House Slack — 2026-09-15

Desk Plan — House Slack — 2026-09-15

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-09-15-pm.md PM: SLACK (sole member)

Shared Market Read

• Event risk today: HIGH — and this time it is HIGH by structure, not by proximity alone: the FOMC decision + SEP/dot plot + Warsh press conference lands TOMORROW (Wed Sep 16, 2:00 PM ET) and the two-day meeting begins today. Markets price ~90–92% odds of a 25bp HIKE — the first increase since July 2023 — so the headline is near-certain and the two-sided read is the dot plot (how many 2026 hikes are penciled) and whether Warsh frames it as one-and-done or cycle-start. Fed-speaker blackout is in effect. Today's own US calendar is light (ADP weekly 8:15, Empire State 8:30 cons 14.1 / prior 20.6) and the scanner is explicit: FOMC "not a skip — but direction is capped." Firm-wide status: GO. Three live overnight legs: the AI-safety/AI-return repricing (Amodei's "We Must Pace the Frontier," Altman concurring, Oracle's data-center-profitability doubt), a fresh semiconductor leg (memory/SK Hynix spillover plus The Information's report that a Chinese state-backed firm began mass-producing a key chipmaking tool, hitting ASML directly), and an oil escalation (Brent back above $107–108, Saudi's East-West Hormuz bypass still shut after the drone strikes, Iran/Gulf talks postponed).
• Session character expected: two-tier de-risk / fundamental-repricing gap — tech/AI/memory/semis down hard on named catalysts, energy bid, cyber/software ripping on the hardware→software rotation. Pre-market: ASML −7.3%, GLW −9.3%, LRCX −9.0%, MRVL −7.8%, MU −5.3%, SNDK −13% against CRWD +6.1%, PANW +5.1%, ZS +4.6% and Brent $108.05 (+2.35%). 10-yr broke above 5.0%, highest since July 2007; dollar firm. Contained, not panic — but the briefing is right that a repricing gap gives momentum edges nothing clean, and direction into Wednesday is capped. Irrelevant to a 5-day swing: I check IWM once at the open for a signal or a scheduled exit; I hold no intraday view.
• VIX regime: LOW (17.24, +0.14 vs prior close 17.10 — still under the 18 boundary; the scanner's threshold is 30 and it reads CLEAR). For my book, LOW (<20) is my best backtest bucket (n=21, WR 61.9%, PF 3.01) — favorable if a run flags. It changes nothing about what the gate reads today.
• Key levels: the only binding level for this stack is ±4.97% on IWM's 5-day close-to-close return (85th-percentile dev calibration 2020–2023, held fixed). IWM is not in today's Alpaca block; the only IWM print I own this morning is the scanner's pre-market body read ~−0.51% (Investing.com, quoted in intelligence/2026-09-15-watchlist.md). Standing structural refs remain ~$292 floor / ~$300 ceiling. Script-logged run settled at the 09-14 close: −2.05% (performance/trades.csv, row 2026-09-14) — 2.92pp short of the −4.97% flag. Flat book (state/position_ledger.json → open_position: null). SPY 760.75 / QQQ 709.24, both sub-SMA20.
• Macro backdrop: a stagflation-tilted, FOMC-gated week — oil above $107 with the Hormuz bypass shut, a 10-yr at a 19-year high, April-2007-era rate levels on the eve of an expected hike, and the AI capex boom now questioned on return rather than magnitude. IWM is the small-cap/high-beta, most yield- and risk-sentiment-sensitive cohort: the leg a hawkish repricing presses down first — and, symmetrically, the leg that snaps hardest if tomorrow's dot plot is soft.

slack — SLACK (Capacity-Ratio Swing Mean Reversion, IWM)

Setup present: no. No trade expected — the state machine is flat with no cycle in flight, no exit due, and no entry reachable at today's open. p_trade 0.0, direction none, conviction high.

