House Delta — 2026-08-21

desk: house-delta date: 2026-08-21 forecasts: delta: p_trade: 0.30 direction: long conviction: low

Desk Plan — House Delta — 2026-08-21

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-08-21-pm.md PM: DELTA

Shared Market Read

• Event risk today: LOW (briefing). No Tier-1 within ~2 days. PCE (Aug 26) and NVDA earnings are 5 days out; today is the quiet Friday before the weekend and the week's real pivots. Today's own macro (flash PMI 9:45, state employment 10 AM, Baker Hughes 1 PM) is non-Tier-1 and modest; light earnings Friday, no major S&P reporters. The real risks are the live coin-flip catalysts, not the calendar: (1) oil/Hormuz — Trump's Iran economic-isolation package with details expected MONDAY, a genuine weekend-gap risk sitting just ahead of the session; (2) rates — 30-yr ending ~5.24% near 2007 highs as the buyback relief faded; (3) consumer/Dow — WMT's comp-miss aftermath (worst day in years). None in DELTA's hard-skip set (FOMC decision/CPI/NFP/PCE; VIX ≥ 30) → hard-skip check CLEAR, firm-wide GO, this desk CAN trade.
• Session character expected: LOW-VIX, near-record, sub-ATR, two-tier, direction UNKNOWN. Per the four-session-confirmed discipline, do NOT pre-assign direction — the honest base case is contained, two-tier, direction unknown, with the overnight fuel the live variable. Yesterday's pullback (SPY -0.84%, QQQ -0.71%, both sub-ATR/contained) was a dip, not a break — uptrend intact, both indices above SMA20.
• VIX regime: 15.70 — LOW (<18), DELTA's best backtest regime (dev PF 2.59 in VIX < 20 vs. 1.30 in 20–30); the ≥ 30 hard-skip barrier is far away. Supportive, licenses trading.
• Key levels (DELTA, prior 08-14 closes, last verified anchor): NVDA 225.16 → +1.5% ≈ 228.5 = clean NVDA→ASML LONG trigger; −1.5% ≈ 221.8 = disabled SHORT. AMD 514.39 → +1.5% ≈ 522.1 = AMD→TSM LONG trigger (the degraded cell); −1.5% ≈ 506.7 = disabled SHORT. Laggards ASML ~1,844 (indicative), TSM 426.35. Runner measures actual first-hour returns — authoritative; the cash open confirms where NVDA/AMD actually gap.
• Macro backdrop: The memory/AI complex is GREEN pre-market for the first time in several sessions (MU, SNDK ~+2%, reversing yesterday's red US pre-market) — a genuine up-tilt on exactly the leg I trade, tracking a firmer Europe and a mixed-but-higher Asia. The tape is otherwise flat-to-slightly-higher with a persistent oil bid (Brent ~$93) and a consumer/Dow leg still fragile on WMT.

PRICE-VERIFICATION LIMIT (stated per dispatch): the pm briefing (Alpaca/IEX-verified) carries none of DELTA's four instruments and no consolidated pre-market prints for them. I carry forward the last verified closes (08-14: NVDA 225.16, AMD 514.39, TSM 426.35; ASML ~1,844 indicative) as indicative anchors only — today's cash open and the runner's 09:30–09:55 measurement are the only authoritative signal inputs.

DELTA — NVDA→ASML, AMD→TSM (LONG-only, Phase 3)

Forecast: p_trade 0.30 | direction long | conviction low

Verdict: the up-branch is the most live it has been in several sessions — the US pre-market itself is green on my complex, not just Asia. This is the key difference from yesterday (08-20): then the up-tilt came only from overnight Asia while US memory was flat-to-red — a divergence I could not resolve; today US memory is green pre-market (MU, SNDK ~+2%), removing that divergence on the complex I actually trade. If NVDA or AMD catches that bid and holds +1.5% UP in the first hour, a LONG is a real, non-fade candidate. The tape is otherwise a two-tier coin-flip (oil/rates/consumer), and the direction discipline says do not pre-assign — but the overnight fuel is, unusually, aligned with the up-side of my exact sector.

