Reversal Desk — 2026-09-25

desk: reversal date: 2026-09-25 forecasts: "null": # quoted deliberately — a bare null key parses as None in YAML p_trade: 0.18 direction: short conviction: low meridian: p_trade: 0.08 direction: short conviction: low

Desk Plan — Reversal — 2026-09-25

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-09-25-pm.md PM: Wicker (persona; own stack retired 2026-07-14 — this plan covers active members Null + Meridian only)

Regime Watch

Regime watch: data-infra/regime/current.json reads as_of 2026-09-22 — status ok: vix_regime LOW (<18, VIX 14.21), spy_drawdown_bucket <2% (−0.58%), spy_vs_sma50 above 50d, spy_day flat, spy_open_to_close_pct −0.08. Stated plainly: the file is present and readable but three sessions stale — written 2026-09-23 21:25, data as_of 09-22, so it contains neither 09-23 nor 09-24. Reported as read (not unavailable), treated as directional only; the live VIX print comes from the briefing — 15.36, still LOW (<18), −0.31 vs the prior close.

Charter consistency: all three shared tags sit at the charter's baseline state (LOW / <2% / above 50d). No charter indicator is moving off baseline; no non-baseline tag has held for the charter's 5-session persistence window; no Desk Review proposal is triggered. Today's plan is consistent with the ## Regime Charter's ratified direction — intraday only, no overnight hold, no regime-keyed sit-out invented for either member (neither carries one). Two sessions the stale file cannot speak to are worth noting for the record: 09-23 was a down day for both index legs and 09-24 recovered to roughly flat (QQQ closed 741.09 vs 741.15 cached 09-23) — nothing in that shape approaches the charter's persistent-shift rule, but the file should ideally be refreshed before the weekly reflection.

Shared Market Read

• Event risk today: MEDIUM — and again NOT a Tier-1 for either member, so neither event gate fires. Bucket A is clear through 09-29 (next Tier-1 is PCE Sep 30, in 5 days, then NFP Oct 2, CPI Oct 14; FOMC Minutes Oct 7, next decision Oct 28). is_event_day('2026-09-25', tier_filter=1) and is_event_day('2026-09-25', tier_filter=2) both return clean — verified this session, not assumed. Firm-wide trade status GO. The MEDIUM rating is headline risk: the Iran/Hormuz seven-day roadmap (Abbas Araghchi's conditional offer — reopen Hormuz at the end of day seven if conditions are met; White House unanswered) is a two-sided, un-timeable clock that runs across the weekend, plus the unchanged rates ceiling (10-year eased to ~5.16–5.19% off Thursday's ~5.208% near-20-year high; CME FedWatch ~77.5% priced for an October hike). Secondary: the 8:30 durable-goods block (cons. cluster −0.3% to −0.5% vs prior +1.1%) and Michigan final 10:00 (prelim 47.8).
• Session character expected: contained, narrow, tech-led gap-UP bias — direction still explicitly unassigned by the briefing, and polarity left unassigned in the instrument table per the 09-24 EOD's fix directive. Futures are modestly higher across all three majors with Nasdaq 100 contracts leading (+0.5–0.6% NQ vs +0.2–0.3% S&P/Dow), the bid narrow and catalyst-driven (chips: AMD/MRVL/INTC ~+2% pre-market; Meta the week's leadership single-name). Every one of the last four sessions resolved at or inside ~1.0× ATR — containment is the base case, not expansion. No QQQ/SPY pre-market print at 07:00 ET on the thin IEX feed: the cash-open gap is unquantified pre-open. I price it, I do not assume its size.
• VIX regime: 15.36 (LOW, <18) — the historically favourable cohort for both members (Null's backtest WR 80.1% in <18; Meridian's edge is concentrated there). No VIX gate exists on either stack. Tailwind note, not a signal.
• Key levels:
• QQQ: prior close 741.09 (Null's gap-fill target), prior-day range 734.66 – 742.63 (range 7.97), prior high 742.63 (Meridian's sweep level, +1.54 pts / +0.21% above the prior close), 20-day SMA 719.72 (price above SMA20 by 2.97%), ATR(14) 9.37.
• SPY: prior close 767.29, prior-day range 763.25 – 768.94 (range 5.69), prior high 768.94 (+1.65 pts / +0.215% above the prior close), 20-day SMA 765.33 (above by 0.26%), ATR(14) 6.70.
• IWM: no briefing block; the cached 1-min series ends 2026-09-23 (20-day average range 3.005 → 1.25× ceiling ≈ 3.76), so 09-24's range and today's prior-high level are unavailable. Meridian's IWM leg is unpriced pre-open; the engine's own precheck governs it at the open.
• Null's gap band off ATR 9.37 / prior close 741.09 (0.10× = 0.937 pts; 0.30× = 2.811; 0.70× = 6.559): gap-down LONG zone [734.53, 740.16] (−0.88% to −0.13%); sub-floor block (740.16, 742.03); tiny gap-up SHORT zone [742.03, 743.90] (+0.13% to +0.38%) — the relaxation path; gap-up ≥0.30× refused by the trend filter by rule [743.90, 747.65]; oversize ≥747.65 or ≤734.53.
• Macro backdrop (one sentence): a firm-wide GO on a no-Tier-1 Friday whose only live risk is a conditional geopolitical headline on a weekend-spanning clock — a rates-and-oil tape that has spent a week capped and contained, now carrying a narrow chip-led bid that points the index's opening bias up, which is the wrong direction for Null's cleared trend filter and the right direction for Meridian's bearish sweep geometry.

Both standing calibration commitments from the 2026-09-23 peer review remain discharged inline in this plan: the running p_trade-bucket frequency tables for both stacks (DELTA → Reversal), and Null's forecast conditioned on the gap-size band rather than set by narrative (VESPER → Reversal). Both are below, and both govern today's numbers.

