3.13) and ER 08-05 inside h10. Excluded on block/reversal artifacts: MWH (0.999), YMM (0.997), ALLY (1.25, already unwinding), FR, GPC (genuine call unwind). Universe hygiene: CCXI (SPAC), VXX (ETN), SPCX (sub-60-td, SpaceX IPO 06-12).
• MOM_SHORT — lane-capped watch-only: MP (−5.50% below 52w low, but ER 08-06 inside h10), PSKY, UBER, ECHO, LBRDK. QS DROPPED — close $4.89, below the $5 floor, fail-closed. No promotion until a DURABLE-N≥30 forward-positive-excess re-validation via /calibration-audit.
• MOM_LONG basket (tail-only, never per-name): ~53% Financial Services — correlation-cluster flag. The lane's "60→15 names" contraction claim is an artifact of its own query LIMIT 15, not a market fact — treated as unverified.
• S2 liquidity-reversion (advisory-only, long-tilt is a crosswind in PULLBACK): ED, TD, DLTR, PENG, PRU, GPN, CYTK, TECK, TFX, VNO, LQDA. ED flagged for signal instability — its dark-pool buy-share flipped 35.5% (07-23) → 92.9% (07-24) in one session.
• S4 sentiment-contrarian (advisory-only, half-cap): PLD (PCR 5.42, ER 10-21, $637M ADV — cleanest; Q2 beat + raised guidance means the put-heaviness is hedging, not informed), VIK (PCR 16.96, ER 08-18), TTAN (PCR 12.67, ER 09-03), XENE (PCR 12.27, ER 08-10). Excluded on informed-put risk: DOX (guidance tightened + CFO transition), NI (KeyCorp cut 07-22), WEN (guidance cuts + PT cuts), KGS (Barclays PT cut). Carried JBL/LI/AMRZ all fell out of the top-5% PCR cross-section.

Risk#

• Correlation: ORCL (Technology) vs PLNT (Consumer Cyclical) — different sectors, different lanes, no cluster collapse required. Portfolio-level flag: both held legs are SHORT and the vol book runs short-premium — three short-gamma-adjacent exposures stacked into the same FOMC/PCE window on fully-negative GEX. Not a formal gate-3 cluster, but a real net-short-beta concentration. The hedge is sizing discipline, not a new position (a hedge outside the validated lane set would violate no-additive-confluence): the ORCL trim is the highest-leverage de-risking action available, and capping 0DTE-VRP at 1.0× avoids a fourth correlated leg.
• Event calendar: FOMC 07-29, PCE 07-30, NFP 08-07 all inside h10, plus MSFT/META and AAPL/AMZN. Applied as −1 tier; it has no room left to bite on new names (all already watch/advisory), so it binds on the held book — reinforcing both the ORCL trim and PLNT's hard pre-print exit.
• Tail caps applied: MOM_SHORT PROVISIONAL (no new starters, no adds to ORCL); OI_FADE fundamentals veto on ALLE; vol-short capped 1.0× for the unsampled left tail; Tier-B earnings wings floored at implied move.
• Data notes: data/prices.parquet / returns / features all still end 2026-07-17 — the panel is stale by 5 sessions despite commit d784c8e claiming extension to 07-18; the extension did not land. All metrics tonight computed from the fresh UW 2026-07-24 panel + live Yahoo chart-API closes. The momentum lane read the regime off that stale parquet and returned ret5 −1.54% / ret10 −0.20% / "CHOP" with a "relief-bounce inflection" warning — wrong, and discarded; Phase A's live numbers (independently re-verified) govern. Corrected: PLNT $-ADV is ~$87M, not the ~$250M figure used mid-session (clears the $50M floor either way).

Lane defect log (for /calibration-audit)#

1. Earnings-gate boundary leak — recurred in TWO lanes tonight. S2 passed 5 names whose ER lands inside its own h5 window (AOS/BAX/KKR/PBF 07-30, CCJ 07-31); S4 passed 4 whose ER lands inside h10 (DOX/NI 08-05, KGS 08-06, WEN 08-07). Both filtered on a boundary 2–3 days short of their stated horizon. Same defect class as the IBM leak from the 07-10 audit. Neither leak propagated — S2/S4 are advisory-only and S4's four were independently caught by the informed-put screen — but the gate should not depend on a second filter catching its misses. Propose: derive the exclusion window from the lane's own horizon rather than a hardcoded date.
2. Stale-parquet regime fallback (momentum lane) — a lane silently computing regime off the truth-set edge produces an inverted read (unconfirmed dip vs confirmed downtrend) and an unwarranted caution narrative. Propose: lanes must either read regime from the Phase A hand-off or fail loudly when the panel edge predates the trade date.
3. implied_move_perc 0DTE contamination in uw insights earnings-play (CDNS, FSLR) — propose a sanity check against the nearest post-event expiry straddle.
4. Carried from 07-23, still un-formalized: persistence_ratio >0.85 block screen (worked well tonight — caught MWH/YMM/ALLY), SPAC/shell exclusion (caught CCXI), sub-60-td listing exclusion (caught SPCX).

---

Regression gate: NOT TRIGGERED — no lane or threshold change applied tonight. All dispositions follow the standing 2026-07-18 calibration-audit rules (MOM_SHORT watch-only cap for new starters, S2/S4 advisory-only, excess-as-currency, no additive confluence, half-cap discipline, universe hygiene). The ORCL invalidation tightening and 50% trim are per-name Phase-D risk judgments on a held position, not standing lane rules; the 0DTE 1.0× cap is a per-session tail judgment, not a threshold change. The four lane-defect items above are logged propose-only and did not alter tonight's zero-starter outcome — no python3 scripts/retro_harness.py --all run required this session.