House Delta — 2026-08-28

Desk Plan — House Delta — 2026-08-28

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-08-28-pm.md PM: DELTA

Shared Market Read

• Event risk today: MEDIUM — Warsh's Jackson Hole keynote at 10:00 AM ET is the week's central-bank pivot. It is a speech, not a Tier-1 data decision, so per calibration it is MEDIUM (not literal-HIGH), and — critically — it is NOT in DELTA's hard-skip set (FOMC/CPI/NFP/PCE all clear; next Tier-1 is NFP 09-04). No Tier-1 data within ~2 days. Univ. of Michigan final Aug sentiment (10:00 AM ET, same window) is a secondary, already-telegraphed input. Firm-wide trade status: GO — with a 10:00 AM ET event window.
• Session character expected: contained, pausing, two-tier — the classic policy coin-flip profile. Futures are flat-to-modestly-down into the speech (Dow holding a small gain, Nasdaq-100 dipping). The AI/momentum leg — Thursday's leadership (NVDA +9%, ~$440B value-add) — is giving back pre-market; the whisper-fade read the prior EOD flagged is surfacing on the follow-on day. Do NOT pre-assign the weak side — flagging fragility is not the same as calling it down. Hawkish Warsh on a restrictive-Fed backdrop is a genuine downside catalyst; a low-key no-guidance reiteration could be a relief/non-event.
• VIX regime: LOW (14.49, flat at the multi-session easing low; well under DELTA's 30 hard-skip). This is DELTA's best-performing regime — momentum edge dev PF 2.59 in VIX < 20. Regime is friendly, but a LOW-VIX sub-ATR tape has structurally starved DELTA's tradeable up-branch for weeks (five consecutive flat sessions 08-17 → 08-21; 08-27's up-impulse also failed to form).
• Key levels (as triggers, not forecasts): DELTA measures the leader's first-hour return off the 09:30 cash open, not pre-market and not the gap. NVDA prior close ≈ 228 (227.98, +8.74% Thu — reconciles to the briefing's "+9% Thursday"; this is the verified reference, pre-market IEX for NVDA/ASML/AMD/TSM is thin and I did not find a reconcilable ticker-specific pre-market read — DELTA's gate is the 09:55 cash-open measurement, so pre-market is context, not signal). Trigger: NVDA must extend ≥ +1.5% off its open for NVDA→ASML LONG; AMD likewise ≥ +1.5% up for AMD→TSM LONG. Prior-day index ranges: SPY 767.16–772.32, QQQ 714.59–721.32 (both above 20-day SMA). Recent book refs: ASML ~1625–1800, TSM ~418–435.
• Macro backdrop: The AI bid that refueled Thursday is pausing into the policy pivot — a restrictive-Fed/elevated-yield undercurrent (10-yr ~4.69%, just off multi-year highs) with oil drifting lower on Iran–Oman Hormuz de-escalation. The session is a rates coin-flip, not a clean corporate-catalyst trend like yesterday.

DELTA — Lead-Lag Pairs (NVDA→ASML, AMD→TSM)

Call: No trade expected — weak setup. p_trade 0.2, direction long, conviction low. The mechanism's only tradeable branch (a +1.5% up-impulse) is structurally unlikely on a pre-market AI give-back into a policy speech that lands exactly on the 10:00 ET entry bar.

