Closing Briefing — Monday, September 14, 2026

1 · Executive Dashboard

2 · Market Hot Spots (ranked by tradability)

3 · Sector Performance — September 14, 2026

Source: Finviz group screener, Performance table view (g=sector&v=140&o=name), read after the close. 1-Day is the Change % column, 1-Week Perf Week, YTD Perf YTD. Finviz classification, not GICS. Read in the in-app browser pane this session: finviz.com was refused by the Chrome extension's domain policy — see Data Notes.

Three groups green, eight red, and the ranking is a near-perfect inversion of Friday's. Friday industrials led at +1.32% and utilities lagged at -0.32%; Monday industrials fell 1.65% and utilities 1.50%, with technology -2.05% and basic materials -2.08% below them. What took the top is the defensive and non-capital-intensive end: communication services +2.68%, its best session of the window, on Alphabet A +3.22%, Meta +2.71%, Fox Corp B +4.60% and Charter; consumer defensive +1.44% on Kroger +4.14%, Walmart +1.80% and Costco +1.56%; and healthcare +1.36% on Boston Scientific +4.82%, Molina +4.82%, Centene +4.50%, Insulet +4.43%, Eli Lilly +2.02% and UnitedHealth +1.18%. A market that sells the data-centre supply chain and buys grocers, insurers and advertising platforms on the same afternoon is rotating along the capital-intensity axis, not the cyclical one.

The financials number conceals the day's other shock. Financial fell only 0.41%, but Bank of America fell 5.14% on 54.58m shares after its chief executive told an investor conference that third-quarter sales and trading revenue would be "roughly flat" year on year, and Goldman Sachs fell 3.96%, Morgan Stanley 3.64%, Citigroup 1.90%, JPMorgan 1.71% and Wells Fargo 1.76%. The group held because the insurers and the exchanges did not follow the banks — Arthur J Gallagher +5.40%, Verisk +6.15% and FactSet +6.63% are all financial-sector names on the Finviz definition. The capital-markets complex had a much worse day than the sector line reports.

The YTD reconciliation, and healthcare is a genuine outlier this session. Compounding each group's 11 September YTD by Monday's one-day move reproduces the published YTD to within 0.07 percentage points for nine of the eleven groups. Worked examples: utilities 0.9819 × 0.9850 = 0.96717, or -3.28% against a published -3.28%, deviation zero; financial 1.0776 × 0.9959 = 1.07318, or +7.32% against +7.32%, deviation zero; energy 1.4138 × 0.9921 = 1.40263 → +40.26% against +40.28%, deviation 0.02; technology 1.2543 × 0.9795 = 1.22858 → +22.86% against +22.79%, deviation 0.07. Two groups miss. Consumer cyclical implies 0.9379 × 0.9961 = 0.93424, or -6.58%, against a published -6.42% — a 0.16 pp gap. And healthcare implies 1.0602 × 1.0136 = 1.07462, or +7.46%, against a published +6.50% — a 0.96 pp gap, the largest single-group deviation of the reporting window. Healthcare rose 1.36% on the day and its published YTD advanced only 0.48 points. That signature is a constituent change inside the vendor's group definition rather than price drift; the vendor figure is published as it stands, flagged here, and carried forward for confirmation.

The composition traps. Technology at -2.05% understates the semiconductor damage badly — SOX -5.86%, Corning -13.70%, Teradyne -13.30% — because the same bucket holds CrowdStrike +13.85%, Palo Alto +13.09%, Fortinet +9.04%, Autodesk +7.78%, Intuit +5.47%, Adobe +5.30% and Salesforce +4.73%. Software and silicon sit in one Finviz group and went eighteen points apart. Basic materials at -2.08% is the worst group on the board on a session with no obvious commodity catalyst beyond the metals, and energy at -0.79% fell on a day Brent rose 1.79% — the fifth consecutive session the group has refused the barrel. Consumer cyclical at -0.39% contains Best Buy +4.44% and Copart +5.94% against Tesla -1.77%.

4 · Movers & Single-Name Catalysts

S&P 500 constituents unless flagged. Closes and volumes from the Investing.com component board, read after the 16:00 ET close; the capture returned 494 of roughly 500 lines, so counts are a ratio rather than a census.

