Market scan · Aug 4, 2026

RegimeUPTREND/REBOUND-THRUST

breadth71.4% green5d+4.11%10d+3.08%Directional tradableS1 stand-down

Vol state

VOL-SUPPRESSION. Dealers net long gamma on both index proxies: SPY total_gex +$1.104B, spot 772.17 vs zero-gamma 767.91
QQQ total_gex +$623.7M, spot 724.09 vs zero-gamma 711.88. VIX ~16.5 (~48th pct). Script and CLI agreed on the gamma sign for BOTH symbols and the orchestrator re-verified independently via `uw options-structure gex --symbol {SPY,QQQ}` -- the zerodte_setup.py sign bug did not bite (THIRD consecutive session of agreement
still unfixed, absence of symptom is not a fix). CAVEAT the lane did not raise: the index-level 30-DAY VRP reads FAIR (-0.002), i.e. IV essentially tracks realized. That is a different and longer measure than the 0DTE VRP and the two can legitimately disagree, but conviction on the index premium-selling book should be read down accordingly and they must NOT be quoted as agreement. dd15 -2.520% (regime_check.py, no schema field for it). s1_standdown fired DECISIVELY on leg (a1): ret5 +4.110% vs the +2.5% trigger, a 1.61pp margin -- versus 0.0139pp on 2026-08-03, so no knife-edge this time. The 2026-08-01 audit measured this guard at real N for the first time: 100 suppressed name-days, mean -1.73%, median -1.99%, hit 0.36, p=0.010 [DURABLE]
day-clustered 15 guard-days mean -1.86%, p=0.070. The guard SAVES money.

2 event-risk item(s) flagged.

Tier mix

No calls cleared the board on this scan.

No call cleared HIGH/MEDIUM/ADVISORY tier in this scan — no new directional edge.

SPY

vol_0dtesell premium

Vol state: GO_PREMIUM_SELL_INTRADAY, size multiplier x0.5 (LOW vol_state). Long gamma CONFIRMED TWO WAYS: zerodte_setup.py reads long-gamma/LOW, and `uw options-structure gex --symbol SPY` reads regime POSITIVE with spot 772.17 above zero-gamma 767.91 -- orchestrator re-verified the CLI leg independently. Gross open-entry mean 0.223%/day -> NET 0.123%/day after a 0.1% round-trip cost assumption. Front-end dte0 25.0% -> dte1 16.0% -> belly 13.0-13.3% at dte6-7 is ORDINARY same-day 0DTE pinning, not a genuine event hump; front_iv/VIX ratio 0.97 sits below the 1.25 caution gate. No VIX-spike stand-aside. Left tail remains UNSAMPLED -- advisory only.
Implied move: 0.89%
Expected range: 0.83%
Net expectancy: 0.001

defined-risk iron fly / short straddle centred ~772.3, wings ~ +/-0.83%

QQQ

vol_0dtesell premium

Vol state: GO_PREMIUM_SELL_INTRADAY, size multiplier x0.25 (LOW vol_state, then HALVED again for the front-end hump). Long gamma CONFIRMED TWO WAYS: script long-gamma/LOW, and CLI regime POSITIVE with spot 724.09 above zero-gamma 711.88 -- orchestrator re-verified. Gross 0.341%/day -> NET 0.241%/day. CAUTION flagged: 0DTE IV is 1.76x VIX. Per the 08-03 lesson the FULL curve was pulled and the near-week hump is GENUINE, not a 2-point-ratio artifact -- dte1 30.8% -> dte2-3 ~28.0% -> trough dte6-7 ~22.8-23.0% -> RE-RISING to 24.2-25.1% by dte8-10. The caution and the x0.25 size are real conditions, not guard artifacts. Left tail UNSAMPLED -- advisory only.
Implied move: 1.83%
Expected range: 1.45%
Net expectancy: 0.002