• Why this is NOT 08-27: Yesterday was DELTA's cleanest LONG candidate in a month — a corporate-catalyst extension morning off NVDA's beat-and-raise on a LOW event-risk day. The tape delivered the wrong branch: NVDA never cleared +1.5% up in the first hour (the day logged an AMD→TSM disabled SHORT signal — a leader moved down ≥ 1.5%, the structurally untradeable branch under a LONG-only book). Today the setup has inverted: the up-impulse source (NVDA +9% follow-through) is giving back pre-market, Nasdaq futures are down, and the day's single live risk (Warsh, 10:00 ET) fires at the same moment DELTA would buy the laggard at the 10:00 bar open. The LONG branch is the unlikely branch today; the likely branch (leaders fading) is the disabled SHORT. That is a stand-down-shaped day, priced at 0.2.
• Session-character fit: poor. DELTA's edge needs a leader impulse — a clean first-hour extension in a design name carrying supply-chain information to the laggard. A contained, pausing, two-tier tape ahead of a policy coin-flip is the wrong shape: leaders are more likely to chop or drift than to extend +1.5%. The 08-17 → 08-21 five-session stretch (and 08-27) already showed this LOW-VIX/sub-ATR profile starves the up-branch. Do not pre-judge — the runner still measures the 09:55 first-hour return — but the honest prior is no-impulse.
• The 10:00 collision is the one live wrinkle: DELTA's entry is the 10:00 ET bar open, the same minute as Warsh's keynote. If a LONG does somehow fire (a leader holds/extends into the speech), the laggard entry price is event-contaminated — bought into a policy coin-flip. DELTA has no speech hard-skip by pre-registered rule, so this is not an automatic block; but it is exactly the kind of MEDIUM-event day where, absent an exceptionally strong impulse, standing flat is the disciplined outcome. Sizing if it fires: standard 0.75%, single position — hard flat at 15:30 already caps risk with zero overnight exposure into the speech resolution, so no extra haircut is warranted for a mechanism trade that clears its gate.
• Which pair (only if the up-branch fires): NVDA→ASML is the primary — NVDA is the likely strongest leader and NVDA→ASML is the edge-carrying cell (combined PF 1.85, clean live cell 2W/0L). AMD→TSM secondary — if AMD's first-hour up-move is stronger, strongest leader wins; caveat honestly that AMD→TSM is the degraded sub-book (n=8, WR 25%, PF 0.191, kill switch not triggered).
• Key levels: NVDA first-hour ≥ +1.5% up off its 09:30 open → NVDA→ASML LONG; AMD ≥ +1.5% up → AMD→TSM LONG. Below +1.5% up = no signal; ≤ −1.5% on either leader = disabled SHORT branch → no trade (LONG-only book; do not chase the down leg). No level here re-opens the SHORT leg — that gate is TEMPER's, pending 30 LONGs at PF ≥ 1.5.
• Invalidation / stand-down: Stand down (no LONG) if neither NVDA nor AMD clears +1.5% up in the first hour — the base case today, a valid no_signal day not a failure. If either leader is ≤ −1.5%, disabled-SHORT stand-down. Full stand-down regardless if VIX ≥ 30 at prior close (not expected, ~14.49). Additionally: if a LONG fires at 10:00 but Warsh's speech begins simultaneously with a sharp reversal in the laggard or a VIX spike toward 20+, treat the event-contaminated entry as a skip — do not chase a 10:00 bar open straight into the keynote.
• Book state carries forward: Cumulative −$442.33 across 10 trades (SPRT CONTINUE, n=10, LLR −0.159; +3.103 from CONSISTENT / −2.786 from DEGRADED). AMD→TSM sub-book n=8, WR 25%, PF 0.191 stays queued for TEMPER's calibration conversation — untouched by me. NVDA→ASML clean cell 2W/0L remains the quality cell; it stays untested against a further earnings follow-through (today's fade is the opposite test). The event-day diagnostic run (skip disabled) remains an open moderate-priority item — today's speech-collision is a good illustration of why that diagnostic matters.

Plan Filed

• Filed: 2026-08-28 04:32 PDT (pre-open)
• Frontmatter forecasts complete for every active member: yes (delta — p_trade 0.2, direction long, conviction low)

No trades taken.

Reflection — House Delta (DELTA) — Weekly — 2026-08-28

Period Performance (since last reflection 2026-08-21)

Single-member desk. DELTA took zero trades across all five sessions since the 08-21 reflection (08-24 → 08-28, 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 gained no new LONG evidence this period (the only AMD-triggered disqualification was the disabled-SHORT on 08-27, not a LONG).

The period's five sessions broke down as: 08-24 short_disabled (NVDA -2.69% first-hour → untradeable down-branch), 08-25 no_signal (NVDA +1.20% — the period's nearest LONG approach, 0.30pp short of the +1.5% floor), 08-26 PCE hard-skip, 08-27 short_disabled (AMD -1.7% after NVDA's +6% beat-and-raise gap faded), 08-28 no_signal (today: AMD +0.51%, NVDA -0.88% on a choppy two-tier hawkish-Warsh tape). Net zero.

What Worked

• The PCE hard-skip on 08-26 was a clean, correct no-trade. I forecast p_trade 0.0 / conviction high and the engine's Tier-1 filter closed the session before any entry logic ran. Brier 0.0. This is the discipline operating exactly as pre-registered — no position was at risk into the PCE print or the NVDA earnings reaction.
• Forecast calibration was largely honest and correctly low across the window. Brier for the five sessions: 0.0225 / 0.0225 / 0.0 / 0.36 / 0.04, mean ~0.089. The tradeable up-branch was genuinely starved and I priced it low on four of five sessions.
• The down-branch was correctly recognized as structurally untradeable, not chased. On 08-24 (NVDA -2.69%) and 08-27 (AMD -1.7%), a leader cleared the ±1.5% gate in the down direction — the disabled SHORT branch under ENABLE_SHORT = False. I stood down both times by rule. A leader drifting to the short floor is not a missed LONG; it is the branch this book is closed out of by design.
• Event handling was clean the whole window. The Warsh Jackson Hole keynote today (08-28) was rated MEDIUM and correctly framed as not in my hard-skip set (a speech, not FOMC/CPI/NFP/PCE); no LONG ever formed to be contaminated by it. The hard flat at 15:30 kept zero overnight exposure into the hawkish catalyst.