Up — security, software and the services tier, on the artificial-intelligence safety debate

CrowdStrike (CRWD) +13.85% to $235.38 on 24.66m shares led the S&P 500. Chief executive George Kurtz argued in reply to Dario Amodei's essay that frontier laboratories will keep advancing "regardless of any single company's decision," and that security is what makes the advance tolerable; RBC Capital, Raymond James and Wedbush raised targets after the Falcon user conference. Palo Alto Networks (PANW) +13.09% to $373.94 on 11.41m shares followed on the same logic, and Fortinet (FTNT) +9.04% to $170.18 — a name that fell 1.75% on Friday.

The services and enterprise-software tier joined without a company-specific catalyst, which is the tell that this was allocation rather than news. Gartner (IT) +9.73% to $197.07, Autodesk (ADSK) +7.78%, ServiceNow (NOW) +7.41% to $142.35 on 15.27m shares, Cognizant (CTSH) +6.83%, Tyler Technologies (TYL) +6.79%, FactSet (FDS) +6.63%, GoDaddy (GDDY) +6.53%, Verisk (VRSK) +6.15%, Accenture (ACN) +6.04% to $195.00, Copart (CPRT) +5.94%, Intuit (INTU) +5.47%, Arthur J Gallagher (AJG) +5.40%, Adobe (ADBE) +5.30%, EPAM (EPAM) +4.83%, Salesforce (CRM) +4.73% and PTC +4.49%. Every one of these is a business whose input cost is people rather than transformers and turbines.

Healthcare and staples carried the rest of the advance. Boston Scientific (BSX) +4.82% and Molina Healthcare (MOH) +4.82%, Centene (CNC) +4.50%, Insulet (PODD) +4.43%, Eli Lilly (LLY) +2.02%, Intuitive Surgical (ISRG) +2.38% and UnitedHealth (UNH) +1.18%; Kroger (KR) +4.14% to $60.91 extended Friday's guidance-cut rally by a further four per cent, Darden (DRI) +4.14%, Best Buy (BBY) +4.44%, AutoZone (AZO) +3.09%, Paychex (PAYX) +2.34%, Microsoft (MSFT) +1.97%, Walmart (WMT) +1.80%, General Mills (GIS) +1.73% — ending a nine-session decline — Costco (COST) +1.56% and Alphabet A (GOOGL) +3.22% on 32.54m shares, Meta (META) +2.71% and Palantir (PLTR) +3.64%.

Down — optical, test, equipment and the power tier

Corning (GLW) -13.70% to $143.60 on 16.32m shares was the worst name in the index, and the optical-fibre supplier to the data-centre interconnect is the purest expression of the trade. Teradyne (TER) -13.30% and Hewlett Packard Enterprise (HPE) -10.77% to $55.41 on 27.98m shares, the latter compounding the safety story with an Evercore ISI downgrade after Friday's 12.44% gain — a full round trip in two sessions. Skyworks (SWKS) -10.29%, Super Micro (SMCI) -8.38% on 35.84m shares, Lam Research (LRCX) -8.29%, Qorvo (QRVO) -7.43%, Monolithic Power (MPWR) -7.39%, Applied Materials (AMAT) -7.07%, Keysight (KEYS) -6.99%, KLA (KLAC) -6.39%, Jabil (JBL) -6.09%, ON Semiconductor (ON) -5.90%, Arista (ANET) -5.90%, Dell (DELL) -5.82%, NetApp (NTAP) -5.55%, Micron (MU) -5.25% on 26.53m shares, NXP -5.24%, Intel (INTC) -5.59% on 93.17m shares, Broadcom (AVGO) -4.77%, Analog Devices (ADI) -4.69%, Western Digital (WDC) -4.53%, Synopsys (SNPS) -4.01%, Microchip (MCHP) -3.68%, Cadence (CDNS) -3.55%, Nvidia (NVDA) -3.37% on 109.40m shares, Seagate (STX) -2.97% and Texas Instruments (TXN) -1.99%.

The power, electrical and real-asset tier that funds the same buildout

GE Vernova (GEV) -8.62% to $874.76, Eaton (ETN) -7.57%, Constellation Energy (CEG) -7.09%, Amphenol (APH) -6.40%, Johnson Controls (JCI) -5.65%, Generac (GNRC) -5.35%, Vistra (VST) -5.16%, NRG -4.40%, Quanta Services (PWR) -4.39%, Caterpillar (CAT) -4.22%, Equinix (EQIX) -3.76%, Digital Realty (DLR) -5.01% and Emerson (EMR) -2.56%. Every name on this list closed green on Friday on Oracle's capital-expenditure guide.