defined-risk iron fly centred ~722.25, wings ~ +/-1.45%

CACI

vol_earningssell premium

Vol state: DTE 1, BACKWARDATION. Historical realized move mean 0.68% (n=8, max 2.5%) -- implied is ~12x the mean, the widest ratio in the book. NO WEEKLY: the nearest listed expiry is 17DTE (2026-08-21).
Implied move: 8.19%

defined-risk iron fly at the 17DTE expiry, wings WIDE

Stand aside TRADEABLE but ORCHESTRATOR FLAG NOT RAISED BY THE LANE: the 17DTE expiry straddles the print PLUS both CPI (08-12) and PPI (08-13) and carries 12 extra decay days. That is unhedged Tier-1 macro stacked on the earnings binary -- a materially different trade from the weekly-expiry names. Size for that or wait.

TTMI

vol_earningssell premium

Vol state: DTE 1, BACKWARDATION, weeklies available (dte3). Historical realized move mean 3.19% (n=7, max 10.6%) -- implied ~6x the mean and still ~2x the fattest historical print.
Implied move: 18.71%

defined-risk iron fly, wings WIDE for the single 10.6% tail

DBX

vol_earningssell premium

Vol state: DTE 2, BACKWARDATION, weeklies available. Historical realized move mean 1.05% (n=8, max 2.4%) -- implied is >2x the WORST print in two years. Cleanest risk/reward in the book. NOTE: DBX also sits on the OI_FADE prior-exclusion list; that is a DIRECTIONAL exclusion and does not bear on a delta-neutral vol trade.
Implied move: 5.92%

defined-risk iron fly, wide wings

COHR

vol_earningssell premium

Vol state: DTE 8, BACKWARDATION, weeklies available. Historical realized move mean 4.12% (n=8, max 7.9%) -- implied clears the historical MAX by a real margin.
Implied move: 8.54%

defined-risk iron fly, wide wings

Stand aside Tradeable, but at DTE 8 the print lands ~2026-08-14, AFTER CPI (08-12) and PPI (08-13). Confirm the chosen expiry does not carry unwanted macro on the way in.

TLN

vol_earningssell premium

Vol state: DTE 1, BACKWARDATION, weeklies available. Historical realized move mean 1.88% (n=8, max 6.0%).
Implied move: 7.60%

defined-risk iron fly, REDUCED size

Stand aside Size down: the LAST print (6.0%) nearly matched implied, i.e. realized moves are closing the gap. Power/AI-datacenter theme name -- the historical sample may not reflect the current regime.

BCO

vol_earningsbuy premium

Vol state: DTE 1, BACKWARDATION. Historical realized move mean 5.15% (n=8, max 12.8%). NO WEEKLY -- nearest expiry 17DTE.
Implied move: 8.65%

extra-wide wings, reduced size, if traded at all

Stand aside MARGINAL -- implied is only ~0.7 sigma over the mean and sits BELOW the fattest historical tail (12.8%). The 17DTE expiry additionally straddles CPI and PPI.

SGI

vol_earningsbuy premium

Vol state: DTE 2, BACKWARDATION. Historical realized move mean 5.41% (n=8, max 11.8%). NO WEEKLY -- nearest expiry 17DTE.
Implied move: 9.83%

extra-wide wings, reduced size, if traded at all

Stand aside MARGINAL -- the historical tail (11.8%) nearly matches implied (9.83%). The 17DTE expiry additionally straddles CPI and PPI.

COR

vol_earningsbuy premium

Vol state: DTE 1, BACKWARDATION. Historical realized move mean 6.02% (n=7) with a 17.4% MAX. NO WEEKLY -- nearest expiry 17DTE.
Implied move: 6.88%

Stand aside PASS -- implied is roughly in line with the mean but the tail is FAT (one 17.4% blowup print) and no near-dated defined-risk instrument exists. The binary's tail is not priced enough to structure comfortably at 17DTE.

IRM

vol_earningsbuy premium

Vol state: DTE 1, BACKWARDATION (front-end-iv-ratio 2.25x), weeklies available. Historical realized move mean 6.14% (n=7, max 9.0%), with the last TWO prints at 7.3% and 9.0%.
Implied move: 5.65%

Stand aside PASS -- implied move (5.65%) is UNDER the historical mean (6.14%) and the two most recent prints BOTH exceeded implied. The crush thesis is inverted; there is no edge to sell here.