• Base case (sub-threshold no_signal): NVDA/AMD open green but their first-hour |return| stays under ±1.5% (fading-bid tape) → no trade. This remains my base case — the sub-ATR LOW-VIX tape has pinned the leaders inside ±1.5% for five straight sessions (08-17/08-18/08-20 no_signal; 08-19/08-15 disabled SHORT), and the memory complex has "flip-flopped direction all week," fading its green pre-market across the last four sessions. Plausible, even likely.
• Second branch (a LONG actually fires — genuinely live today): for a LONG I need NVDA or AMD to clear +1.5% UP in the first hour. More likely than on any recent session because the US pre-market itself is bidding my complex — the up-move can develop at the cash open rather than needing Asia to transmit. If it fires, the operative cell question is the one I've been carrying all month: NVDA → clean NVDA→ASML cell (2W/0L live) vs. AMD → degraded AMD→TSM cell (n=8, WR 25%, PF 0.191, three straight losses).
• Third branch (down — disabled SHORT shadow): if the green pre-market fades and NVDA/AMD slide to −1.5% in the first hour, that's a disabled SHORT (structural no-trade under LONG-only), logged to the shadow book (AMD_TSM shadow n=9→n=10, or NVDA_ASML's next live disabled-SHORT datapoint). A real branch — the complex has flip-flopped down all week — but somewhat less likely today given the green pre-market.

Calibration math (day-conditional): P(signal) ≈ 0.55 — the green US pre-market on my exact complex is the strongest single-sector signal in a while and brings real first-hour volatility to the leaders; but the flip-flop/fade pattern across the last four sessions keeps it shy of 0.60+. P(LONG | signal) ≈ 0.55 — two-way coin-flip, but the green pre-market tilts the resolution up more than on recent sessions (where the divergence forced 0.44). Net tradable-signal probability ≈ 0.30. Direction long by construction (LONG-only book — the only direction that counts as a real trade). Conviction low: the most probable single outcome is still a sub-threshold no_signal, and the LONG branch — if it fires — most likely hands me the chronically-degraded AMD→TSM cell unless NVDA leads.

Sizing vs. event risk: standard 0.75%, single position — no change. No same-day Tier-1 hard skip (event risk LOW; no FOMC/CPI/NFP/PCE). Today's macro (flash PMI 9:45, state employment 10 AM) prints before/at the 09:30–09:55 measurement and 10:00 entry and is absorbed into the open; no scheduled event sits inside the 10:00→15:30 holding window. The weekend-gap risk (Iran measures due Monday) is a real tail, but it is not a same-day DELTA Tier-1 and not a holding-window stop-out — per-trade risk control (0.75% with a 2.0×ATR stop) and hard flat at 15:30 remove all overnight. Entry selectivity is the only non-standard dimension, and it is cell-quality-driven, not event-driven: the clean NVDA→ASML cell (2W/0L live) is the only LONG I'd want at all; the degraded AMD→TSM cell (n=8, WR 25%, PF 0.191) firing on a green-bid flip would be a double-conviction-negative (lossy cell + a fade-tape risk).

Key levels: NVDA +1.5% ≈ 228.5 → clean NVDA→ASML LONG (2W/0L live, the cell that makes money — the only LONG worth taking); NVDA −1.5% ≈ 221.8 → disabled SHORT (shadow). AMD +1.5% ≈ 522.1 → degraded AMD→TSM LONG (n=8, WR 25%, PF 0.191, three straight losses — the chronic loser; the least desirable entry today, doubly so on a possible fade); AMD −1.5% ≈ 506.7 → disabled SHORT (shadow). Laggard cash-open gaps are the second read: ASML (~1,844) / TSM (426.35) gapping with their leader closes the lead-lag spread (late entry); a laggard not gapping is the only genuine lead-lag shape left. But the first read is direction, and today it tilts up on the overnight fuel while the tape itself stays a coin-flip — the cash open and the runner's 09:30–09:55 measurement decide whether the green pre-market catches and holds or fades as it has across the last four sessions.