Null — QQQ (gap fade / FVG; MNQ proxy, gap measured at the cash open)

Setup present: NO on the modal path — p_trade 0.18, direction SHORT (mechanical: the only live path is the tiny gap-up short), conviction low. Both calendar gates read clean — no Tier-1, and the engine's mega-cap earnings check returns clean — but for today that second "clean" is a data artifact, not a verification, and I am flagging it rather than resting on it (below).

1. The direction of the tape is up, and that is what binds. Futures are modestly higher with Nasdaq leading (+0.5–0.6% NQ). QQQ sits 3.0% above its 20-day SMA (741.09 vs 719.72), and the cleared HTF filter (strategy_v2.md rule 3) is explicit: above the SMA20, take gap-DOWN fades only; gap-up fades are refused above 0.30× ATR. The exception (a tiny gap < 0.30× ATR) is the only tradeable form on the up side today. So the day's live paths are narrow in a way the last several sessions' plans were not: the modal futures-implied open is in a refused band.
2. The conditioning table (VESPER's ask, maintained): fires by gap band, all logged Null sessions since 2026-07-14. 43 sessions carry a parsed gap_ratio — 15 fires (0.349). Sub-floor (<0.10×) N=7, fires 0; tiny (0.10–0.30×) N=13, fires 10 (0.769); in-band (0.30–0.70×) N=14, fires 5 (0.357); oversize (>0.70×) N=9, fires 0. The gate is doing exactly what it was cleared to do: the modal blockers are the two tails, not the band. Split by direction × SMA20, which is the split that matters today: tiny gap-up above SMA20 fired 5/5 (08-05 0.168, 08-17 0.158, 08-19 0.220, 09-08 0.219, 09-18 0.244 — all winners); in-band gap-up above SMA20 was refused 3/3 by the trend filter (08-04 0.563, 08-21 0.379, 08-27 0.527 — no_trade:trend_filter); gap-down above SMA20 fired 3/3 (08-06 0.427, 08-20 0.325, 08-28 0.135). Blocker census since 07-14: gap_filter 16, trend_filter 7, no_signal 5, event days 10. The trend filter is a real, recurring binder — not a hypothetical one.
3. The knife-edge, in numbers. Prior close 741.09, ATR(14) 9.37: tiny gap-up short 742.03–743.90 (+0.13% to +0.38%); refused band 743.90–747.65; sub-floor 740.16–742.03; gap-down long 734.53–740.16; oversize outside. The futures-implied QQQ open (+0.5–0.6% on NQ) is ≈ 744.5–745.5 — roughly 1–2 points inside the refused band. The tradeable path therefore requires the cash open to pare back to ≤+0.38%; a fully-held futures bias is a rule-based no-trade, and a pare-through to flat is a sub-floor no-trade. That asymmetry — refused band a point away, tradeable window 1.9 pts wide — is exactly what 0.18 prices.
4. Even a qualifying gap must still produce a drive. 09-24 gave the desk an in-band gap-down (0.659, dir=long) and the mechanism blocked at the pattern level: no_signal:pattern max_retrace_ratio=0.00 floor=0.50 drive_ok=False — no open drive existed to retrace, so the reversal impulse rule never armed. That is the fifth no_signal block since 07-14 and the second-recent reminder that a structurally-correct gap is a precondition, not a trade. On an up-gap open the drive is the usual shape, so this leg is less of a handicap today than it was yesterday — it is still a live way to spend a qualifying open on nothing.
5. The counter-weight, stated honestly. The last four eligible opens read 0.786 (block) → 0.244 (fire) → 0.045 (sub-floor) → 0.038 (sub-floor) → 0.659 (pattern block) — a mixture of oversize, compression and one mechanism miss. Nothing in that run says the band is favourable today; it says the gate stack is doing its job. What is new today is that the band the tape is likely to occupy is on the refused side, for a reason (trend alignment) that has blocked 7 sessions this quarter.
6. The one filter that can close the session before the band is even measured: pre-market NQ volume > 2× its prior 5-day average (no_trade:premarket_volume, strategy rule 4) — a live path on a headline-clock day, and one I cannot measure from the IEX feed I have.

⚠️ Live-vs-cleared divergence — flagged, not fixed (no gate change). strategy_v2.md rule 5 names COST among the QQQ top constituents and says: no entry on any session where one of them reported earnings after the prior session's close or pre-market the same day. Costco reported FQ4 after yesterday's close (EPS $6.75 vs ~$6.54 est., comps +9.4% — a beat; the stock gave back its after-hours pop to settle near flat). Read plainly, today is a skip-eligible session for Null. The live engine will not skip it: data/event_calendar/qqq_earnings_dates.csv is stale through 2026-06-03 (last row AVGO 06-03), so is_event_day('2026-09-25', tier_filter=2) returns clean — verified this session. I am not touching the gate; the standing rule is no unilateral parameter or gate change. Three consequences I am recording rather than acting on: (a) my p_trade prices the ENGINE's behaviour (eligible), not the cleared rule's (skip) — the two would differ materially today, and the forecast is scored against trades.csv; (b) this is the session where the stale CSV stops being a footnote and starts being load-bearing — the desk flagged it as a transparency note on 09-24, and today it is the difference between a tradeable and a skipped session; (c) the same stale file will miss Micron's FQ4 on 09-30 (MU is not on the named list, so that one is a residual-vol question rather than a hard skip, but it lands on a PCE day and the desk should carry it deliberately). Considered as v2 rule 6's territory — the log-only T+1 residual-vol monitor — today is exactly the sample that monitor was built for.

Sizing vs event risk: standard — 1 MNQ, 0.5% risk, one position at a time, 11:00 ET hard exit. No changes. Not upsized on the LOW-VIX cohort; not downsized for MEDIUM headline risk. Null's cleared rule set has no headline skip and I am adding none.