Nothing in this scan landed at WATCH, STOOD_DOWN or DROP.

laneMOM_LONGBasket watchLongregime fit1.00

final_size=watch. horizon_validated=10, validated_excess=+0.0023. 49-name basket (top 5 by proximity to 52w-high listed in candidates), regime-appropriate and mechanically clean, but out on TWO INDEPENDENT bars. (1) Lane excess +0.23% is below the +0.30% starter bar, and the median is NEGATIVE (-0.91%) -- the mean is tail-carried, so the basket is the only honest expression. (2) risk-sizer.md bars MOM_LONG from starter sizing OUTRIGHT until a DURABLE-N (>=30) forward window re-clears positive excess, to be re-enabled 'via a /calibration-audit recommendation, not ad hoc'. Either bar alone is sufficient. Sector concentration max 24.5% (Industrials), inside the 1/3 cap of gate #7. Funnel: 184 screened -> 49 present in prices.parquet -> 135 FAIL-CLOSED on the coverage gap. LANE MISREPORT: the lane returned 'Sized: 49 names' and gave forward excess as 'not yet resolved', which is not a starter-bar test at all; it is wrong on both grounds above.

SWKJCINEUETNNUE

laneMOM_SHORTStood downShortregime fit0.00

final_size=watch. horizon_validated=10, validated_excess=-0.0008. TRIPLY out: (a) s1_standdown fired decisively (ret5 +4.110% vs +2.5%, 1.61pp margin); (b) invariant-#6 watch-only cap for new starters; (c) baseline excess is NEGATIVE, the ONE recorded exception to 'no lane negative-excess'. CLBK -- VERIFIED FALSE SHORT, correctly excluded by the lane: screener week_52_low $10.60 vs Yahoo $6.2091, true pct_range 81.9% not the screener's 0.9%; its week_52_high is ALSO wrong ($25.83 vs $11.7409), so the row is doubly broken. GME -- SEPARATE LANE MISS: a genuine near-52w-low the lane's screen did not return at all. Yahoo and the screener now AGREE (low 18.545 set 2026-08-03, close 19.21, pct_52w_range 6.96%, in-range and correct), so this is not the stale-field failure mode -- the lane simply dropped it. No live consequence because the lane is stood down, but GME belongs on the watch list and did appear there on 08-03. 101 of 2,870 liquid common stocks still carry an out-of-range pct_52w_range on the trade-date panel.

GMECLBK

laneOI_FADENo name clearedShortregime fit0.30

final_size=watch (no candidate survived). horizon_validated=10, validated_excess=+0.0071. FIFTH consecutive zero-starter session (07-29, 07-30, 07-31, 08-03, 08-04). All 15 top-relative-build candidates die on at least one INDEPENDENT gate: 9 mechanism/persistence>0.85 (HNI 0.999, HXL 0.993, HIG 0.988, INFY 0.984, LZM 0.983, AZN 0.950, EXLS 0.869, plus MRCY 1.260 and APAM 1.068 which exceed 1.0, i.e. prior days were net NEGATIVE); 3 earnings-in-window (SOLV 08-05, MRCY 08-18, LZM); 3 ETP by screener issue_type (MTUM, SPCU, DIG); 6 liquidity (HNI $29.8M ADV, APAM $39.9M, MYGN $3.17 close, LZM $3.74 close, SPCU $14.4M, DIG $2.6M); 1 sustained fundamentals VETO (AVTR, 16 prior mentions, build organic at persistence 0.427 and catalyst-live -- killed ONLY by the veto). ORCHESTRATOR RE-RAN EVERY GATE AND REPRODUCED THE LANE EXACTLY, name for name -- second consecutive honest report from this lane, worth recording alongside the failures. Shorting a decisive V-rebound thrust is a live headwind on top of the mechanics. Contributes ZERO new rows to the maturing 08-03..08-07 post-fix h10 cohort that is the real test of the Cohort-B decay signal (-3.73% mean, hit-base -0.53).

