House Delta — 2026-09-21

Desk Plan — House Delta — 2026-09-21

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-09-21-pm.md PM: DELTA

Shared Market Read

• Event risk today: LOW — no Tier-1. No FOMC/CPI/NFP/PCE through 2026-09-28; next hard-skip for this desk is PCE Sep 30 (then NFP Oct 2, CPI Oct 14; FOMC Minutes Oct 7). Today's US schedule is macro-silent: only the Chicago Fed National Activity Index (7:30 AM) plus Fed speakers Goolsbee/Jefferson/Williams/Barkin — none a market binary. Firm-wide status: GO. The session's structural catalyst is not a macro print — it is the rebalance: both the S&P 500 and the Nasdaq-100 quarterly rebalances are effective prior to today's open (passive flow cluster). Monday is a light-earnings day; no major S&P reporter today.
• Session character expected: Mixed / two-tier with a tech-led bid — with a passive-flow overlay. The overnight tape opened firm and AI-chip-led (Nasdaq futures ~+0.7–1.1%, S&P/Dow ~+0.5–0.7%), the AI "pick-and-shovel" complex (Nvidia modestly higher, memory/semis bid) leading — the same complex that led Friday's close, this despite net-negative AI weekend headlines (Anthropic IPO delay, OpenAI cash-burn, Oracle debt). But the briefing is explicit: do not pre-assign a one-day polarity on a rebalance + geopolitical-coinf-flip Monday, and the weekend Houthi strikes on Riyadh / the Aramco Yanbu export hub + the Bab el-Mandeb battle keep crude two-way overnight-to-intraday (Brent ~$98–102 after sliding ~2% on UNGA de-escalation hopes). The rebalance effective date and the oil vector are the dislocators — passive index flow is a diffuse rotational-cluster force, not clean directional impulse.
• VIX regime: 14.87 (+0.06 vs prior 14.81) — LOW (<18), nowhere near the 30 hard-skip. This is DELTA's strongest regime by the backtest (dev PF 2.59 in VIX < 20; the family-inverted dependency — high macro vol adds noise to the supply-chain transmission rather than amplifying it). The vol backdrop is favorable; the texture of the session is the concern.
• Key levels (this desk's instruments — adopted as direction/sector-dispersion only, not owned levels; live confirmation at the cash open): NVDA prior close $222.27 (+1.34% on 09-18); AMD prior close $559.82 (+2.70% on 09-18); ASML premarket ~$1,705.90 (+1.55%); TSM prior close $434.67 (+1.02% on 09-18). Premarket single-name breadth on the complex is firm and broad (INTC +5.5%, NVDA +0.94%, QQQ +1.01%, SPY +0.62%). The only load-bearing number for this desk is each leader's first-hour return off the 09:30 cash open vs. the ±1.5% floor — and NVDA's premarket +0.94% sits inside it. Index context (briefing, Alpaca-verified): QQQ prior close 721.36, above SMA20 713.25; SPY 761.62, below SMA20 764.26 — the continuing tech-over-value, leadership-concentrated (breadth ~31%) shape.
• Macro backdrop: A post-Fed-recovery two-tier market (QQQ > SMA20, SPY ~SMA20) whose live risks are external and diplomatic: the S&P/Nasdaq rebalance flow cluster effective today, the Houthi/oil supply-line escalation, and the Trump–Xi summit (Wed) / UNGA opening a diplomatic week — an AI-safety notification mechanism is a constructive possible readout, a hostile tariff/trade or Iran readout would reverse the risk-on. For a stat-arb desk whose edge lives on NVDA/AMD-specific idiosyncratic moves, the summit (Wed) is a latent China/NVDA two-way, but it lands after today's session — today's live structural feature is the rebalance flow.

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

Call: No trade expected — p_trade 0.15, direction long, conviction low. The engine runs (LOW event risk, no Tier-1, VIX 14.87 LOW, firm GO, NOT a skip day), but the third-strongest reason to be flat in a row is the same structural one, and today adds a rebalance-layer on top. This is the mechanism's structural stand-down, not a risk stand-down: the live driver is complex-wide systematic beta (not an idiosyncratic design→manufacturer cascade), and today's passive-flow-cluster tape is the texture the edge does not sample — per the firm's own 2026-06-22 TEMPER/calendar entry, quarterly rebalance effective dates are exactly the diffuse, below-ATR, low-conviction rotation sessions that depress directional-institutional-momentum strategies.