Key levels: QQQ prior close 741.09 (fill target), prior high 742.63, prior low 734.66, SMA20 719.72 (price +2.97% above — the filter that decides which gaps are even eligible), ATR(14) 9.37. Live zones: [734.53, 740.16] (gap-down LONG) ∪ [742.03, 743.90] (tiny gap-up SHORT); refused by rule [743.90, 747.65]; sub-floor (740.16, 742.03).

Invalidation / stand-down: stand down if the cash open prints ≥743.90 (gap-up above the tiny threshold — refused by the trend filter), in (740.16, 742.03) (sub-floor), ≤734.53 (oversize), ≥747.65 (oversize up), or if pre-market NQ volume runs >2× its 5-day average. The 11:00 ET hard exit is unchanged. A refused-band or sub-floor block is not a near-miss and must not be read as one — it is the cleared filter working.

Mechanism health: SPRT CONTINUE — n=15 W14/L1, LLR +2.461, cumulative +$1,965.99, WR 93.3%. First-loss watch stays N=1 (the 09-03 trend-day fade remains un-clustered; no filter conversation until same-shaped losses reach N ≥ 3). No tweak.

Null — p_trade-bucket calibration table (DELTA's ask; 49 scored plan days since 07-14, 13 fires, realised 0.265, mean p 0.244, mean Brier 0.160):

Read: the higher buckets are genuinely better-calibrated (0.40–0.49 → 0.625), while 0.20–0.29 is the over-priced zone (realised 0.143–0.167 over 13 days). 0.18 sits in the 0.15–0.19 bucket, which has realised 0.400 on N=5 — so today's price is on the conservative side of its own bucket, which is the right side to err on when the binding gate is a rule that has fired 7 times this quarter. Provenance note (verified this session): Null's rows in intelligence/plan-calibration.csv are recorded with a blank stack field (49 rows neither meridian nor wicker) — the quoted "null" frontmatter key in the 09-24 plan did not restore the label, though the p_trade / traded / brier columns scored correctly. The table above is built from those unattributed rows; the label loss is a tooling artifact for the operators, not a scoring failure, and I am recording it rather than silently working around it.

Meridian — QQQ/SPY/IWM (kill-zone prior-high sweep + RSI(2) ≥ 88, bearish only)

Setup present: NO — but this is the best pre-session geometry the stack has had in weeks, and I am pricing the mechanism, not the geometry. p_trade 0.08, direction SHORT (mechanical — the stack is bearish-only), conviction low.

1. The base rate is the forecast. Meridian has fired 3 times in 73 logged sessions (0.041) and once in the last 28 sessions (0.036). The p_trade-bucket table below is blunt: 23 plan days priced at 0.20 or above produced zero fires; both scored fires landed in the 0.10–0.19 range. Anything at or above 0.20 without a new mechanism claim is optimism dressed as a forecast. 0.08 is deliberately above the unconditional 0.041 — the pre-session gates look better than usual today (points 2–4) — and deliberately below the desk's historic 0.20 habit, which the evidence has never supported on this stack.
2. The volatility gate (Step 1) has likely RELEASED, and that is the change. It reads 09-24's range. QQQ: prior range 7.97 vs a 1.25× ceiling of ≈9.0 (my reconstruction off the cached 1-min series: 20-day average range 7.229 through 09-23, ceiling 9.036) → passes with ~12% cushion. SPY: prior range 5.69 vs ceiling ≈6.33 (engine's own 20-day average read 4.99 on 09-24 → ceiling 6.24) → passes with ~9–10% cushion. Note the direction of travel: on 09-24 the gate blocked SPY and IWM and released only QQQ; today two of three legs look released. IWM is unprojectable — no briefing block, cache ends 09-23 — and its own 20-day average range (3.005 → ceiling ≈3.76) was already exceeded by 09-23's 4.34, so IWM has been the wide leg and I price it as more likely blocked than released.
3. The ADX band is unknowable pre-open, and I will not invent a number. Step 2 is computed at the open off a continuous 15-min series that includes today's pre-market. Nearest anchors: QQQ read 27.9 on 09-24 — inside 18–32 (09-23 it was 38.4, above); SPY and IWM were blocked at Step 1 on 09-24, so no read exists for them. The honest statement is that QQQ has the best chance of landing inside the band and the other legs are unknown. Note the mechanism's own tension for today: a one-way chip-led up-open is friendly to forming a sweep but lifts ADX, and the band's 32 ceiling is the gate that would bite exactly then.
4. The sweep geometry finally points the right way — the load-bearing new fact. QQQ's prior high 742.63 sits only +1.54 pts (+0.21%) above the prior close 741.09; SPY's 768.94 sits +1.65 pts (+0.215%) above 767.29. These are the tightest levels this desk has had in weeks (09-24: QQQ's level was 9.02 pts away on a gap-down open, and the blocker was no_sweep). More importantly, the pre-market tape is higher and chip-led — pointing INTO the levels, which is the standing 2026-06-18 lesson satisfied for once: "reachable in principle" is not "likely to be reached", and a down-biased open is the wrong structural backdrop for an upside sweep. Today the backdrop is the right one.
5. But exhaustion is still the binder, and its live record is stark. Since 07-20 the engine has logged 11 sweep legs — and zero RSI(2) ≥ 88 prints. The near-misses: IWM 84.2 (09-21), SPY 80.0 (08-17), 78.4 and 73.0 (08-10), 59.3 (09-02), 51.6 (07-22), 46.5 (08-25), 40.4 (07-21), 23.9 (07-27), 11.1 (09-18), 4.5 (07-20). On 09-18 QQQ swept twice and the RSI(2) read 11.1 against 88 — a 76.9-pt shortfall, not a graze. So "a sweep forms" has moved from improbable to plausible; "a sweep forms with genuine exhaustion" is 0-for-11 live. That ratio — not the level proximity — is what 0.08 prices.
6. What a live setup would require today: a leg inside ADX 18–32 and volatility-released, price pushing up into a prior high 0.21% away, a spike through prior_high that closes back below it in the same 5-min bar with RSI(2) > 88, then a bearish 1-min FVG within 15 minutes. Three legs to try, two of them plausibly gate-released for the first time in a week — and the pattern-abandonment variants from 09-18/09-22 already tell me not to read a sweep as a trade.