VERDICT: ZERO directional starters across all five lanes; book was FLAT coming in (PLNT closed 2026-08-03 on its h10 time stop at +0.33% realized short excess) and stays FLAT. calls[] is deliberately EMPTY rather than carrying watch/basket placeholders -- per the 2026-08-01 audit proposal #4 (applied at de91fe9), synthetic basket and watch entries belong in lane_status[], where they cannot land as INCONCLUSIVE in a future audit. The vol book is the only actionable output, for the fifth consecutive session.SIZING-MAP CONFLICT CLOSED (open since 2026-07-31, re-derived from scratch on four separate nights). risk-sizer.md:45 reads: validated excess >= +1% & sign-stable -> half (HIGH never auto-fulls); +0.3-1% -> starter; '< +0.3% OR ADVISORY LANE -> watch'. The trailing clause is the one that kept being dropped. An advisory lane NEVER sizes regardless of measured excess -- which is why S4 at +0.51% historical AND +1.16% forward is watch, not starter. It is not a bar S4 narrowly misses; it is structurally out. S2 is likewise advisory. Phase C's 'MEDIUM = +0.3-1%' label describes MEASURED EDGE, not an entitlement to size.LANE_STATUS[S2_dp_revert] (no schema slot -- lane_status.lane admits only MOM_SHORT/MOM_LONG/OI_FADE, so this advisory lane is recorded here): disposition=ADVISORY_WATCH direction=long regime_fit=0.3 candidates=GDS,BNS,LYB,SWK,AMP,SPGI,LPLA,CORZ,CAKE,TT. final_size=watch. horizon_validated=5, validated_excess=+0.0021. 10 mechanically clean names, ALL re-verified by the orchestrator on the TRADE-DATE panel (earnings h5, liquidity, issue_type). Out on the sizing map's 'advisory lane -> watch' clause regardless of measured excess. UPTREND/REBOUND-THRUST is a documented headwind for a mean-reversion long: on 08-03 all 8 candidates landed below starter on exactly this. 35 of 45 FAIL-CLOSED on the prices.parquet coverage gap. GDS reports 2026-08-13 -- clear of h5 by a SINGLE session, would be BLOCKED at h10. LANE HYGIENE DEFECTS: (a) it read issue_type from stock-screener-2026-07-23.parquet, a STALE panel, on an 08-04 scan -- re-run on the trade-date panel showed no ETP leaked and the outcome is unchanged, but the input was wrong; (b) it reported CAKE 'clean' while prior_verdicts.py returns a prior blocking verdict (2026-W31), which is a MOM_LONG basket note rather than an S2 veto, so no practical effect -- but the gate result does not reproduce.LANE_STATUS[S4_pcr_fade] (no schema slot -- lane_status.lane admits only MOM_SHORT/MOM_LONG/OI_FADE, so this advisory lane is recorded here): disposition=ADVISORY_WATCH direction=long regime_fit=1.0 candidates=ICLR,LYV,NVS,TCOM. final_size=watch. horizon_validated=5, validated_excess=+0.0051. 4 viable names (PCR: ICLR 61.97, LYV 10.52, NVS 7.62, TCOM 6.82), all re-verified clean on earnings (h5) and liquidity by the orchestrator. NVS additionally carries the h3 ivrank_chg_5d tilt (+6.86) -- ORTHOGONAL evidence, noted, NEVER summed (invariant #2). 