Setup present: no — structurally incompatible, on two accumulated grounds. (1) The dominant reason, third consecutive session: the driver is complex-wide, not idiosyncratic. The LEAD leg is the whole AI-chip complex up together premarket (INTC +5.5%, NVDA +0.94%, memory/semis bid), the same one-factor bid that failed to lift the leaders to the floor on 09-17 (AMD +0.98% / NVDA +0.39%) and on 09-18 — complex-wide beta reprices leader and laggard simultaneously on one systematic factor, which is by construction not the supply-chain cascade (leader-specific news telegraphing to the laggard over hours) my edge is built on. A one-session-old and a two-session-old read both show a strong sector bid leaving the leaders inside the ±1.5% floor; nothing in today's premarket (NVDA +0.94%) argues that changes. (2) The rebalance-effective tape is the wrong texture for my measurement window — TEMPER's own framework says so. Item A of the 2026-06-22 learnings entry: Nasdaq-100 quarterly rebalance effective dates should be a session-character modifier flagging lower setup likelihood for strategies whose edge requires directional institutional momentum or clean sweep formation — passive program volume dominates intraday structure with diffuse rotational churn rather than price discovery. The Nasdaq-100 rebalance effective today directly implicates NVDA, AMD, ASML, TSM (all Nasdaq-listed; NVDA/AMD are Nasdaq-100 members). A first-hour leader move on a rebalance-open is more plausibly flow/rotation than leader-specific information — the same class of concern as Friday's witching, albeit milder and a different mechanism. I am not going to dress up a fifth-week structural absence as a disciplined refusal — the honest statement is that the room is being blocked by the shape, and it is the same shape that has blocked every session since 08-03.

Sizing vs. event risk: n/a — no position expected. If — against both shape reads — a leader clears +1.5% on a move that reads genuinely idiosyncratic (laggard-leg news carrying independently of the leader), standard 0.75% single-position risk applies unchanged: event risk LOW, and my hard 15:30 flat + 2.0× ATR(14, 5-min) stop carry zero overnight and zero event exposure by rule. No parameter is adjusted. One note specific to today: because the rebalance-open distorts the first-hour measurement window, any leader clear is presumed flow until a laggard-leg catalyst proves otherwise — that is a read on signal quality, not a sizing change.

Session-character fit: poor, and poor for the same accumulating reason. A rebalance-effective + geopolitical-coinf-flip Monday is a passive-flow/rotation session, not a pairs-convergence session. The one clean upside path remains a leader that decouples from the sector bid and builds an idiosyncratic first-hour impulse; the rebalance flow cluster makes that decoupling less likely to be distinguishable from rotation churn. The 09-17 evidence stands as the cleanest recent read and it is two sessions old: a strong sector bid does not produce a qualifying idiosyncratic leader move in this regime.

Key levels (mine only): the only load-bearing number is the leader's first-hour return off the 09:30 cash open, read at 09:55, vs. ±1.5% — NVDA ≥ +1.5% → buy ASML at the 10:00 bar open (primary, edge-carrying cell, live 2W/0L, last live trade 08-03) and AMD ≥ +1.5% → buy TSM at the 10:00 bar open (secondary; degraded sub-book n=8, WR 25.0%, PF 0.191). Leader ≤ −1.5% = the disabled-SHORT branch → no trade by rule. No instrument-level price is adopted for any of the four legs (see Key levels above — premarket prints are direction/sector-dispersion only, all four require live confirmation at the cash open).

Invalidation / no-trade reasons (all binding today): (1) the live driver is complex-wide AI-chip beta, not an idiosyncratic cascade — the standing structural stand-down, at its strictest because the repricing sits on my own instruments, and 09-17/09-18 both show the sector bid failing to lift the leaders to the floor; (2) the rebalance-effective tape is diffuse passive-flow rotation, not the clean price-discovery my edge samples (TEMPER 2026-06-22 item A — rebalance effective dates = lower setup likelihood for directional-institutional-momentum strategies); (3) leaders opening broadly down ≥1.5% = the disabled-SHORT side (ENABLE_SHORT = False) — no trade by rule; (4) any intraday VIX ≥ 30 closes consideration; (5) the remaining risk-derived stand-downs are unchanged. Next hard-skip: PCE Sep 30 (NFP Oct 2, CPI Oct 14). The Trump–Xi summit (Wed) is a live NVDA/China two-way but lands after today's session — it is watch-briefing, not today's gate.