Sizing vs event risk: standard — $250/leg, caps 10 MNQ / 5 MES / 5 M2K, 2-loss daily cap, hard close 11:30 ET. No changes. A no-trade forecast is not a reason to size down, and a MEDIUM headline day is not a reason to alter a standard that already assumes normal conditions. iwm_leg remains n=0 — the honest live anchor for IWM is ~56%, not the 73.3% headline.

Key levels: QQQ prior high 742.63 (sweep level, +0.21% from prior close 741.09); SPY prior high 768.94 (+0.215% from 767.29); IWM unavailable. ADX band 18–32 (nearest read: QQQ 27.9 on 09-24; 09-23 was QQQ 38.4 / SPY 16.1 / IWM 15.7). Volatility ceilings ≈ 9.0 QQQ / 6.24–6.33 SPY / 3.76 IWM (1.25× the 20-day average range; engine and my cache differ by <2%). ATR(14) daily QQQ 9.37 / SPY 6.70 → the stop is the engine's own 2.0× ATR(5m) with the same-sign guard voiding any malformed stop.

Invalidation / stand-down: stand down if the ADX precheck lands outside 18–32 on every leg (the base case I half-expect, and I have no pre-open read), or if the volatility precheck blocks all legs (IWM most likely), or if a leg sweeps but never prints a close-back-below bar with RSI(2) ≥ 88 (0-for-11 live), or if no leg reaches its prior high. The 11:30 ET hard close is unchanged. I do not touch the 88 line: it is cleared-parameter territory and only TEMPER can move it.

Meridian — p_trade-bucket calibration table (DELTA's ask; 49 scored plan days, 2 fires, realised 0.041, mean p 0.158, mean Brier 0.065):

Read: 23 plan days priced ≥0.20 produced zero fires. Every fire the desk has ever scored on this stack landed at 0.10–0.19. (The scored table shows 2 fires vs. 3 in trades.csv — the 2026-07-10 QQQ fire predates the calibration file's 07-14 start; noted so the counts reconcile.) The desk's mean p of 0.158 against a realised 0.041 is the honest diagnosis of the historical habit, and 0.08 is the correction carried forward from the two peer-review dispositions.

Standing tolerance (DELTA's ask, enforced): if Meridian produces two consecutive non-fires while this desk calls the setup "one gate away", the label reverts from regime to calibration and the price drops — with the next two sessions priced from the realised bucket frequency alone. A single Brier score is not cited as validation; it is one observation in the table above.

Plan Filed

• Filed: 2026-09-25 (pre-market, 04:44 PT / 07:44 ET)
• Frontmatter forecasts complete for active members only (Null, Meridian): yes. Wicker (PM): strategy retired 2026-07-14 (WS2c Tier-2 NO-GO) — no plan, no forecast, no per-member file; persona continues desk oversight only.
• Firm-wide GO; event risk MEDIUM (no Tier-1 through Sep 29 — next PCE Sep 30 — but the Iran/Hormuz clock is two-sided and runs across the weekend). Both members' Tier-1 event gates are clean (verified). Null p_trade 0.18 / direction short / conviction low; Meridian p_trade 0.08 / direction short / conviction low.
• Base-case desk outcome: Null no-trade, on a rule-based block — the cash open lands in the refused gap-up band (≥743.90, trend filter) most likely, or sub-floor/oversize — with the tradeable path requiring the open to pare to ≤+0.38% and a drive-and-retrace to form. Meridian no-trade, on adx_precheck outside 18–32 (no pre-open read), the IWM volatility block, or rsi_reject behind a sweep that does form — with the caveat that today is the first session in weeks where the sweep geometry is genuinely favourable (levels +0.21%, open biased toward them).
• One thing I am watching, both stacks: whether the cash open pares back from the +0.5–0.6% NQ futures bias. For Null that single number decides refused-band versus tiny-gap-short; for Meridian a modest up-open that holds is the backdrop its sweep needs. The second thing: QQQ's ADX at the open — 27.9 yesterday means it can be inside the band, and the band is Meridian's first gate.
• Divergence carried, not acted on: today is a COST-reporting session by the cleared v2 rule's plain text, but the live engine's stale qqq_earnings_dates.csv (through 2026-06-03) reads it clean. Recorded for TEMPER/ZEUS; no gate touched. This is the first session where that staleness is load-bearing rather than a footnote.
• Reconfirmation ledger: no open item gates today. Meridian IWM expansion CLOSED 2026-08-20 (strategy_v7); standing ~6-month decay re-backtest next due ~2027-02-20. Wicker's feed/session re-run remains the desk's highest-priority open reconfirmation (own stack retired, not a today blocker).
• Method note (transparency, no execution impact): every pre-session number that actually matters is the engine's own precheck at the open. The volatility ceilings, band edges and gap zones here are my own arithmetic off the briefing's authoritative prices and the cached 1-min series (data/market_data/assay/*_1min_sip_exthrs.parquet, last cached session 2026-09-23 — the cache does not contain 09-24, so QQQ/SPY ceilings use the briefing's 09-24 ranges and IWM is marked unavailable). Nothing here touches execution; no member's gate stack was read from or written to.
• Peer-review disposition carry-in (2026-09-23, both "acted"): DELTA's p_trade-bucket tables → maintained above for both stacks, updated through 09-24; VESPER's conditioning ask → Null's number is set off the gap-band conditioning (tiny 0.769 / band 0.357, split by direction × SMA20), not by a habitual 0.20; Meridian's price stays anchored to its realised fire rate (0.041), with the conditional uplift argued explicitly rather than assumed.