9 of 15 fail-closed on coverage. This is the ONLY lane positive FORWARD (+1.16% mean, hit-base +0.26, DURABLE) as well as historically (+0.51%), and it is fully regime-appropriate -- yet it does NOT size, because retro_harness.py tags the lane structurally size:'advisory' and the sizing map reads '< +0.3% OR ADVISORY LANE -> watch'. POOL NOTE: the lane used the full >=p95 cohort (40 names at a 3.07 cutoff) where retro_harness truncates at LIMIT 15; both are defensible readings of 'top 5% PCR' and NVS/TCOM are legitimately above p95, but the +0.51% prior was measured on the LIMIT-15 definition, so transferring it to names outside that pool is a prior-transfer stretch. LANE MISREPORT: it justified the no-size call with a '+0.7% starter bar' that exists NOWHERE in the repo -- right answer, invented reasoning, which would have failed on a different night. No fundamentals-gate was spawned because no name reached the sizing stage (the gate is specified to run on 'only the names about to be sized').MOM_LONG basket (49, WATCH, h10 closes 2026-08-19): sector max 24.5% Industrials, inside the 1/3 cap; 135 names fail-closed on the prices.parquet coverage gap. Top by proximity: SWK, JCI, NEU, ETN, NUE. Barred from starter sizing until a /calibration-audit re-enable, INDEPENDENT of the +0.23% vs +0.30% miss.MOM_SHORT (STOOD_DOWN): GME -- genuine near-52w-low, Yahoo and screener AGREE (low 18.545 set 08-03, close 19.21, pct_52w_range 6.96%). Kept out by the crash gate and the invariant-#6 cap, NOT by any data defect. The lane failed to return it at all; re-check next session.MOM_SHORT (EXCLUDED, verified false short): CLBK -- screener week_52_low $10.60 vs Yahoo $6.2091 AND week_52_high $25.83 vs $11.7409; true pct_range 81.9%. Do not re-propose without a fresh Yahoo verification.OI_FADE prior-exclusion set carried forward: AVTR (fundamentals VETO, 16 mentions), HXL, INFY, DBX, APAM, LZM, SOLV, SSNC, ALLE, KBR, HRI. PLNT remains EARNINGS-BLOCKED for re-entry (reports 2026-08-06, inside the 08-04..08-19 h10 window).KKR still blocked by the unresolved 2026-07-21/22 EQT/KKR/Advent takeover-rumor tail-block. Its 07-24 h5 earnings block HAS resolved (reported 07-30), but the tail-block was never re-verified and still stands.WRITE-BACK CONFIRMATION: persisted to analyses/scan/2026-08-04/conviction_2026-08-04.json. Book is FLAT with zero open positions and zero new starters, so the write-back carries watch/exclusion state only -- there is nothing to reconcile as a held position next session.ORCHESTRATOR ACTION ITEMS (9) are recorded in analyses/scan/2026-08-04/conviction_2026-08-04.json under orchestrator_action_items -- the envelope schema closes the top level, so they live with the conviction file as in prior sessions. Headline items: prices.parquet coverage gap now SEVENTH consecutive session; earnings_gate.py one-sided-bound bug bit again (LZM, past-dated print); FIFTH lane misreport of the cycle (momentum); next /calibration-audit DUE 2026-08-08 with the post-fix h10 cohort maturing 08-03..08-07.SCHEMA GAP for /calibration-audit: decision_envelope.v2 restricts lane_status.lane to ['MOM_SHORT','MOM_LONG','OI_FADE'], so the two ADVISORY lanes (S2_dp_revert, S4_pcr_fade) have no structured slot and must be recorded as free-text write-back strings. On a night like tonight, where every lane is watch and the advisory lanes carry the only regime-appropriate names, that pushes real content out of the queryable part of the envelope. Propose extending the enum.

