TRAI quant day-trading research

← Research · knowledge/reports/2026-10-23-0900-semis-selloff.md

2026-10-23 09:00 ET — Why semis broke on Oct 22, who's in TRAI's buckets, and today's BMO gap risk

Slot: Fri 2026-10-23, 9 AM ET (first hourly report). Status: updated ~9:30 AM ET with research-partner (whiggity) consult — see the consult section. Screening/research, not investment advice.

TRAI data used (pulled 2026-10-23 ~13:00 UTC): /api/latest premarket run 2026-10-22_premarket_2026-10-23T12-53-19-641Z (8:53 AM ET; daily stats = Oct 22 session), /api/events (asOf 2026-10-23), /api/correlations (60-session window, asOf 2026-10-22), /api/regime (2026-10-22), /api/macro (2026-10-22; FRED timed out, Treasury/Yahoo fallbacks used), /api/scorecard (replay 2026-04-23 → 2026-10-21). All under https://rva.guru/trai.

1. Focus summary

  • Tape (Oct 22): SMH −5.99%, QQQ −2.86%, SPY −1.18% (TRAI /api/latest, 8:53 AM ET 2026-10-23). PHLX Semiconductor Index −6.72%; Nasdaq Composite −2.60%; S&P 500 −1.21%; Dow +0.02% (Reuters via Investing.com, 2026-10-22).
  • Regime: VIX 19.98 → label transitional (VIX9D ≤ VIX, mild contango); quant regime score 17.5 = neutral, trend choppy, ADX 13.5 (TRAI /api/latest 8:53 AM ET; /api/regime 2026-10-22). Risk per trade stays 0.25% (VIX < 22).
  • Rotation, not liquidation: 20-day leaders XLP +10.6%, XLU +9.3%, XLE +8.4% vs laggards XLK −5.9%, SMH −12.4% (TRAI /api/macro summary, 2026-10-22). 10y yield 4.96%, −0.14pp on the week (TRAI /api/macro, US Treasury fallback, 2026-10-22); WSJ called the 10y's ~0.2pp drop the sharpest one-day fall in more than three years (WSJ live, 2026-10-22).

2. Why semis fell on Oct 22

Trigger — Lam Research (LRCX) Oct 21 AMC guide. Record Sept-quarter revenue $8.22B and non-GAAP EPS $2.20; Dec-quarter guide revenue $8.60B ± $0.4B, non-GAAP EPS $2.33 ± $0.15, GM 52.5% ± 1pt, op margin 40.5% ± 1pt (Lam IR release, 2026-10-21). Consensus was $2.18 EPS / $8.13B revenue for the quarter and $2.32 / $8.54B for the guide (Investing.com, 2026-10-21) — i.e., a thin beat and an in-line guide. What sold it:

  • Supply constraints: CFO Doug Bettinger: the guide "would have been a bit higher if not for some near-term supply chain constraints"; affected December shipments move into March; he declined to confirm an analyst's $300–400M impact estimate (Yahoo Finance, 2026-10-21/22). Size of the held-back revenue: unknown (not quantified by management).
  • China roll-off: China was 35% of Sept-quarter revenue (26% in June) and is expected to be "meaningfully down" in December, driven by domestic Chinese customers (same source).
  • Guide shape: midpoint implies < 5% sequential growth after three quarters of acceleration (TIKR, 2026-10-21); wide range ($8.2–9.0B) (Investing.com, 2026-10-21).
  • Bullish offsets management gave (ignored on the day): CY2026 systems revenue growth "well over 40%", initial view that 2027 WFE could exceed $200B, advanced packaging > 70% growth in 2026, customer visibility through 2028 (Investing.com call transcript summary, 2026-10-21).
  • Reaction: LRCX gapped −10.1% and closed −11.76% (TRAI /api/events reaction record for 2026-10-21 AMC: gap −10.09%, close-to-close −11.76%, day-1 drift −1.86%); $230.15 close (Investing.com/Reuters table, 2026-10-23). Analyst moves: Evercore ISI PT $355 → $315 (Outperform kept); Goldman Buy, $380 (Investing.com, 2026-10-22).