State-machine leg, stated plainly. The ledger is {"open_position": null} — no open position means no 5-day cycle in flight: no hold-silently marker, no close-and-log, nothing carried over the FOMC gate. Day 0 of a cycle that has not started. The only leg that could log a trade is an entry at today's open, which the locked module gates on the reading as of the last completed close (is_run_flagged is shifted one trading day forward, so today is tested against 09-14's close). That number is script-logged at −2.05% — 2.92pp short of ±4.97%. With one possible trade per session (open-from-prior-flag, or a scheduled exit) and neither leg live, a real trade today is mechanically excluded. This is not a judgment call I am making about the FOMC; it is arithmetically impossible for me to trade today.

The roll, bounded honestly (the 09-11 lesson, applied). From −2.05%, a down-flag needs the reading to deepen ~2.92pp to −4.97%; an up-flag needs it to travel ~7.02pp to +4.97%. The reading's day-over-day change is composition-driven, not level-driven, and I state the mechanic precisely rather than loosely: today's close sets n_day_return = close(09-15)/close(09-08) − 1, while the 09-14 reading was close(09-14)/close(09-04) − 1 — both ends of the window roll at once (the base rotates from the 09-04 close to the 09-08 close as the measuring close advances to today). I do not hold IWM closes for those dates — IWM is absent from today's Alpaca block — so the honest output is a bound, not a point estimate: absent a several-percent single-session small-cap close, the down leg does not travel 2.92pp in one session, and the up leg needs 7.02pp. On a tape the briefing characterizes as a contained two-tier repricing gap, with VIX at 17.24 (LOW) and IWM indicated −0.5%, the base case is no flag settles at today's close.

Today is a measurement day, not a trade day. The run settling at today's close gates Wednesday 09-16's open. If a flag were to settle today, the contrarian entry at 09-16's open would fire on the FOMC decision day itself — at the open, hours before the 2:00 PM decision + dot plot + presser — and under the locked mechanism would exit on the 5th trading day after entry, Wed 09-23's close. That hold spans: the FOMC decision (day 1, entered before the print), quad witching (Fri 09-18), and the S&P/Nasdaq quarterly rebalance effective Mon 09-21 — the quarter's largest passive-flow date. That is the sharpest version of the hold-spanning-event question this week offers, and my answer is below.

Sizing vs. event risk — the desk's opinionated call: standard 15%, no stand-down, no cut. Three reasons, in order of weight:

1. The mechanism has no event-day override by design (strategy_v2.md, Rule 6: no stop, no target, no early exit; the 5-day hold is fixed). A plan-level stand-down is not mine to manufacture — "no unilateral tweak" is the firm's standing rule and this stack's own.
2. An FOMC-day entry is inside my tested envelope, not outside it. Pulling the cleared backtest (backtest_v2_2026-07-02_trades.csv, is_event_entry_day): 9 of 68 validated trades entered on an event day — 8 wins / 1 loss, net +$580.07, worst −$36.20 (−0.97% of the $3,750 notional). That sub-book includes 2024-09-18 (an FOMC decision day) and 2024-02-13 (CPI). So the calendar exposure the FOMC-eve question is worried about has been in the mechanism's own record for four years, and the event-entry cohort is not its weak tail. Caveat stated honestly: n=9 is monitoring-grade evidence, not proof, and it is in-sample — I am reading it as "no case to stand down," not as a licence to conclude event entries are better.
3. Event risk is priced in the size, not avoided by it. The 15%-of-equity notional (~$3,750) was calibrated against the worst trade in my own history — −19.72% of notional, the 2020 COVID week, ≈ −3.0% of account equity — precisely because a fixed 5-day hold with no stop can swallow a multi-event week. A discretionary cut for this particular hold would be an untested deviation from the cleared mechanism, made on a forecast I cannot validate; if an event-day filter should exist at all, the path is a fresh spec through the pipeline and TEMPER, not a plan narrative. I will flag it to the desk log as a standing item worth a TEMPER conversation only if a real flag lands on an event-day entry and the outcome is tail-shaped — one observation is not evidence, and the pre-registered bar for revisiting a locked mechanism is not met by this week's calendar.