Reversal Desk Reflection — 2026-09-25

PM: Wicker (persona; own stack retired 2026-07-14 — desk covers active members Null + Meridian only) Briefing reference: dadbrain/Analysis/briefings/2026-09-25-eod.md Cadence rule: weekly (Friday) or same-day-on-trade — WS5, in effect 2026-07-11. Today is a Friday weekly reflection AND Null traded — both triggers fire. Period covered: 2026-09-21 through 2026-09-25 (prior desk reflection 2026-09-18). Five sessions. One trade — Null's 09-25 QQQ gap-fade short, a win. Four Null no-trades (three gap-size blocks, one pattern-level block) and five Meridian no-trades. Desk scope net: +$122.10.

What Happened — Desk Summary

One trade, one win, four no-trade Null sessions and five Meridian sessions — and the period resolves to regime on both stacks, with one qualification I am naming rather than burying: Meridian's ADX floor was repeatedly approached this window (five leg-blocks below 18 across three sessions, the closest 1.6 points under), which is the desk's first genuinely calibration-shaped near-miss cluster on that stack. The tape was contained throughout — every session resolved at or inside ~1.0× ATR, VIX fell from 15.67 to 14.87 (LOW) — and that single fact drives both members' period. Null's binder was gap SIZE, on both tails, and it released: 09-21 gapped up 6.675 pts (0.786× ATR — oversize, 0.086 over the 0.70 ceiling), 09-22 gapped down 0.420 pts (0.045× — sub-floor, 0.055 under), 09-23 gapped down 0.350 pts (0.038× — sub-floor, 0.062 under), 09-24 finally printed an in-band 0.659× gap-down — and the gap gate passed it, only for the pattern gate to block (max_retrace_ratio=0.00 floor=0.50 drive_ok=False, the fifth no_signal block since 07-14) — and then 09-25 printed the tiny in-band gap-up the plan named as its only live path (0.171×, above the SMA20), and Null shorted 744.155 @ 10:01, hit the prior close 741.09 @ 10:14 for +$122.10. Meridian never assembled its conjunction: 09-21 the band blocked above the ceiling on both index legs (QQQ 42.5 / SPY 45.0) while IWM swept ten times and peaked at RSI(2) 84.2 against 88 — the closest exhaustion approach on this stack's record (3.8 points short); 09-22 the volatility gate blew wide on QQQ (15.41 vs 7.06 avg) and SPY (8.86 vs 4.88) with IWM ADX 11.8; 09-23 all three legs blocked at the ADX precheck (38.4 / 16.1 / 15.7); 09-24 QQQ reached the sweep stage for the first time in the window (volatility cleared, ADX 27.9 inside the band) and no_sweep bound with the prior high +9.02 away; 09-25 the volatility gate released on all three legs — exactly as the plan forecast — and the binder moved downstream to the ADX floor (QQQ 16.4, SPY 14.9) and to exhaustion (IWM sweeps=2 max_rsi2=29.9 thresh=88). Both SPRTs CONTINUE — no boundary crossed.

• Null — TRADED 09-25, WIN +$122.10 (short QQQ, entry 744.155 @ 10:01, target 741.09 @ 10:14, 13 minutes). The trade is the tiny-gap relaxation path, the exact path the plan enumerated and priced: QQQ prior close 741.09, cash open 742.64 (+1.55 pts / +0.21%, gap_ratio 0.171, gap_up=True above_sma20=True) — inside the tiny gap-up short zone [742.03, 743.90] and below the 0.30 relaxation line, so the trend-alignment filter never engaged. It sold the morning pop (QQQ ran to a 745.84 high, then faded to the prior close) and hit target in thirteen minutes. SPRT CONTINUE — n=16 W15/L1, LLR +2.673, cumulative +$2,088.09, WR 93.8%, now 0.271 from the CONSISTENT-WITH-BACKTEST boundary (+2.944) — the closest this book has been to a decided boundary. First-loss watch stays N=1 (the 09-03 trend-day fade remains un-clustered).
• Meridian — NO-TRADE all five sessions. Binder sequence: 09-21 adx_precheck above the ceiling (QQQ 42.5 / SPY 45.0) with IWM sweep_chain sweeps=10 rsi_reject=10 max_rsi2=84.2 thresh=88 — the mechanism's closest approach to its authorizing shape in its entire live record; 09-22 volatility on QQQ and SPY, adx_precheck 11.8 on IWM; 09-23 adx_precheck on all three (38.4 / 16.1 / 15.7); 09-24 QQQ no_sweep (prior high 747.13 vs day high 738.11, +9.02 away, adx_open 27.9 — inside the band, so volatility had cleared), SPY volatility (6.55 vs 4.99), IWM volatility (4.34 vs 2.95); 09-25 adx_precheck QQQ 16.4 / SPY 14.9, IWM sweep_chain sweeps=2 rsi_reject=2 max_rsi2=29.9. SPRT CONTINUE — n=3 W2/L1, LLR +0.222, cumulative +$385.22, WR 66.7%; spy_leg n=1 WR 100%; iwm_leg n=0 all period.

Forecast calibration for the period (rows 09-21 → 09-25, intelligence/plan-calibration.csv): Null brier 0.04 / 0.0625 / 0.04 / 0.09 / 0.6724; Meridian 0.0225 / 0.0025 / 0.04 / 0.0025 / 0.0064. Null's 0.6724 is the same shape the desk scored on 09-18: a low-p call that fired and won (0.18 forecast, traded, direction short correct). Per the 09-23 peer-review disposition I am no longer citing a single Brier as validation — it is one observation, and it lands in a bucket (0.15–0.19) that now reads N=6 plan days, 3 fires, realised 0.500, i.e. the bucket is on the conservative side of what it has realised.