Amplifiers (same night / same morning):

  1. Meta's capex and financing — Q3 FCF −$2.12B; FY2026 capex range floor raised to $135–145B (from $130–145B) (Yahoo Finance/Stocktwits, 2026-10-22; Meta IR, 2026-10-21); plans to issue new shares to help fund AI investment (Yahoo Finance/Stocktwits, 2026-10-23). META −6.98% (TRAI /api/latest). The market read "more spending + external funding" as AI-funding stress rather than as demand for chips.
  2. AI-funding narrative — Jensen Huang told CNBC "there is a scarcity of capital" for AI infrastructure (Yahoo/Stocktwits, 2026-10-23); Blackstone's Jon Gray: "With AI, you're building capacity, and you're hoping you can keep up with the demand" (Yahoo/Stocktwits, 2026-10-22). CoreWeave near 52-week low ~$55.5 (FinancialMarkets.com, 2026-10-22).
  3. Memory-oversupply worry already in the tape the prior week; Huang said he sees no memory glut by 2029 (NexuAlpha citing CNBC, 2026-10-23). (Secondary source; CNBC original not fetched.)
  4. Oil / geopolitics — Brent +4% to $113.17 on reports of possible US/Israeli action against Iran after the midterms (Yahoo/Stocktwits, 2026-10-22).
  5. Tesla cut its near-term spending target; MarketWatch tied chip selling partly to that (FinancialMarkets.com, 2026-10-22).

Texas Instruments (TXN) was a beat-and-raise, and still fell −1.81%. Q3 revenue $6.06B (+28% y/y, +11% q/q), EPS $2.62; Q4 guide revenue $5.80–6.30B, EPS $2.45–2.79 (TI release, 2026-10-21); Q4 guide above the $5.73B consensus (Reuters via Investing.com, 2026-10-21). End markets: industrial +40%+ y/y, auto mid-teens, data center ~+95% y/y (TI Q3 call transcript, 2026-10-21). IBD framed the day as "chip stocks tumble despite solid earnings" (IBD, 2026-10-22). Read: the selloff was about AI-capex funding/positioning, not fundamentals in analog/industrial — TXN's small move vs LRCX's is consistent with that.

Breadth of the damage — "one trade, many exposures": GE Vernova −11.5%, Bloom −10.7%, Vistra −10.5%, Lumentum −10.3%, CoreWeave −10.2% intraday; SanDisk −8.7%, Seagate −8.1%, WDC −6.8%; Russell 2000 only −0.2% and XLV +1.2% (FinancialMarkets.com, 2026-10-22). Closing moves in TRAI's universe (TRAI /api/latest): LRCX −11.76%, GEV −10.64%, VST −9.81%, KLAC −9.55%, MRVL −8.70%, AMAT −8.16%, AMD −8.10%, INTC −8.04%, STX −7.86%, VRT −7.81%, ARM −7.28%, ANET −7.19%, ASML −6.93%, DELL −6.78%, WDC −6.43%, MU −6.26%, CEG −5.74%, SMCI −5.20%, ADI −4.98%, TSM −4.52%, NVDA −4.40%, AVGO −4.31%, QCOM −2.88%, TXN −1.81%.

Overnight / premarket Oct 23: chips bouncing modestly — Intel +1.1%, AMD +0.94%, Nvidia +0.51% premarket (TipRanks, 2026-10-23); SOXX up < 1%, SK Hynix +4% after ending plans for a US IPO of Solidigm; 10y back to 5.00%; Brent ~$115, highest since early May (Investopedia via Yahoo, 2026-10-23). TRAI indicated gaps at 8:53 ET agree: KLAC +0.28%, MU +0.88%, ANET +0.54%, ADI +0.32%, INTC +0.28%; NVDA −0.22%, AVGO −0.57% (TRAI /api/latest).