Lane glossary (5)

vol_0dte: Vol book — 0DTE variance-risk-premium premium-selling. Non-directional, delta-neutral, 0 directional points.
vol_earnings: Vol book — sells volatility into an earnings IV-crush. Non-directional, delta-neutral, 0 directional points.
MOM_LONG: Momentum, long leg — near-52w-high relative strength. Tail-driven; basket-only, never sized per-name (h10).
MOM_SHORT: Momentum, short leg — near-52w-low relative weakness. Regime-gated against momentum-crash/squeeze rebounds (h10).
OI_FADE: OI-flow fade — fades persistent multi-day net call-OI building; heavy call-OI build precedes underperformance. The most robust lane measured (short, h10).

Regime & Verdict#

Regime: UPTREND/REBOUND-THRUST · ret5 +4.110% · ret10 +3.080% · dd15 −2.520%
Vol-state: VOL-SUPPRESSION — dealers net long gamma on both index proxies. SPY total_gex +$1.104B, spot 772.17 vs zero-gamma 767.91; QQQ total_gex +$623.7M, spot 724.09 vs zero-gamma 711.88. VIX ~16.5 (~48th pct). Script and CLI agreed on the gamma sign for both symbols, independently re-verified by the orchestrator — the zerodte_setup.py sign bug did not bite (third consecutive session; still unfixed).
Breadth: 71.37% green (359 adv / 141 dec of 503).
directional_tradable = TRUE · s1_standdown = TRUE
Bottom line: no directional edge today — ZERO new starters across all five lanes. The vol book is the only actionable output, for the fifth consecutive session. Book was flat coming in and stays flat.

Every lane is out for a sourced, structural reason tonight, not a marginal miss — see the table below.

Directional Book (excess-scored)#

(empty — no name in any lane reached a sizeable tier)

Ticker	Lane	Dir	Horizon	validated_excess	regime_fit	invalidation	gates
—	—	—	—	—	—	—	—

Why each lane is out#

Lane	Disposition	Binding reason (sourced)
MOM_LONG	BASKET_WATCH (49 names)	Two independent bars. (1) Lane excess +0.23% < +0.30% starter bar, and its median is −0.91% — the mean is tail-carried. (2) `risk-sizer.md` bars this lane from starter sizing outright until a DURABLE-N (≥30) forward window re-clears positive excess, re-enabled "via a `/calibration-audit` recommendation, not ad hoc." The lane reported "Sized: 49 names" — that is wrong on both counts.
MOM_SHORT	STOOD_DOWN	`s1_standdown` fired decisively (ret5 +4.110% vs the +2.5% trigger — a 1.61pp margin, versus 0.0139pp yesterday). Plus the invariant-#6 watch-only cap for new starters and a negative baseline (−0.0008). Triply out.
OI_FADE	NO_NAME_CLEARED	Fifth consecutive zero-starter session (07-29→08-04). All 15 candidates die on ≥1 independent gate.
S2_dp_revert	ADVISORY_WATCH (10 names)	Advisory lane → watch by the sizing map, regardless of measured excess. Regime is a documented headwind for a mean-reversion long.
S4_pcr_fade	ADVISORY_WATCH (4 names)	Advisory lane → watch. `retro_harness.py` tags this lane `size:"advisory"` structurally. Note this is the only lane positive forward (+1.16%, hit−base +0.26, DURABLE) as well as historically (+0.51%).

The sizing map (risk-sizer.md:45) is ≥+1% & sign-stable → half · +0.3–1% → starter · < +0.3% **or advisory lane** → watch. This finally resolves the Phase-C/sizing-map conflict that has been carried as an open action item since 07-31: an advisory lane never sizes regardless of its measured excess. S4 at +0.51% is watch because of the advisory clause, not because it misses a bar. The S4 lane's claimed "+0.7% starter bar" does not exist anywhere in the repo — it was invented.

Vol Book (non-directional)#

Delta-neutral, advisory, quoted net of cost. Left tail remains unsampled.

0DTE-VRP premium-selling#

Symbol	Verdict	Gamma (script / CLI)	Gross	Net	Structure
SPY	GO_PREMIUM_SELL_INTRADAY	long / POSITIVE, spot 772.17 > ZG 767.91 ✓	0.223%/d	+0.123%/d	iron fly, wings ≈ ±0.83%
QQQ	GO_PREMIUM_SELL_INTRADAY	long / POSITIVE, spot 724.09 > ZG 711.88 ✓	0.341%/d	+0.241%/d	iron fly, wings ≈ ±1.45%

QQQ's halved size is not a ratio artifact: the full term structure was pulled and the near-week hump is genuine (dte1 30.8% → trough dte6-7 ~22.8% → re-rising to 24.2–25.1% by dte8-10). SPY's dte0 25.0% → dte1 16.0% hump is ordinary same-day pinning, correctly not flagged.