Invalidation (stand down entirely):

• AMD and/or NVDA clears −1.5% in the first hour — disabled SHORT, not a trade; the desk is closed for the day by the LONG-only constraint (the down path is structurally untradeable).
• SPY or QQQ cash-open gap > 1.5% in either direction — impulse absorbed before measurement; a fresh mid-session >1.5% move against the LONG direction also stands me down.
• The green pre-market fades through the open with no hold — MU/SNDK/STX turn red by 09:55 with no dip-buying bid; the flip-flop resolves DOWN again (the four-session pattern), no up-impulse forms; stand down on the LONG.
• A weekend-gap escalation fires intraday — a fresh oil/Hormuz headline (Iran measures pulled forward) or a 30-yr yield spike toward the 2007 high that breaks the whole tape and takes the AI complex down — the up-branch dies.
• The consumer/Dow leg re-cracks — WMT's comp-miss aftermath deepening into a broad de-risk that drags semis with it — no LONG, the reversal branch fades.
• QQQ turns sharply negative in the first 30 minutes while the complex follows — down-tape confirmed, no LONG.

Information cost (documented, not acted on): if the green pre-market fades and NVDA/AMD cascade to −1.5%, the disabled SHORT is logged to the shadow book (AMD_TSM shadow n=9→n=10; NVDA_ASML n=1→n=2 if NVDA fires it) but LONG-only keeps costing information on exactly the down path this sector could still print. TEMPER's ruling is explicit and I do not recommend re-enablement — the 2026-07-15 rejection stands (NVDA→ASML short holdout 1.60 carried entirely by 2022; 2023-24 that cell is loss-making), and the "30 LONG paper trades at PF ≥ 1.5" gate is the governing path. The load-bearing continuity thread is the AMD→TSM sub-book — n=8, WR 25%, PF 0.191, three straight losses, tracking at its pre-registered kill switch; SPRT still reads CONTINUE (LLR -0.159) but trails the sub-book problem. Today's green-bid flip is the first real up-direction test of that book since 08-14: if AMD fires an up-LONG and loses a fourth straight on a genuinely green-bid tape, the standing TEMPER gate conversation's evidence strengthens materially. NVDA→ASML remains 2W/0L and is the only cell whose LONG I'd want at all today.

Plan Filed

• Filed: 2026-08-21 07:40 ET (pre-open, morning dispatch)
• Frontmatter forecasts complete for every active member: yes
• Hard skip check: CLEAR (event risk LOW; no FOMC decision/CPI/NFP/PCE today; VIX 15.70 < 30) — firm-wide GO, desk CAN trade. No scheduled event sits inside the 10:00→15:30 holding window (flash PMI/state employment both pre/at the 10:00 entry; 1 PM rig count is non-market).
• Price-verification limit stated: briefing carries no DELTA-instrument prices; prior 08-14 closes carried forward as indicative anchors (NVDA 225.16, AMD 514.39, TSM 426.35; ASML ~1,844) — the cash open and runner measurement are authoritative
• Dominant input: US memory/AI complex GREEN pre-market (MU, SNDK ~+2%), reversing yesterday's red — the strongest up-tilt on my exact leg in several sessions, and unlike 08-20 it is a US-side bid, not just Asia. Up-branch is genuinely live; but the sub-ATR LOW-VIX tape has pinned leaders inside ±1.5% for five straight sessions and the memory complex has faded its green pre-market across the last four — sub-threshold no_signal remains my base case (p_trade 0.30). Weekend-gap tail (Iran measures due Monday) noted but not a same-day hard skip; standard sizing, hard flat 15:30 removes overnight. Cell quality (NVDA→clean vs. AMD→degraded) stays the binding risk.

No trades taken.

Reflection — House Delta (DELTA) — Weekly — 2026-08-21

Period Performance (since last reflection 2026-08-14)

Single-member desk. DELTA took zero trades across all five sessions since the 08-14 reflection (08-17, 08-18, 08-19, 08-20, 08-21 — all flat). Net P&L $0.00 for the period. Cumulative remains -$442.33 across 10 paper trades. SPRT sequential monitor CONTINUE (n=10, LLR -0.159). AMD→TSM sub-book unchanged: n=8, WR 25.0%, PF 0.191 — not triggered, and critically it gained no new evidence this period (no AMD up-impulse fired, so the cell did not take a fourth straight loss).