3. Semis / AI-hardware names in TRAI's buckets (8:53 AM ET run)

BucketNameOct 22Indicated gap 8:53RelVolResidual zNext earnings (TRAI /api/events)
Short momentumLRCX−11.76%−0.15%2.63−2.32reported 10-21 AMC
Short momentumKLAC−9.55%+0.28%2.20−2.4110-28 AMC (elevated)
Short momentumAMAT−8.16%+0.15%2.02−1.5311-12 AMC
Short momentumADI−4.98%+0.32%2.20−1.9011-24 BMO
Short momentumTSM−4.52%+0.08%1.51−1.472027-01-14 (est.)
Short momentumTXN−1.81%−0.30%2.51−0.21reported 10-21 AMC
Short momentumANET (Hardware)−7.19%+0.54%1.54−1.5911-03 AMC
Short momentumVRT (Hardware)−7.81%−0.76%1.63−1.1510-28 BMO (elevated)
Short momentumGEV (Industrials, AI-power)−10.64%−0.38%2.67−3.3310-28 BMO (elevated)
Short momentumVST (Utilities, AI-power)−9.81%−0.59%1.69−2.8411-06 BMO
Mean reversion (fade-long)ASML−6.93%−0.01%1.38−2.46reported 10-14 BMO
Mean reversion (fade-long)AMD−8.10%+0.07%1.43−1.4111-03 AMC
Mean reversion (fade-long)NVDA−4.40%−0.22%1.21−1.1611-17 (timing unknown)
Mean reversion (fade-long)INTC−8.04%+0.28%1.35−1.1510-29 AMC (elevated)
Mean reversion (fade-long)MU−6.26%+0.88%1.21−0.6912-23 (est.)

Source: TRAI /api/latest (8:53 AM ET 2026-10-23) and /api/events (asOf 2026-10-23). The engine's own cluster flags: "Semis ×6" (short), "Hardware ×2" (short), "Semis ×5" (mean reversion).

4. Sector linkages — who moves together and why

  • One giant correlation cluster. TRAI's 60-day clustering puts 24 names in cluster 1 (avg ρ 0.588): ON, MCHP, TXN, ADI, CAT, ANET, VRT, GEV, ETN, AVGO, MU, TSM, ASML, KLAC, AMAT, LRCX, MRVL, ARM, AMD, INTC, NVDA, ORCL, SMCI, DELL (TRAI /api/correlations, asOf 2026-10-22). AI-power/grid (GEV, ETN, VRT) and CAT trade with semis, not with Industrials. FinancialMarkets.com describes the common thread as "all have been bought as beneficiaries of data-center construction" (2026-10-22). VST/CEG form their own tight pair (ρ 0.836) but sold off with the theme on Oct 22.
  • Tightest sub-links (top pairs): AMAT–LRCX ρ 0.914, LRCX–KLAC 0.879, AMAT–KLAC 0.866, KLAC–ASML 0.861, LRCX–ASML 0.843 (equipment / WFE complex — Lam's guide is a read-through for all four); TXN–ADI 0.875 (analog); WDC–STX 0.828 (storage); CEG–VST 0.836 (AI-power IPPs) (TRAI /api/correlations).
  • Sector-level: Semis intra-sector ρ 0.59; Semis vs Consumer Staples −0.35, vs Telecom −0.27, vs Hardware +0.30, vs Crypto/high-beta +0.30 (TRAI /api/correlations sectorMatrix). The Oct 22 mirror image (staples/health care up, semis down) is the normal sign of that relationship, not an anomaly.
  • Next shared catalysts for the cluster: KLAC 10-28 AMC, VRT & GEV 10-28 BMO, INTC 10-29 AMC (TRAI /api/events); MSFT and GOOGL 10-28 AMC, AMZN 10-29 AMC (TRAI /api/events) — FinancialMarkets.com notes hyperscaler capex plans "will do more to settle the debate than any single supplier's guidance" (2026-10-22).

5. Typical post-catalyst reaction (TRAI /api/events stats, last 4 reports each unless noted)

  • Universe aggregate (n = 622 events): median |gap| 3.76%, avg |gap| 1.52 ATR, gap fill rate 38.7%, continuation 49.2%, avg day-1 drift in gap direction −0.16%, avg drift days 2–5 in gap direction +0.02%. Big gaps (n = 172, median |gap| 8.75%): fill rate only 2.3% (TRAI /api/events aggregate, asOf 2026-10-23). → After a big earnings gap the gap essentially never fills same day, and there is no average post-earnings drift to lean on.
  • LRCX: avg |gap| 9.73% (1.34 ATR), fill 50%, continuation 75%. KLAC: avg |gap| 5.66% (1.28 ATR), fill 50%, continuation 75%. AMAT: 7.30%, fill 50%, continuation 0%, day-1 in gap dir −4.29%. INTC: 10.46% (1.99 ATR). MU: 10.09%, continuation 0%. NVDA: 3.14% (0.98 ATR). VRT 8.08%; ANET 8.87% (2.0 ATR) (TRAI /api/events tickers).