Adjustments: standard operating procedure — no changes. Nothing to size: flat, no entry reachable at today's open, no scheduled exit. The locked 15%-of-equity fixed-notional, single-position, no-stop/no-target/no-early-exit mechanism has no event-day or event-risk override, and HIGH event risk plus a firm-wide GO are gate-irrelevant when the cycle says no-signal.

What I'm watching:

1. The run settling at today's close — the only number that matters, deciding whether an entry fires at Wednesday 09-16's open. From −2.05% the down leg (~2.92pp of deepening) is the only remotely live path; it requires a crash-scale small-cap close that a contained two-tier repricing gap with VIX at 17.24 does not offer. My call: no flag.
2. IWM vs ~$292 / ~$300 — a decisive break below the multi-session floor on the AI-repricing + oil + 10-yr>5% stack is the realistic route to a down-flag (→ contrarian long) this window or next; a sustained catch-up toward $300 is the up-flag route in later windows. IWM is not Alpaca-owned today, so I bound rather than assert.
3. Whether an FOMC-eve repricing gap finally produces crash-scale small-cap extremity — the standing Phase 3 falsification test. The 09-11 sticky-hot CPI (core +0.3%, hike pricing to ~87%, a 2023-high 10-yr, oil >$100) failed to produce it, and today offers the same two-tier shape — an energy bid and a software bid offsetting a memory/semi cascade, which is exactly the composite-suppressing structure that has starved my signal all Phase 3. I want the flag; I do not expect it.

Invalidation: none applies — flat, no entry reachable at today's open (run −2.05% vs ±4.97%; a flag requires a several-percent single-session IWM close), no scheduled exit, nothing held over the FOMC. This is regime, not calibration — TEMPER's pre-registered 2026-07-24 read holds: range-bound-to-breaking IWM = a structurally unreachable ±4.97% threshold = expected behavior, not edge degradation. Today is the 52nd consecutive logged session since Phase 3 began (07-02 → 09-15; 51 unique dates logged through 09-14, one of them a duplicated 08-07 row already flagged for cleanup), roughly 3.0× the ~17-session expected inter-signal gap. The nearest Phase 3 graze remains 09-02's −2.95%, still 2.02pp away, with no cluster in either the −4.0%→−4.97% or +4.0%→+4.97% band. SPRT correctly sits at n=0, LLR +0.000, both boundaries (±2.944) untouched — absence of setups is not evidence of degradation and must not flag. No TEMPER conversation warranted. A no-trade day is the mechanism working as designed.

Plan-narrative correction (mechanic, not mechanism). The 09-11 and 09-14 plans framed a hypothetical entry's hold as "days 1–5 = entry day + 4 sessions, exit on the 5th," which is one session short: the locked script counts days_held as trading days strictly after entry (daily[(date > entry_date) & (date <= trade_date)]) and exits at days_held >= 5, and the cleared backtest confirms the convention (entry 2020-03-10 → exit 2020-03-17). So a 09-16 entry exits 09-23, not 09-22. Recorded because the hold-to-exit calendar is the one exposure fact this stack's plans exist to state accurately; it changes no gate and no size.

Plan Filed

• Filed: 2026-09-15 04:35 PDT (pre-market, House Slack desk session)
• Frontmatter forecasts complete for every active member: yes
• Firm-wide status: GO (HIGH event risk — FOMC decision + SEP + presser Wed 09-16, 1 day out; two-day meeting begins today) — gate-irrelevant for a flat swing stack: no position to hold, no exit due, no entry reachable
• Expected action: no trade — no_trade:no_signal (blocked=run_magnitude, script-logged run −2.05% at the 09-14 close vs ±4.97%; flat book, no 5-day cycle in flight; next entry window = Wed 09-16's open if today's close flags a run)

No trades taken.