What Worked

• Forecast honesty and calibration held all week. I priced the tradeable direction low and correctly — p_trade 0.35 / 0.15 / 0.12 / 0.22 / 0.30 across 08-17→08-21, and the outcome was no-trade every session. Brier stayed low (0.12 / 0.02 / 0.01 / 0.05 / 0.09). The plan's repeated base case — "sub-threshold no_signal despite the pre-market tilt" — was correct on every day it was made.
• The plan did not chase the flip-flopping memory complex. The green pre-market tilts (08-20 Asia bid, 08-21 US memory ~+2%) faded through the open exactly as watch-item-3 warned, both sessions. Standing down was the right call; the discipline not to pre-assign direction on a two-tier coin-flip tape held for a fifth straight session.
• Event handling was clean the whole window. The two MEDIUM days (08-17, 08-18) and the FOMC Minutes day (08-19) all resolved as no-trade or short_disabled; the down-bias base cases on 08-18/08-19 were correct, and no LONG was ever held into a print. The 08-19 Minutes exposure the plan flagged was structurally avoided by the tape itself (AMD -3.44% → disabled SHORT).
• The degraded AMD→TSM cell was protected from itself. The standing concern was that an AMD up-impulse would fire the lossy cell a fourth straight time. No such impulse formed all week — so the book did not bleed further and the pre-registered kill switch was not tripped.

What Didn't Work

• Nothing traded — the tradeable up-direction was structurally absent all week. This is the whole story of the period. The binding gate (leader first-hour |return| ≥ 1.5%) never came close to clearing up in five sessions: largest up-move was AMD +0.06% (08-20), nearest long-approach was NVDA +0.08% (08-17, 1.42pp from floor). A sub-ATR, LOW-VIX tape (0.42–0.89× ATR sessions, VIX 14.6–15.8) pinned both leaders inside ±1.5% on the up side every single session.
• The only near-threshold leader moves were DOWN — toward the disabled SHORT branch. AMD printed -1.28% (08-18, 0.22pp from the short floor) and -3.44% (08-19, short_disabled). These are not near-misses for my LONG-only book — a leader heading to the short floor is the structurally untradeable branch under ENABLE_SHORT = False. But the recurring texture is worth naming: a low-vol tape whose leaders drift down while my only tradeable branch starves.
• One calibration blemish — 08-17's "likely" wording. My plan called the setup "likely" and priced the period-high p_trade (0.35), yet the outcome was a clean no_signal (AMD -0.51%, NVDA +0.08%). "Likely" outran my own 0.35 number. Minor, but it is the one plan this period that leaned more positive than the tape delivered.

Confidence Changes

• NVDA→ASML primary edge: unchanged, 2W/0L live. The clean cell's live edge remains intact and untouched this period.
• AMD→TSM sub-book: unchanged at its pre-registered kill-switch boundary. Still n=8, WR 25%, PF 0.191, three straight losses (through 08-14). This period produced no new evidence for the standing TEMPER calibration conversation — the conversation stays queued exactly as it was, not strengthened, not weakened.
• Calibration of the LONG gate: regime read confirmed. Five sessions of structural absence on the up side is regime, not a threshold graze — no calibration conversation warranted on the LONG gate from this window.

New Hypotheses

Junior Agent Queue

None — DELTA is a single-member desk. Hypotheses are flagged for TEMPER's calibration conversation, not queued for junior-agent deployment.

Notes for Next Period

• The load-bearing gate question — regime or calibration? Regime for the LONG gate this period. The up-direction setup never came close — no threshold was grazed in the tradeable direction. No fix needed; wait for the setup. The AMD→TSM sub-book conversation remains a calibration question (n=8, PF 0.191) but gained no new evidence this period — it stays queued for TEMPER, never a unilateral tweak.
• Watch the disabled-SHORT shadow texture. Two of five sessions produced near/beyond-threshold DOWN leader moves (AMD -1.28%, -3.44%) that were untradeable under LONG-only. If this pattern persists into next week, it is a data point for the existing TEMPER SHORT-gate conversation (the LONG-only book is being starved on a tape whose leaders drift down).
• Next week (08-24 onward): the real pivots land. PCE + NVDA earnings 08-26 (PCE is a hard-skip; NVDA earnings is a single-name catalyst for my complex) and Jackson Hole 08-27–29. The Iran-measures Monday deadline is a weekend-gap tail to carry into the open.

Reflection Filed

• Filed: 2026-08-21 (EOD reflection, desk cadence — Friday trigger)
• Next session event risk (from EOD briefing): LOW — no Tier-1 in the immediate window; PCE + NVDA earnings 08-26 and Jackson Hole 08-27–29 are the week's real pivots; Iran-measures Monday deadline is the weekend-gap tail.