6. Secondary: today's BMO reporters and the big long-momentum movers

American Express (AXP) — reported BMO today (gapRisk `imminent`). Q3 revenue +9% to $20.1B, billed business +8% to $454B; FY2026 adjusted EPS guide raised to $17.60–17.90 (midpoint above $17.63 LSEG consensus); T&E spend +10% (airline +13%, partly ticket-price inflation per the CFO); plans to eliminate 5% of global workforce in a restructuring (Reuters via Investing.com, 2026-10-23 08:55 ET update). GAAP EPS $5.42, adjusted $4.68 (search summary of AXP 8-K via StockTitan, 2026-10-23 — not yet verified against the EDGAR original). Premarket direction is conflicting across sources: "edged higher" early (Investopedia via Yahoo, ~6:30 ET); TRAI's 8:53 ET indicated gap −2.43% (TRAI /api/latest). History: avg |gap| 2.1% (0.81 ATR), fill rate 0%, continuation 75%, avg day-1 in gap dir +2.29% (TRAI /api/events, n = 4). Today's −2.43% ≈ 1.37 ATR (ATR 1.78%) — larger than its usual earnings gap.

SLB — reported BMO today (gapRisk `imminent`). Q3 revenue $9.45B (+5% q/q, +6% y/y), GAAP EPS $0.49, EPS ex-charges $0.64, FCF $828M; dividend $0.295 (SLB IR, 2026-10-23); beat ~$0.62 / $9.27B consensus (Investing.com, 2026-10-23). Middle East recovery commentary: from a search summary only — unverified. TRAI 8:53 ET indicated gap +2.00% (ATR 2.85% → ~0.7 ATR); history avg |gap| 2.21%, fill rate 75%, continuation 50% (TRAI /api/events). SLB–HAL ρ 0.798 (TRAI /api/correlations) — HAL indicated +1.24%.

Other BMO/AMC names in the window: only AXP and SLB are imminent for today; VZ is imminent for Mon 10-26 BMO (TRAI /api/events). Data gap: Western Digital reported Oct 22 after the close (beat; FQ2 guide revenue $5.0–5.2B, non-GAAP EPS $5.60–6.00) (WD release, 2026-10-22) and fell ~2% premarket (Yahoo/Stocktwits, 2026-10-23), but TRAI /api/events shows no upcoming date for WDC (next = null) so it was never flagged. Same for NEM (reported, fell premarket per Investopedia via Yahoo, 2026-10-23; next = null in TRAI). TRAI indicated: WDC −2.33%, NEM −1.84%.

Blackstone (BX) +8.38% (long momentum, residual z 6.19). Q3 (BMO Oct 22): distributable EPS $1.32 vs $1.26 consensus (down from $1.52 y/y); inflows $63.3B; FRE $1.6B (+6%); net realizations −34% to $332M; GAAP revenue $3.06B, below estimates; management expects Q4 realizations to "rise significantly" and double-digit base-fee growth in 2027 (Yahoo Finance/TIKR, 2026-10-22); AUM $1.37T; stock had fallen nearly 30% YTD into the print (Reuters via Investing.com, 2026-10-22); AI-infra/data-center marks helped (infrastructure +6.4% in Q3) (MarketWatch, 2026-10-22). It gapped −2.05% then ran +10.65% open→close (TRAI /api/events). Catalyst = distributable-EPS/fundraising beat in a name down ~30% YTD (interpretation: a positioning reversal — not sourced); bank ETFs also rose as long-term Treasury yields fell that day (FinancialMarkets.com, 2026-10-22). Indicated −1.67% at 8:53 ET.