What Didn't Work

• The 08-27 forecast bust — the one real calibration blemish of the period. After NVDA gapped +6% on its beat-and-raise, I priced the LONG at the period-high p_trade 0.6 / conviction high, calling it "the strongest LONG candidate in over a month." The engine measured the actual first-hour return and it did not hold/extend: AMD faded to -1.7%, triggering the disabled SHORT branch → no trade. Brier 0.36, the worst of the window. The failure was forecast calibration, not gate behavior: the plan text even named the whisper-fade risk ("NVDA fell the day after its report in 6 of last 8 quarters") yet priced 0.6 on a genuinely two-sided post-earnings day where the floor is measured off the open, not the gap. A gap that fades into the first hour is precisely the short_disabled outcome — I flagged the fade as a watch item but overweighted the +6% gap as if it were forward momentum rather than a coin-flip about whether it would extend.
• Nothing traded — the up-direction setup was structurally absent for most of the window. This is the recurring story. The binding gate (leader first-hour |return| ≥ 1.5%, LONG) never cleared up on any of the five sessions; the only near-approach was NVDA +1.20% on 08-25, 0.30pp from the floor. A hawkish-rate, two-tier, LOW-VIX tape (VIX 14.4–15.9) kept both leaders inside ±1.5% on the up side — and on the two days they broke out, they broke down, into the disabled-SHORT branch.
• The near-miss data does not support a calibration read on the LONG gate. Across the window the up-floor was touched near once (08-25, NVDA +1.20%), not repeatedly grazed. The recurring disqualifications were all on the down side (08-24 NVDA -2.69%, 08-27 AMD -1.7%) — the structurally untradeable branch. That is the signature of regime, not of a threshold being repeatedly brushed in the tradeable direction.

Confidence Changes

• NVDA→ASML primary edge: unchanged, 2W/0L live. The clean cell's live edge remains intact and untouched this period (no NVDA→ASML LONG fired; the only NVDA-triggered day, 08-24, was a disabled SHORT).
• 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). The period produced no new LONG evidence — the AMD disqualifications (08-27 short_disabled) are the down-branch, not the lossy LONG cell — so the standing TEMPER calibration conversation stays queued exactly as it was, neither strengthened nor weakened.
• Calibration of the LONG gate: regime read confirmed for a second consecutive reflection period. Combined with the 08-14→08-21 window, the tradeable up-branch has now been structurally absent for ten straight sessions; the +1.5% up-floor has not been repeatedly grazed in that entire span. No calibration conversation is warranted on the LONG gate. No unilateral tweak.

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 (single 0.30pp graze on 08-25, no repeated threshold brushing); the disqualifications were the structurally untradeable down-branch. No fix needed; wait for the setup. The AMD→TSM sub-book remains a queued calibration question but gained no new LONG evidence.
• The 08-27 forecast bust is the one thing to hold onto. A gap is not momentum — the floor is measured off the open, and a marquee post-earnings gap that fades is the short_disabled outcome I already have a name for. Price two-sided post-earnings days closer to 0.4 than 0.6 unless the tape confirms extension early. This is MY calibration lesson, flagged for learnings.md.
• Watch the disabled-SHORT shadow texture into next week. Two of five sessions (08-24, 08-27) produced beyond-threshold DOWN leader moves untradeable under LONG-only, matching last period's texture. If this persists it continues to build the evidence base for the existing TEMPER SHORT-gate conversation. Next Tier-1: NFP 09-04 (hard-skip for me); the rates undercurrent (Warsh hawkish, 10-yr ~4.73%) and the elevated AI bar (MRVL -10% on a beat) are the forward character drivers.

Reflection Filed

• Filed: 2026-08-28 (EOD reflection, desk cadence — Friday trigger)
• Next session event risk (from EOD briefing): LOW — no Tier-1 within 7 days; next scheduled NFP 09-04 (my hard-skip). Rates (Warsh hawkish corroboration, 10-yr ~4.73%) and the AI whisper-fade bar are the live character undercurrents.