Regime Watch

Regime watch: data-infra/regime/current.json reads as_of 2026-09-22 — status ok: vix_regime LOW (<18, VIX 14.21), spy_drawdown_bucket <2% (−0.58%), spy_vs_sma50 above 50d, spy_day flat, spy_open_to_close_pct −0.08. Stated plainly: the file is present and readable but now three sessions stale — written 2026-09-23 21:25 with data as_of 09-22, so it contains none of 09-23, 09-24 or 09-25. Reported as read (not unavailable), treated as directional only; the live print is the briefing's — VIX closed 14.87, LOW (<18), −0.80 vs the prior close 15.67. This file has now missed three consecutive sessions across two reflections; it should be refreshed before the next one, and I am recording that rather than working around it.

Charter consistency: all three shared tags sit at the ## Regime Charter's baseline state (LOW / <2% / above 50d). No charter indicator is moving off baseline, no non-baseline tag has held for the charter's 5-session persistence window, and no Desk Review proposal is triggered. Today's plan and this reflection are consistent with the charter's stated direction — intraday only, no overnight hold, no regime-keyed sit-out invented for either member (neither carries one). Positively: today's tape — contained, mildly-up, Dow-led, both legs 0.65–0.87× ATR — is precisely the charter's clause-1 "ordinary two-sided, non-trending session in a calm tape" in which both members expect to make money, and Null made money in exactly that way (fading the morning pop into the prior close). Meridian's clause-1 shape was structurally absent: the ADX band never opened while a prior high sat within reach, and its clause-2 selloff case did not apply (the tape was up). Charter status note: DRAFT — not ratified. I am holding to its stated baseline direction, as the desk's plans have since 09-23, and flagging the draft status rather than treating it as ratified.

Desk Performance Summary

Desk net paper P&L (scope): +$122.10 (Null's 09-25 win; Meridian $0). No SPRT boundary crossed.

Per-Member Binding Gate Analysis

Null — TRADED 09-25 (WIN). Binding gate through the period = gap SIZE on both tails, then one pattern-level block; REGIME, released, and converted

Null's binder this period was the gap-size filter working its two-sided job and then handing the mechanism a setup it took: 09-21 gap_ratio val=0.786 band=[0.1,0.7] (gap +6.675, ATR 8.49 — a modest oversize, 0.086 past the ceiling, not the 2× overshoots of 09-14/09-17); 09-22 0.045 (−0.420, ATR 9.28) and 09-23 0.038 (−0.350, ATR 9.19) — two consecutive sub-floor prints, misses of 0.055 and 0.062 against the 0.10 floor; 09-24 0.659 (in-band gap-down, above the SMA20 — the row my conditioning table said had fired 3/3) where the gap gate PASSED and the pattern gate blocked (max_retrace_ratio=0.00 floor=0.50 drive_ok=False — no open drive existed to retrace, so the reversal-impulse rule never armed); then 09-25 0.171 (tiny gap-up, above the SMA20) — fired, short 744.155, target 741.09, +$122.10.

• Regime, not calibration. The three gap blocks were one modest oversize (0.086 past the ceiling) and two sub-floors — the tails, in opposite directions, not one threshold grazed from one side. And the sub-floor misses are not a threshold problem in the way a graze is: a 0.038–0.045× ATR gap is a ~0.4-point open on a 9.2-point ATR — there is no imbalance there to fade, which is exactly why the floor exists. Widening the floor would admit sessions in which the mechanism's precondition is absent. The release pattern is the same one the 09-18 reflection recorded: when the tape finally produced a qualifying gap, the mechanism took it and won. No gate conversation warranted.
• The one caveat I will not leave implicit: the sub-floor cluster is now N=4 since 09-01 (0.081 on 09-02, 0.053 on 09-15, 0.045 on 09-22, 0.038 on 09-23), all misses of 0.019–0.062 against the 0.10 floor. I read that as compression — a contained LOW-VIX tape handing out near-zero overnight gaps — and not as evidence the floor is mis-set, because the observations cluster where the mechanism has nothing to trade rather than just short of a line it could cross. But the desk's job is to say which it is, not to let "regime" be unfalsifiable (the 09-23 DELTA critique), so I am naming it and stating what would flip it: if the sub-floor cluster reaches N ≥ 6 within a rolling 20 sessions while the modal open is sub-floor, the floor becomes a TEMPER question — a question, filed by the desk, never a tweak by the desk.
• The period's real calibration finding is mine, not the gate's. My 09-24 price of 0.30 was built off a conditioning table whose rows — "tiny gap-up above SMA20 fired 5/5", "in-band gap-down above SMA20 fired 3/3" — measure P(gap gate passes), not P(trade). There is a second gate behind it. Since 07-14 there have been 19 gap-gate passes: 14 fires and 5 no_signal:pattern blocks → a conditional fire rate of ≈ 0.74. On 09-24 the difference was the entire call: the gap gate passed at 0.659 and the pattern blocked. The honest structure is P(trade) = P(gap gate passes) × P(pattern passes | gap gate passes) ≈ 0.74, and the next plan's number will be built that way. That is a forecast fix, not a gate change.
• Also a rule-divergent win, and I am saying so. Today was skip-eligible by the cleared strategy_v2.md rule 5's plain text — Costco reported FQ4 after the prior session's close — while the live engine read the session clean because data/event_calendar/qqq_earnings_dates.csv is stale through 2026-06-03. The engine traded; the trade won; the win does not validate the staleness. Had the engine honoured the cleared rule the desk would have logged $0 and my 0.18 would have scored clean, which is a statement about the CSV, not about the trade. Flagged, not fixed (no unilateral gate or calendar change) — see Null's member reflection.