Spotify (SPOT) +11.10% (long momentum, residual z 4.32). Q3 (BMO Oct 22): Premium subs 306M (+9% y/y, 1M above guidance), MAUs 788M, revenue €5.0B (+16%), GM 33.0% (above 32.9% guide), op income €672M (Spotify newsroom, 2026-10-22; 6-K Ex. 99.1, SEC EDGAR, 2026-10-22). Q4 guide: MAUs 818M, subs 316M, revenue €5.2B, GM 33.8%, op income €820M (same 6-K). TRAI recorded EPS surprise −12.7% vs estimate yet a +6.84% gap and +3.99% day-1 drift (TRAI /api/events) — the market traded subs/margin/guide, not EPS. Beta60 −0.49 (TRAI /api/latest): SPOT is a hedge-like name vs the tech tape. Indicated −0.97% at 8:53 ET.

7. Macro today (Fri Oct 23)

  • 9:45 AM ET S&P Global flash US manufacturing & services PMIs (Oct) (TradingKey premarket, 2026-10-23).
  • 10:00 AM ET University of Michigan consumer sentiment, final Oct (FinanceCalendar, event page). New home sales is not today (scheduled Tue Oct 27) per Scotiabank's Oct 2026 release calendar (via search summary; page not fetched).
  • TRAI /api/macro has no calendar feed; its dashboard state (2026-10-22): 10y 4.96%, 2y 4.65%, 10y–2y +0.31pp, 10y real 2.64%, breakeven 2.32%, DXY 101.56, HYG/IEF −2.49 pts over ~1 month (credit underperforming). FRED was unreachable on the last run (fallbacks used).
  • Oil: Brent ~$115, WTI ~$101 premarket (Investopedia via Yahoo, 2026-10-23).
  • PMI/UMich actual prints: unknown at time of writing (released after this report).

8. TRAI scorecard context for today's buckets (replay 2026-04-23 → 2026-10-21, next-open→close, net of costs)

  • Short momentum overall: n = 412, hit 43.5%, mean −0.46%/trade, t = −3.72. In VIX-transitional days: n = 329, hit 41.3%, mean −0.51%, t = −3.66. In trend-choppy days: n = 173, hit 38.7%, mean −0.60%, t = −2.75 (TRAI /api/scorecard byRegime). Today is transitional + choppy — the bucket's worst historical cell.
  • Mean reversion fade-long (after big down day): n = 212, hit 54.3%, mean +0.165%, t = 1.19; transitional n = 167, +0.18%, t = 1.18; quant-neutral n = 73, hit 57.5%, +0.37%, t = 1.49 (TRAI /api/scorecard). Positive but not statistically significant.
  • Long momentum: n = 385, mean +0.13%, t = 0.82; quant-neutral n = 65, hit 41.5%, mean −0.31%, t = −0.9 (TRAI /api/scorecard).
  • Caveats from TRAI's own assumptions: daily-bar replay approximates VWAP; survivorship bias (today's universe). Live scorecard: 0 scored picks so far (TRAI /api/scorecard live).

9. ACTIONABLE for today (Fri 2026-10-23) — screening notes, not advice

  1. Semis short-momentum signals face (a) a small premarket bounce and (b) a historically losing cell. The 6-name semis short cluster plus ANET/VRT/GEV/VST are mostly indicated flat-to-up (KLAC +0.28%, ANET +0.54%, ADI +0.32%), and TRAI's replay shows short momentum at −0.51%/trade (t −3.66) in transitional regimes and −0.60% (t −2.75) in choppy trend. If the engine's rules allow discretion, the evidence argues for skipping or downsizing the short-momentum semis/AI-power names today; at minimum only one per cluster (see 2).
  2. Cluster cap is under-counting. GEV and VST are labeled Industrials/Utilities, so the engine's sector clusterFlag treats them as separate from "Semis ×6" — but GEV, VRT, ANET, CAT, ETN sit in the same 60-day correlation cluster as the semis (TRAI /api/correlations cluster 1, avg ρ 0.588). The "max 1 per correlated sector cluster" rule should treat LRCX/KLAC/AMAT/ADI/TSM/TXN/ANET/VRT/GEV (and ASML/AMD/NVDA/INTC/MU on the fade-long side) as one exposure. Note a fade-long on, say, AMD and a short on KLAC are offsetting bets on the same factor.
  3. Earnings inside the 5-day window for short-bucket names: KLAC (10-28 AMC), VRT and GEV (10-28 BMO), INTC (10-29 AMC, fade-long bucket). No overnight holds per TRAI rules, so no direct gap exposure today, but the cluster's next leg is hyperscaler capex (MSFT/GOOGL 10-28, AMZN 10-29).
  4. AXP gap larger than its norm, and it shares a cluster with BX. AXP indicated −2.43% (≈1.37 ATR vs a 0.81-ATR historical average; fill rate 0/4, continuation 3/4). AXP is not in a bucket, but it sits in TRAI correlation cluster 8 with BX, KKR, BLK, COF, BA (avg ρ 0.582). The BX long-momentum candidate (already indicated −1.67%) therefore carries same-day read-through risk from AXP's reaction — and BX's own earnings-day-2 drift has no support in TRAI's stats (aggregate days 2–5 drift ≈ +0.02%).
  5. SLB +2.0% indicated (≈0.7 ATR, inside its normal 2.2% earnings gap; fill rate 3/4 historically) — SLB/HAL not in buckets; oil-driven tape (Brent ~$115) supports energy, but SLB's history suggests the gap often fills.
  6. Event-calendar gap: WDC and NEM reported but were not flagged (TRAI next = null). WDC indicated −2.33% and is in a 0.828-ρ pair with STX (STX reports 10-27 AMC, elevated). Neither is in a bucket today, but fix the calendar before Monday.
  7. Macro timing collides with the entry window: flash PMIs at 9:45 and UMich at 10:00 ET — exactly when TRAI's "no entries before 10:00" window ends. Consider waiting for the 10:00 print to clear before the first entry.