Tempering caveat: the index-level 30-day VRP reads FAIR (−0.002) — IV essentially tracks realized. The 0DTE VRP is a different, shorter measure and can be positive while the 30d is flat, but conviction on the index premium-selling book should be read down accordingly. These are not the same number and should not be quoted as agreement.

Earnings IV-crush (SELL-VOL vs implied move)#

Ticker	DTE	Implied	Hist. realized (mean, n, max)	Verdict
CACI	1	8.19%	0.68% (n=8, max 2.5%)	SELL_VOL — implied 12× mean; no weekly, 17DTE expiry
TTMI	1	18.71%	3.19% (n=7, max 10.6%)	SELL_VOL — ~6× mean, ~2× fattest print
DBX	2	5.92%	1.05% (n=8, max 2.4%)	SELL_VOL — >2× worst print in 2yr
COHR	8	8.54%	4.12% (n=8, max 7.9%)	SELL_VOL — clears historical max
TLN	1	7.60%	1.88% (n=8, max 6.0%)	SELL_VOL, size down — last print 6.0% nearly matched implied
BCO	1	8.65%	5.15% (n=8, max 12.8%)	MARGINAL — below fattest tail; no weekly
SGI	2	9.83%	5.41% (n=8, max 11.8%)	MARGINAL — tail nearly matches implied; no weekly
COR	1	6.88%	6.02% (n=7, max 17.4%)	PASS — fat tail unpriced, no near-dated instrument
IRM	1	5.65%	6.14% (n=7, last two 7.3% & 9.0%)	PASS — implied under historical mean; crush thesis inverted

Best of book: CACI, TTMI, DBX, COHR. All defined-risk, wide wings for the binary tail.

Orchestrator event-risk flag (not raised by the lane): CACI, BCO and SGI have no weekly — the nearest listed expiry is 17DTE (2026-08-21), which straddles the print and both CPI (08-12) and PPI (08-13). That is an unhedged Tier-1 macro exposure stacked on top of the earnings binary, for 12 extra decay days. Size these accordingly or wait; it is a materially different trade from the weekly-expiry names.

The uw insights earnings-play endpoint returned iv_rank=100 for every result — a saturated/broken field. It was correctly not used for ranking.

Watch / Stood-down#

MOM_LONG basket (49, watch): sector max 24.5% (Industrials), inside the ⅓ cap. 135 names fail-closed on the prices.parquet coverage gap. Top by proximity: SWK, JCI, NEU, ETN, NUE.
MOM_SHORT (stood down):
- CLBK — verified FALSE SHORT, correctly excluded. Screener week_52_low $10.60 vs Yahoo $6.2091; true pct_range 81.9%, not the screener's 0.9%. Its week_52_high is wrong too ($25.83 vs $11.7409) — doubly broken. The lane got this one right.
- GME — genuine near-52w-low the lane MISSED. Yahoo and the screener now agree: low 18.545 (set 08-03), close 19.21, pct_52w_range 6.96%, in-range and correct. The lane's screen returned only CLBK. GME is stood down by the crash gate regardless, so there is no live consequence — but it belongs on the watch list and was dropped. 101 of 2,870 liquid common stocks still carry an out-of-range pct_52w_range.
S2 advisory (10, h5 closes 2026-08-12): GDS, BNS, LYB, SWK, AMP, SPGI, LPLA, CORZ, CAKE, TT. 35 of 45 fail-closed on coverage. GDS reports 2026-08-13 — clear of h5 by a single session, blocked at h10.
S4 advisory (4, h5 closes 2026-08-12): ICLR (PCR 61.97), LYV (10.52), NVS (7.62), TCOM (6.82). NVS additionally carries the h3 ivrank_chg_5d tilt (+6.86) — orthogonal, noted, never summed. 9 of 15 fail-closed on coverage.
OI_FADE killed (15): mechanism/persistence >0.85 — HNI 0.999, HIG 0.988, HXL 0.993, INFY 0.984, LZM 0.983, AZN 0.950, EXLS 0.869, MRCY 1.260, APAM 1.068; earnings-in-window — SOLV, MRCY, LZM; ETP — MTUM, SPCU, DIG; liquidity — HNI, MYGN, APAM, LZM, SPCU, DIG; fundamentals VETO — AVTR (16 mentions).