Up-branch / calibration status (carried forward, unchanged): the queued flag — a third NVDA up-graze inside ~0.3pp turning regime into a TEMPER calibration question — remains queued and untouched, not advanced today unless a leader actually clears the floor. The ledger stands: NVDA +1.23% (09-02), +1.02% (09-03), AMD +1.20% (08-31); 09-08 NVDA −1.55% (down-side, disabled); 09-10 NVDA −1.44%; 09-14 AMD −0.27% / NVDA −0.21%; 09-17 AMD +0.98% / NVDA +0.39%; (09-15/09-18 unobserved — runner/watchdog pipeline states, never counted as grazes). No unilateral gate change.

Book state (carries forward untouched): cumulative −$442.33 across 10 LONG trades, 4W/6L, WR 40.0%, PF 0.494 (SPRT n=10, LLR −0.159, CONTINUE). AMD→TSM sub-book n=8, WR 25.0%, PF 0.191 — the 09-05 stop-calibration diagnostic read NO (widening the stop does not rescue the cell: PF 0.073 → 0.119 across 2.0–4.0×), pending TEMPER cadence-ratification; a signal/regime watch, not a parameter fix. NVDA→ASML 2W/0L remains the quality cell; its trigger has not recurred since 08-03. Rolling 6-month PF monitor unchanged, not escalated.

✔ Operational item / p_trade honesty — carried forward at desk-top priority: the per-agent run-trades watchdog. learnings.md records DELTA starved by the per-agent 60s watchdog on cold-cache sessions (09-09, 09-10, 09-17; this desk's own record adds 09-15 and today's prior 09-18 as genuine runner_failed/starvation holes). My forecast is scored against trades.csv — if the watchdog trips again today, the run produces no row at all and my 0.15 is scored against a pipeline absence, with the session's leader first-hour returns lost on precisely the rebalance day I would most want to observe. A missing observation is never evidence of a near-miss and must never be narrated as a strategy stand-down. Standing asks: backfill any standalone holes, and wire in the ledger-freshness post-run check. The 0.15 is a probability of a mechanical/first-hour trigger on a broad, firm premarket into a rebalance-open — it is not an endorsement of the resulting trade.

Watch items (informational only — not entry signals): (1) whether NVDA/AMD decouple from the sector bid and build genuine idiosyncratic first-hour up-impulses — distinguished as always by the laggard-leg (ASML/TSM) news carrying independently of the leader; on a rebalance-open I must also judge whether any such move is distinguishable from rotation flow. (2) Whether crude re-escalates (Houthi/Riyadh/Yanbu headline) and rattles the two-tier risk-on — the live two-way under a firmer open. (3) Whether a LOW-VIX, no-Tier-1, complex-bid session finally produces a clean ≥1.5% first-hour up-impulse on a leader — the same watch every session since 08-03, and the one the open event-day diagnostic run (skip disabled, still unperformed) exists to interrogate.

Plan Filed

• Filed: 2026-09-21 04:35 PDT (pre-open)
• Frontmatter forecasts complete for every active member: yes (delta — p_trade 0.15, direction long, conviction low)
• Briefing event risk confirmed: LOW — no Tier-1 through 2026-09-28; S&P 500 + Nasdaq-100 quarterly rebalance effective at today's open is the session's structural catalyst
• Verdict: No trade expected — the engine runs (LOW event risk, no Tier-1, VIX 14.87 LOW, firm GO), but the session is structurally incompatible with the mechanism on the accumulating complex-wide driver (the whole AI-chip complex is the lead, systematic beta not an idiosyncratic cascade; 09-17 and 09-18 both show a strong sector bid failing to lift the leaders to the ±1.5% floor), reinforced today by the rebalance-effective passive-flow-cluster tape that TEMPER flags as the wrong texture for a directional-institutional-momentum edge. p_trade 0.15 reflects the nontrivial chance a broad firm premarket into a rebalance-open mechanically clears a first-hour floor — not an endorsement; direction long, conviction low.