10. Unknowns / to verify

  • Size of Lam's supply-constrained revenue (management declined to quantify).
  • AXP prior FY guide range and GAAP/adjusted EPS — verify against the EDGAR 8-K original (only a StockTitan summary was seen).
  • SLB's Middle East commentary and guidance — read the full release/call.
  • CNBC original for Huang's memory comments (seen only second-hand).
  • Today's PMI/UMich actuals.

Research partner (whiggity) consult

_Consulted 2026-10-23 ~9:15 ET (Monica research bot "whiggity", new chat; it errored twice before answering). Treated as a second researcher, not a source of truth: every whiggity claim below is tagged [whiggity] and kept as fact only where I could fetch a public source or check TRAI data myself. Everything else is an open question. Its BX/SPOT/AXP/SLB section was generic and, by its own admission, unsourced, so it is not used._

Divergence: whiggity did not mention Lam's guide

whiggity explained Oct 22 as a macro/positioning story (AI-ROI doubts, failed technical breakout, CTA de-risking, flight to safety) and did not mention LRCX's Oct 21 guide, which §2 treats as the trigger. One of whiggity's own sources disagrees with it: Investopedia's Oct 22 live blog says LRCX was "the worst-performing stock in the S&P 500 ahead of the opening bell, with shares down nearly 9% after the company posted an underwhelming outlook that was just in line with consensus projections," and that LRCX closed with "a 12% pullback" (Investopedia, 2026-10-22). Read: both accounts hold up and fit together. LRCX + META set off the move on the day, and the macro/positioning backdrop whiggity describes is why the reaction was so large.

Verified (kept, with sources)