Meridian — NO-TRADE all five. Binding gate = the ADX band (floor side dominant), with volatility and no_sweep as the earlier binders; REGIME for the outcome, with the floor-side cluster named as this period's one calibration-shaped candidate

Meridian's binder moved through the period and never sat on one gate: 09-21 the ADX ceiling (QQQ 42.5 / SPY 45.0) while IWM ran a ten-sweep chain to RSI(2) 84.2 vs 88 — a 3.8-point shortfall, the closest this stack has ever come to authorizing; 09-22 volatility wide on QQQ (15.41 vs 7.06 avg — 1.74× the ceiling) and SPY (8.86 vs 4.88 — 1.45×), IWM 11.8 below the ADX floor; 09-23 adx_precheck on all three legs, above the ceiling on QQQ (38.4) and below the floor on SPY (16.1) and IWM (15.7); 09-24 QQQ no_sweep at +9.02 with the band inside (27.9) for the first time in the window, SPY/IWM volatility-blocked; 09-25 the volatility gate released on all three legs (as the plan forecast) and the binder moved downstream to the ADX floor (QQQ 16.4, SPY 14.9) and to exhaustion (IWM 29.9).

• Regime for the trade outcome. The authorizing conjunction — a leg inside ADX 18–32, volatility-released, price pushing up into a prior high, a spike through it closing back below in the same 5-min bar with RSI(2) ≥ 88, then a bearish 1-min FVG — never assembled on any leg on any session. Where a sweep did form, exhaustion failed by 3.8 points (09-21 IWM) and 58.1 points (09-25 IWM) — one near-approach and one categorical absence, not a repeated graze of the same line. Where the band was inside (09-24 QQQ), a different gate bound. Exhaustion is now 0-for-13 live sweep legs since 07-20.
• But the ADX floor is the one thing this period I will not file silently under "regime." The band's floor blocked five leg-instances across three sessions (09-22 IWM 11.8; 09-23 SPY 16.1, IWM 15.7; 09-25 QQQ 16.4, SPY 14.9) — the closest 1.6 points under 18 and 1.9 points under 18, on a threshold that has now blocked the floor side in six of the last eight sessions. That is the shape the desk's own definition calls calibration rather than regime, and given the 09-23 peer review's charge that this desk labels every miss "REGIME" in a way nothing could contradict, the consistent move is to name it: I am carrying the 18 floor to TEMPER as a question — is it calibrated for a contained LOW-VIX tape in which 15-min ADX structurally sits in the 11–17 band on orderly days, or is it describing a state that no longer occurs? — and I am not touching it. Two pieces of counter-evidence go with the question, because they bound its value: (1) on 09-24 the band was inside and no_sweep bound anyway, so opening the floor would not by itself have produced trades; (2) exhaustion remains the deepest binder (0-for-13), so a floor change would buy access to a gate stack whose last filter still refuses. The escalation is honest, not opportunistic.
• The band is also unstable session-over-session, which is a separate finding. QQQ read 27.9 on 09-24 (inside) and 16.4 on 09-25 (below the floor) — an 11.5-point swing between two sessions of near-identical character (both contained, both sub-ATR, SPY +0.34% / +0.53%). An inside-band read one day therefore carries almost no information about the next, which matters because the desk's plans lean on the nearest ADX read as a guide. Flagged for learnings.md in Meridian's member reflection.
• The geometry finding, stated precisely, because it is the mechanism's structural problem in this regime. The plan called 09-25 the best pre-session sweep geometry in weeks (QQQ's prior high 742.63 only +0.21% above the prior close) and the geometry was confirmed — QQQ's 745.84 session high cleared the prior high, and IWM swept twice. But QQQ never reached its sweep chain (the ADX precheck stood the leg down first) and IWM's sweeps never approached exhaustion. So on the two sessions where the level was close enough to sweep (09-24, 09-25) the band was either inside-with-the-level-9-points-away or below-the-floor-with-the-level-0.21%-away: on this tape, "level within reach" and "ADX inside 18–32" have not co-occurred. That mutual exclusion, not any single threshold, is the honest reason this stack has fired 3 times in 75 sessions.
• Tolerance status (DELTA's ask, carried from 09-24): armed, not triggered. The desk did not call 09-25's setup "one gate away" — the plan explicitly priced the mechanism rather than the geometry, and wrote that "a sweep forms with genuine exhaustion" is 0-for-11 live. So the two-consecutive-non-fire tolerance does not fire, and I am stating that rather than letting it lapse.

Cross-Member Synthesis

Narrative finding: one tape, two members, and the period's single trade came from the mechanism whose gate finally released — while the member whose geometry finally lined up was blocked by a different gate one step earlier. Null's suppression this period was gap size, on both tails, and it broke the moment a qualifying gap appeared; Meridian's was the ADX band, and the new detail is that the band's floor, not its ceiling, is now the frequent binder on a contained tape. Neither member's live gate stack should be touched, and no parameter is being moved by this desk. What is new and worth carrying: Meridian's two closest approaches in its live history are now 09-21 (IWM RSI(2) 84.2) and 09-25 (a confirmed sweep geometry that the ADX floor pre-empted) — the mechanism is getting to its last gate and being stopped there, not being stopped by the absence of a setup.

Peer-Review Disposition Check (2026-09-23, both "acted")

• DELTA → Reversal (constant p_trade, unfalsifiable "regime"): discharged and enforced. The p_trade-bucket tables for both stacks have been maintained in every plan since 09-24; Meridian's price was re-anchored to its realised fire rate (0.20 on 09-23 → 0.05 on 09-24 → 0.08 on 09-25); a stated tolerance (two consecutive non-fires while the desk calls the setup "one gate away" → the label reverts to calibration and the price drops) was written into the 09-24 plan and is armed, not triggered today; and no single Brier is cited as validation. This reflection applies the same discipline to the ADX floor rather than filing it as regime.
• VESPER → Reversal (Null's price anchored on the last fire, not the gap distribution): discharged. Null's number is set off the gap-band conditioning split by direction × SMA20, not a habitual 0.20 — and this period's outcome shows the next correction the ask implies: my conditioning table measured the gap gate, not the trade, which is why 09-24's 0.30 missed (see the Null section).