House Delta — Desk Reflection — 2026-09-21

Cadence rule: Weekly/Friday or same-day-on-trade (WS5 per-desk cadence). Today DELTA logged a real trade, so the desk reflects same-day. Period covered: 2026-09-18 (last desk reflection) → 2026-09-21. Scheduled session in window: 09-21 (Mon).

Single-member desk — the desk reflection is DELTA's own analysis at desk scope; the per-member compatibility file (agents/delta/reflections/daily_2026-09-21.md) is written separately and remains first-person.

Desk Performance Summary

Desk total for the period: +$375.06 — the desk's first real trade since 08-06 and its best single-trade result of Phase 3.

Period session map:

SPRT moved sharply in the desk's favor: LLR −0.159 → +0.219 on the single win — the first positive LLR movement in the cycle, now +2.725 from the CONSISTENT-WITH-BACKTEST boundary and −3.164 from DEGRADED. Cumulative P&L −$442.33 → −$67.27 (net of commission). AMD→TSM sub-book advances to n=9, WR 33.3%, PF 0.62 (gross win $542.31 / loss $874.00) — still sub-1.0, but its first win since 08-06 and the strongest reading since the 09-05 diagnostic.

Per-Member Binding Gate Analysis (delta — the whole load)

Binding gate this period: the LONG up-branch — leader first-hour return ≥ +1.5% off the 09:30 cash open, read at 09:55. It both bound and cleared this period, and the clearing vindicates the regime reading.

1. The floor finally cleared, decisively, and the trade fired through the cell that had been written off. AMD printed +3.69% in the first hour — 2.19pp above the floor, nowhere near a graze — on a clean, idiosyncratic catalyst-class move (AMD's $1T-market-cap rally on AI demand, the marquee driver of a big trending tech day, QQQ 2.30× ATR). This is the exact CASE the desk's standing "regime, not calibration" verdict held for: the accumulated up-approaches since 08-28 (NVDA +1.23%/+1.02%, AMD +1.20% 08-31, +0.98% 09-17) were genuine approaches to a live setup, not evidence of a broken threshold. When a decisive leader-specific move finally arrived — not the diffuse sector-wide beta the five-week stand-down feared — it cleared with massive room and the LONG edge worked as backtested (HARD_FLAT into a held close).
2. No single threshold was repeatedly grazed — so no calibration conversation on the up-floor. A floor crossing at +3.69% is definitionally not a graze. The decision to change nothing on the gate stack (no ATR-stop, no floor, no pair-weighting tweak) is confirmed correct in hindsight. The 09-04 watch-to-escalate (a third NVDA up-graze inside ~0.3pp → calibration conversation) never fired then and now is structurally moot — the floor proved reachable and profitable. Verdict: REGIME, not calibration — confirmed by a decisive crossing, not a graze.
3. The finding that matters is forecast calibration, not gate calibration. The desk priced today p_trade 0.15 (direction long, conviction low) on the plan's structural stand-down narrative (rebalance-effective passive-flow rotation, complex-wide bid), and the engine then fired a +$375.06 win → Brier 0.7225, worst class. This is the 4th documented under-forecast in DELTA's p_trade series (07-22 catalyst blind spot, 08-13 weak-cell-live, 08-03 opposite-anchor, today). The recurring root is the same: the aggregate is anchored on the prior no-trade regime and the degraded AMD→TSM cell is scored as effectively not-live — so on big trending tech days where that cell is the one that clears, the probability is structurally under-stated. This is a DELTA-owned p_trade-calibration conversation for TEMPER (already an established learnings series), not a gate change.