  1. [whiggity] Flight to safety, yields plunged, metals also sold off. VERIFIED. 10y yield ~4.97% late afternoon, −16 bp vs Wednesday, "as investors fled to the relative safety of government bonds"; gold futures −2.6% to $4,040; SOXX ~−6.5%; Roundhill Memory ETF (DRAM) −7% (Investopedia, 2026-10-22). Verified Investing also noted gold and silver falling with yields, an unusual divergence (Verified Investing, 2026-10-22). This matches TRAI /api/macro (10y 4.96%) and the WSJ line in §1.
  2. [whiggity] AI revenue/capex ROI skepticism as backdrop. VERIFIED as background, not as Oct 22's trigger. Reuters (Oct 15): US chip stocks were on track to lose $2T in market value over a 7-session selloff "driven by rising bets of an interest-rate hike and concerns over the cost of the AI buildout"; SOX −3.3% that day after a $1.64T loss over six sessions (Reuters via Investing.com, 2026-10-15). The reuters.com link whiggity gave returns 401 to our fetchers, so the same story was read on Investing.com. whiggity's "Nasdaq" link is a paid PRNewswire release from Stock Preachers hosted on nasdaq.com, not Nasdaq reporting. It says semis came off October highs after an FT report that OpenAI's annualized revenue was nearing $50B vs ~$70B in media reports (Stock Preachers via Nasdaq.com, 2026-10-10). Low-quality source; the FT original was not checked. The Leverage Shares link is dated Aug 24, 2026 and covers the summer correction, not October (Leverage Shares, 2026-08-24). It is context only.
  3. [whiggity] CTA deleveraging. PARTLY VERIFIED, not as a same-day driver. JPMorgan (note published 06:55 ET Oct 22, before the session) said it "suspects that hedge funds and momentum-based investors such as CTAs have largely unwound their elevated long semiconductor exposures," noted semis take ~1/5 of global equity risk capital, and said positioning resets "could start to limit further downside" (Investing.com, 2026-10-22). That supports earlier CTA selling. It cuts against CTA selling as a big new driver on Oct 22. No source I found measures CTA flows on Oct 22 itself.
  4. [whiggity] Failed Nasdaq trendline breakout; counter-evidence of SMH at Fibonacci support, Nasdaq channel support near 25,036, late-session bounce, oversold. VERIFIED that the source says this. These are one technician's levels, not facts about the market. Verified Investing: the Nasdaq failed to hold a breakout above a declining trendline; next support in its long-term rising channel ~25,036; SMH is testing a 50% Fibonacci retracement from the March 2026 low plus a retest of a broken trendline, with daily RSI oversold; there was a late-session bounce that "did not repair the technical damage." S&P levels cited were 760.40 then 748.45 (on its "tracking chart", which looks like an ETF scale) (Verified Investing, 2026-10-22). Caveat: the same page puts the 10y "beneath the 5.21% area," which conflicts with ~4.97% (Investopedia; TRAI /api/macro). Treat its yield levels as unreliable.
  5. [whiggity] Macro dates late October. VERIFIED, plus one item whiggity missed. Q3 advance GDP and September Personal Income & Outlays (PCE) are on Thu Oct 29, 8:30 ET (BEA release schedule). Q3 ECI is on Fri Oct 30, 8:30 ET (BLS ECI schedule). Missed by whiggity: the FOMC meets Oct 27–28 (Federal Reserve FOMC calendar), and the decision lands Wed Oct 28, the same day as MSFT/GOOGL/KLAC AMC and VRT/GEV BMO. Given Reuters' "rising bets of an interest-rate hike" (item 2), this may be the biggest single risk in the window.
  6. [whiggity] Earnings catalysts next week. PARTLY CORRECT (checked against TRAI /api/events, asOf 2026-10-23). Correct: MSFT and GOOGL 10-28 AMC, AMZN 10-29 AMC, KLAC 10-28 AMC, INTC 10-29 AMC. Wrong: META already reported (10-21 AMC, see §2). ORCL is 12-14, not this window. AMD is 11-03 AMC and ON is 11-02 AMC, so they report the week after next. NXPI is not in TRAI's universe, so its date was not checked. whiggity also left out STX 10-27 AMC, VRT/GEV 10-28 BMO and CAT 10-29 BMO, all of which are in TRAI cluster 1 or the storage pair.

Checked against TRAI data

  • [whiggity] Analog (TXN, ADI, NXPI) has lower beta to AI capex. TRUE ONLY FOR TXN. TRAI /api/latest beta60 (vs market): TXN 1.22, ADI 1.65, ON 2.54, MCHP 2.68 (compare NVDA 1.94, LRCX 3.13, AMD 3.18). Average 60-day ρ to the core-semis basket (NVDA, AMD, AVGO, MU, TSM, ASML, AMAT, LRCX, KLAC; TRAI /api/correlations asOf 2026-10-22): TXN 0.54, MCHP 0.50, ON 0.51, ADI 0.65 (same as MU/AMD). Oct 22: TXN −1.81%, ADI −4.98%, ON −4.82%, MCHP −3.41%. TXN is the defensive one; ADI behaved like an AI name. All five analog names are in TRAI cluster 1. NXPI is not in the universe.
  • [whiggity] Co-movement map. MOSTLY MATCHES, except the hyperscalers. In TRAI cluster 1 (24 names, avg ρ 0.588): TSM; ASML/AMAT/LRCX/KLAC; MU; AVGO/MRVL/ANET; SMCI/DELL/VRT/ETN/GEV; ORCL. WDC/STX form their own pair (cluster 19, ρ 0.828) but sit at ρ ~0.50–0.56 to core semis. Not in TRAI's universe: ASX, COHR, LITE, HPE, NXPI. Hyperscalers do not co-move on daily returns: average ρ to core semis is MSFT 0.33, AMZN 0.17, GOOGL 0.07, META 0.05. They sit in their own clusters (MSFT 36, META 31, GOOGL+AMZN 20). The link is through catalysts (their capex guides move the supply chain, as META did on Oct 22). Return correlation doesn't capture it, so correlation-based gap-risk flags (S-004) will miss it. See S-010.