Open TEMPER Items (Carried Forward)

• Meridian — ADX band floor (18), NEW this period. Five leg-blocks below the floor across three sessions (closest 1.6 pts under), and floor-side blocks in six of the last eight sessions. Carried to TEMPER as a question about whether the floor is calibrated for a contained LOW-VIX tape — with the counter-evidence that on 09-24 the band was inside and no_sweep bound, and that exhaustion (0-for-13) is the deeper binder. No gate touched by this desk.
• Meridian — exhaustion (RSI(2) ≥ 88) remains the deepest binder (strengthened). 09-21 IWM reached 84.2 — the closest approach on record (3.8 short) — and 09-25's sweeps reached only 29.9. Sweeps form; exhaustion does not. Monitoring only.
• Meridian — ADX band instability, flagged for learnings.md. QQQ 27.9 (inside) → 16.4 (below floor) across two sessions of near-identical contained character.
• Null — stale qqq_earnings_dates.csv (through 2026-06-03) now load-bearing. 09-25 was skip-eligible by the cleared v2 rule's plain text (COST reported after the prior close) and the engine traded it; the trade won, which validates nothing. The same staleness will not see Micron's FQ4 on 09-30, which lands on a PCE day. Flagged for learnings.md; no calendar or gate change made.
• Null — forecast conditioning error, self-identified. The gap-band table measured P(gap gate passes); the trade rate is that multiplied by P(pattern | gap gate) ≈ 0.74 (14 fires / 19 gap-gate passes since 07-14). Fix carried into the next plan.
• Null — sub-floor compression cluster, N=4 since 09-01 (0.081 / 0.053 / 0.045 / 0.038, misses 0.019–0.062 under the 0.10 floor). Read as regime (compression, nothing to fade); tripwire stated — N ≥ 6 within a rolling 20 sessions while the modal open is sub-floor → a TEMPER question.
• Null — first live loss + 2026-WR watch (carried, N=1). The 09-03 trend-day fade (−$192.61) remains un-clustered; the period added a win. n=16 W15/L1, WR 93.8%, SPRT LLR +2.673, 0.271 from the CONSISTENT-WITH-BACKTEST boundary — the closest this book has been to a decided outcome; the 30-trade 2026-WR review threshold is not approached.
• Meridian — iwm_leg sub-book monitoring (carried). n=0 all period (ADX 11.8/15.7 below, volatility 4.34 wide, and on 09-25 a two-sweep chain that failed exhaustion at 29.9). Honest live WR anchor ~56%, not the 73.3% headline.
• Meridian — standing ~6-month periodic re-backtest decay check (carried). Next due ~2027-02-20.
• Wicker — backtest re-run under corrected feed/session config (carried). Standing highest-value reconfirmation; unchanged (own stack retired).

Reconnaissance Ledger

• Wicker backtest re-run under corrected (2026-07-05) feed/session config — open.
• Standing rule: no parameter re-tuning / gate loosening to manufacture throughput — unchanged and observed this period. Null's three gap blocks were tails in both directions and the mechanism converted the first qualifying gap it was given; Meridian's blocks were a moving binder with one named calibration-shaped cluster carried to TEMPER as a question. Nothing was touched on either live stack.
• data-infra/regime/current.json — stale for three sessions across two reflections. Refreshing it is an operational item, not a desk action; recorded so it is not silently worked around again.

Desk Summary for Orchestrator

One trade, one win, both members REGIME — with one named qualification. Null traded and won: blocked at the gap-size filter on 09-21 (0.786 oversize, 0.086 past the ceiling), 09-22 (0.045 sub-floor) and 09-23 (0.038 sub-floor), then on 09-24 the gap gate passed (0.659 in-band gap-down) and the pattern gate blocked (max_retrace_ratio=0.00 floor=0.50), then 09-25 printed the tiny in-band gap-up (0.171, open 742.64 vs prior close 741.09, above the SMA20) — short 744.155 @ 10:01, target 741.09 @ 10:14, +$122.10, SPRT CONTINUE n=16 W15/L1 LLR +2.673 (0.271 from the boundary), WR 93.8%, cumulative +$2,088.09, first-loss watch N=1. Meridian no-trade all five: ceiling-side ADX blocks (42.5/45.0) with IWM's record-closest exhaustion at 84.2 vs 88 on 09-21; volatility-wide on 09-22 and 09-24; floor-side ADX blocks on 09-23 and 09-25 (16.4/14.9); no_sweep +9.02 on 09-24; IWM sweeps=2 max_rsi2=29.9 on 09-25 — SPRT CONTINUE n=3 W2/L1 +0.222, iwm_leg n=0, cumulative +$385.22. Desk scope net +$122.10. Carried to TEMPER as a question (nothing touched): Meridian's ADX floor (18) — five leg-blocks below it across three sessions, closest 1.6 pts under, floor-side blocks in six of the last eight sessions — with the counter-evidence attached that the band was inside on 09-24 and a different gate bound. Flagged for learnings.md: the stale qqq_earnings_dates.csv made a skip-eligible session tradeable (the trade won; that validates nothing) and will miss Micron on 09-30; Meridian's ADX band swung 27.9 → 16.4 between two near-identical contained sessions. Watched: Null — whether the pattern gate keeps eating gap-gate passes (5 of 19 since 07-14) and whether the sub-floor compression cluster (N=4 since 09-01) persists into next week's PCE/NFP window; Meridian — whether "level within reach" and "ADX inside 18–32" ever co-occur on a contained tape.

Reflection Filed

• Filed: 2026-09-25 (EOD)
• Per-member compatibility files written: agents/null/reflections/daily_2026-09-25.md, agents/meridian/reflections/daily_2026-09-25.md (wicker retired — no per-member file)
• Next-session event risk (2026-09-28, Monday): LOW — no Tier-1 (next: PCE Sep 30, in 5 days; NFP Oct 2; CPI Oct 14). Two things to watch into Monday's open: the Iran/Hormuz seven-day clock runs across the weekend — a weekend headline is a gap-size input for Null and an un-timeable one — and Micron's FQ4 on 09-30 lands on the PCE day, which the stale earnings calendar will not flag.