Two escalations this period (both desk-blind, on the desk's own book):

• Rolling 6-month PF monitor FIRED — TEMPER binding condition #2, first trigger. Trailing PF reads 0.923 < 1.2 on 11 trades. The driver is concentrated in the AMD→TSM cell (n=9, WR 33.3%, PF 0.62); NVDA→ASML remains clean (2W/0L). The 2026-09-05 stop-calibration diagnostic (read NO — widening the stop does not rescue the cell) stands unchanged; today's win is N=1 live evidence the degraded cell can work, but the sub-book PF is still sub-1.0, so a calibration-shaped TEMPER conversation on the rolling PF trigger and the AMD→TSM cell is warranted — never a unilateral gate tweak.
• Pipeline recurrence + stale output — unresolved operational debt. The per-agent 60s watchdog tripped on DELTA again at today's 1:30 run-trades (5th flag in the streak: 09-09/09-10, 09-15, 09-18, today). The orchestrator's direct backfill recovered the observation this time — daily_2026-09-21.md and trades.csv now carry the real +375.06 WIN, proving the backfill path works when invoked. But the 18:00 auto EOD desk summary still read DELTA as "no performance file / no trade" despite the daily log existing at 15:31 — so the ledger-freshness / stale-output post-run check is still not implemented. Repeated backfills are the mitigation, not the cure.

Implication — no action on the gate stack; escalate p_trade calibration + rolling-PF trigger to TEMPER; fix the watchdog/freshness pipeline. The desk's strategy stood up exactly where it was validated; the losses this cycle were forecast bias and a degraded sub-book, both now surfaced for review with live evidence.

Open TEMPER Items (Carried Forward)

Reconnaissance Ledger Items

• Watchdog root fix + ledger-freshness post-run check — wire the freshness/stale-output check so a missing or stale performance/daily_.md after run-trades is flagged and the EOD summary reflects the recovered state. Today proved the backfill recovers the data but not the reporting: the 18:00 desk summary still listed DELTA as having no performance file. This is the desk's top operational priority, now at 5 flags.
• TEMPER conversation — rolling PF 0.923 trigger + AMD→TSM cell — the rolling-6-month monitor fired for the first time; pair with the pending 09-05 stop-calibration ratification. Desk-blind; on DELTA's own book.
• TEMPER conversation — p_trade calibration series (4th instance).
• Event-day diagnostic (skip-disabled) remains a standing obligation.

Desk Summary for Orchestrator

House Delta: the LONG up-branch finally cleared and the desk won. AMD first-hour +3.69% (> +1.5% floor by 2.19pp) on its $1T-milestone rally → TSM LONG 57 @ 438.50, HARD_FLAT 445.08 @15:30, +$375.06 WIN — first real trade since 08-06, first AMD→TSM LONG since 08-14. Cumulative −$442.33 → −$67.27 (11 trades, WR 45.5%, gross PF 0.923); SPRT LLR −0.159 → +0.219 (CONTINUE). Binding gate: LONG up-floor — verdict REGIME, not calibration, now confirmed by a decisive crossing (the accumulated prior up-approaches were a live-setup warm-up, not a grazed threshold; no gate change made or warranted). Two escalations, desk-blind: (1) the rolling 6-month PF monitor FIRED (0.923 < 1.2, TEMPER binding condition #2, first trigger), driver concentrated in the AMD→TSM cell (n=9, PF 0.62) — a TEMPER calibration conversation, on top of pending 09-05 stop-diagnostic ratification; (2) p_trade under-forecast (0.15/Brier 0.7225 on a winning day — 4th instance of the under-forecast series). One thing watched operationally: the per-agent 60s watchdog tripped on DELTA a 5th time at today's 1:30 run-trades — the orchestrator backfill recovered the observation (real +375.06 is in trades.csv/daily log), but the 18:00 EOD row still showed DELTA as no-performance-file, so the ledger-freshness post-run check is still not wired.

Reflection Filed

• Filed: 2026-09-21 15:35 PDT (EOD, desk cadence — same-day-on-trade trigger)
• Next session: Tue Sep 22 — event risk LOW, no Tier-1 through 2026-09-28; next hard-skip PCE Sep 30 (then NFP Oct 2, CPI Oct 14). Live week two-way: Trump–Xi summit Wed Sep 24 (AI-safety/export-control agenda) lands after tomorrow's session — watch, not gate. Watch: whether the AMD→TSM cell's return to profit (after today's +$375.06) is N=1 or the start of sub-book recovery, and whether NVDA→ASML (the clean cell, dormant since 08-03) finally clears its floor on the same tape.