Not verified (open questions)

  • [whiggity] INTC −5.24% on Oct 22 from expanded U.S. export controls on robotics/automation chips (Tickeron, 2026-10-22). The −5.24% / $88.09 figure conflicts with TRAI's close of −8.04% and was probably an intraday snapshot. I found no primary BIS source for a new robotics-chip export rule; a search turned up only a draft AI-chip licensing proposal and a separate Section 232 robotics import probe. Tickeron's page is auto-generated. Unverified, don't use.
  • [whiggity] Today's weekly opex/gamma effect at the open. Not sourced. Oct 23 is a regular weekly expiration (not the monthly), and I have no dealer-gamma data. Open.
  • [whiggity] "Oversold" bounce case. Supported only by the technician's RSI comment (item 4) and the small premarket bounce in §2. No positioning data beyond the JPMorgan note.

whiggity's testable hypotheses (status)

  1. [whiggity] Next-day drift after SMH < −5% with VIX ≥ 19 and the 10y falling > 5 bp. Oct 22 qualifies: SMH −5.99%, VIX 19.98, 10y −16 bp. Quick check (mine, Yahoo daily bars SMH/^VIX/^TNX, 2010-01 → 2026-10-21, not TRAI's engine): all SMH < −5% days n = 38: next-day gap mean +0.34% (hit 63%), open→close mean +0.33% (median −0.12%, hit 50%). With the VIX/10y filter n = 16: next-day open→close mean +1.00%, median +0.49%, hit 56%, t ≈ 1.2. Excluding the three March-2020 days: n = 13, mean +0.42%, median −0.40%. Not significant. Any edge sits in the overnight gap, which TRAI can't trade (no overnight holds), and in crisis outliers. Status: NEEDS DATA (intraday, sector names, not just SMH). See S-011.
  2. [whiggity] 10:00 VWAP + reclaim of the first-15-minute high as a long filter, vs fading broken trendlines. Can't be tested without intraday bars; TRAI's replay uses a daily-bar VWAP proxy (§8). Open.
  3. [whiggity] Lead-lag: "low-beta" customers/power (MSFT, META, VRT) lead the semis bounce on 5-minute relative strength after a >3σ sector shock. The premise is wrong in TRAI data: VRT beta60 3.34 and MSFT 2.13 are not low beta (META 1.12 is). Intraday data needed. Open; I'd reframe it as "names with the lowest ρ to core semis" before testing.
  4. [whiggity] Residual reversion: names with the most negative beta-adjusted residuals and no news beat the next day open→close. TRAI already computes residualZ (e.g., GEV −3.33, ASML −2.46, KLAC −2.41) and can replay this on daily bars. Today's candidates mostly have news (LRCX, META) or are short-momentum names. Status: testable in TRAI's scorecard. See S-012.

Net changes to the actionable list (§9)

  • Add: FOMC Oct 27–28 (decision Wed Oct 28) on top of the 10-28 earnings cluster. GDP/PCE Thu Oct 29 8:30, ECI Fri Oct 30 8:30. No change for today, since TRAI holds nothing overnight, but next week's semis/AI-power signals carry stacked event risk.
  • Refine §9.1: the JPMorgan positioning note and the technician's support levels argue against new short-momentum entries today. That backs the existing "skip/downsize" note; it is not a signal to go long.
  • Refine §9.2: keep TXN separate from the rest of the analog group (lowest beta and ρ); ADI trades like the AI names.
  • No changes from whiggity's BX/SPOT/AXP/SLB notes (unsourced).

Sources (accessed 2026-10-23)