Risk#

Event calendar: CPI 2026-08-12 (Tier-1) and PPI 2026-08-13 (Tier-1) both land inside an h10 window opened today (h10 ends 2026-08-19; h5 ends 2026-08-12, i.e. CPI lands on its final day). No Tier-1 print falls in the first six sessions. Nothing directional is open, so this binds only the vol book.
Correlation clusters: none — zero positions.
Tail caps: none applied; nothing sized.
Hedge note: no directional exposure to hedge. Vol book is delta-neutral by construction.
Advisory divergence tell: price breadth (71.37% green) is much broader than options-flow breadth (41.1% bullish, 2,585 bullish vs 3,701 bearish). Price is running ahead of flow positioning. The repo's defined divergence check (green tape with pct_green < 50) does not fire, so this is noted, not scored.
uw risk market-regime self-labels TRANSITIONAL while its own spy.trend field reads UPTREND. That is the CLI's internal heuristic, not the mechanical ret5/ret10 rule this repo scores on. Not adopted.

Process findings#

Momentum lane misreported twice. It claimed "Sized: 49 names" (barred on two independent grounds) and returned forward excess as "not yet resolved" — which is not a starter-bar test. It also missed GME, a clean in-range near-52w-low name. Fifth lane misreport this cycle; caught only by independent re-run.
S4 invented its bar. The "+0.7% starter bar" appears nowhere in the repo. The real reason it does not size is the advisory-lane clause. Right answer, wrong reasoning — which would have failed on a different night.
S2 read a stale screener (stock-screener-2026-07-23.parquet) for the issue_type ETP cut on an 08-04 scan. Re-run on the trade-date panel: no ETP leaked, outcome unchanged — but the input was wrong.
S2 reported CAKE "clean"; prior_verdicts.py returns a prior blocking verdict (2026-W31). Substance is a MOM_LONG basket note, not an S2 veto, so no practical effect — but the gate result does not reproduce.
OI_FADE reproduced EXACTLY — persistence, liquidity, ETP classification and earnings, name for name. Second consecutive honest report from this lane. Worth recording alongside the failures.
earnings_gate.py one-sided-bound bug bit again: LZM was BLOCKED on a 2026-07-29 print that had already happened. Fails closed, so safe, and LZM died on mechanism anyway. Still must be fixed paired with retro_harness.py:63,76,138, which carries the identical bound and imports nothing.
prices.parquet coverage gap — 7th consecutive session. 783 tickers vs a 2,466-ticker features spine. Cost tonight: 135 MOM_LONG + 35 S2 + 9 S4 fail-closed. Root cause is known (static universe.json frozen 2026-06-28). Deliberately not fixed mid-scan — it shifts the harness baseline panel and requires a regression-gate re-run.
held_book.py fix verified working. It now reports read 1 position(s): 0 open, 1 closed and names the exclusion, instead of the bare "no open positions" that hid a live position on 07-31 and 08-03. 27/27 tests pass.

Aug 4, 2026

Tier mix

Directional book

Vol book

SPY

QQQ

CACI

TTMI

DBX

COHR

TLN

BCO

SGI

COR

IRM

Watch / stood-down

Lane status

Watchlist write-back

Regime & Verdict#

Directional Book (excess-scored)#

Why each lane is out#

Vol Book (non-directional)#

0DTE-VRP premium-selling#

Earnings IV-crush (SELL-VOL vs implied move)#

Watch / Stood-down#

Risk#

Process findings#