The banks, and a guidance warning the sector line hides

Bank of America (BAC) -5.14% to $59.47 on 54.58m shares after its chief executive told an investor conference that third-quarter sales and trading revenue would be roughly flat year on year. Goldman Sachs (GS) -3.96% to $988.45, Morgan Stanley (MS) -3.64%, Citigroup (C) -1.90%, Wells Fargo (WFC) -1.76% and JPMorgan (JPM) -1.71% followed. Oracle (ORCL) -3.65% to $144.79 on 39.12m shares and SLB -4.89% round out the large-cap decliners.

Analyst actions, with the arithmetic

Single-name events worth the desk's attention

BioNTech (BNTX, not S&P 500) reported a lung-cancer candidate doubling median survival to 18.5 months against 10 months for chemotherapy, and GSK (not S&P 500) a lung-cancer therapy with a 94% tumour response rate. Apollo Global is in talks to acquire Johnson & Johnson's orthopaedics unit at a potential $20bn valuation. Element Solutions (ELMT, not S&P 500) won a roughly $2bn Defense Logistics Agency tungsten contract. Tesla (TSLA) -1.77% captured 52% of the U.S. electric-vehicle market through August, up from 43% a year earlier. Rumble (RUM, not S&P 500) rose on a report of a $13.7bn Anthropic computing deal — the same company whose chief executive triggered the day's selloff.

5 · S&P 500 Earnings Calendar — Current & Next Week (S&P 500 components only)

Sourcing, disclosed. The Earnings Whispers day pages remain behind a cookie-and-usage-agreement consent banner, which this unattended session did not accept. The rosters below are captured from the Nasdaq earnings calendar API for each date and screened name by name against the 494-line Investing.com S&P 500 component capture taken the same session. Nasdaq publishes a before-open or after-close bucket rather than a clock time, so no clock times are asserted this session; confirm every time against company investor relations before trading a date.

Current week (Sep 14 - Sep 18) — remaining sessions

Tue 9/15. No S&P 500 reporter on either bucket.

Wed 9/16. AMC: Lennar (LEN).

Thu 9/17. No S&P 500 reporter on either bucket.

Fri 9/18. No S&P 500 reporter on either bucket.

Next week (Sep 21 - Sep 25)

Mon 9/21. No S&P 500 reporter on either bucket.

Tue 9/22. BMO: AutoZone (AZO).

Wed 9/23. BMO: Cintas (CTAS), Paychex (PAYX), General Mills (GIS).

Thu 9/24. BMO: Darden Restaurants (DRI). AMC: Costco Wholesale (COST).

Fri 9/25. No S&P 500 reporter on either bucket.

Changes vs. the prior calendar (9/11 report):

6 · U.S. Treasury Yields — Official Par Curve

Source: U.S. Department of the Treasury daily par yield curve, 14 September 2026 row, read from the Text View with the month-scoped query. All fourteen tenors were extracted; the table publishes ten. Week-on-week is versus the 4 September row. The Wall Street Journal real-time cross-check was unavailable this session — see Data Notes.

A bill-led bear flattener with the long end richening for a second consecutive session, and the location of the peak is the diagnostic. The largest move on the fourteen-tenor strip is off the published table: the 6-month rose 6 bp to 4.18%, with the 3-month +4 bp to 4.11% and the 4-month +3 bp to 4.18%. Inside the coupon curve the peak sits at the 3-year, +4 bp to 4.73%, with the 2-year and the 5-year both +2 and the 7-year and 10-year both +1 — and then the sign flips again: the 20-year fell 1 bp to 5.37% and the 30-year 1 bp to 5.34%. Friday's move peaked at the 1-year and 2-year; Monday's peaks at the 6-month bill and the 3-year. The market is no longer adding to the first hike — it is adding to the second and third, which sit inside the window a 6-month bill and a 3-year note both span. 2s10s tightened 1 bp to 32 bp, a fresh cycle low, and 2s30s 3 bp to 69 bp.

The 10-year's intraday statement is bigger than its close. Investrade and TheStreet both record the yield touching 5.00% during the session, its highest since October 2023, before settling back; the official par close is 4.97%, up 1 bp, and Bloomberg's real-time board reads 4.99%, +2 bp at 16:59 ET. A market that trades a four-year yield high and closes a basis point above the prior day has absorbed the level rather than broken through it — but the fact that it printed at all, on a session when equities fell and the front end added only 2 bp at the 2-year, says the selling pressure at ten years is coming from supply and term premium rather than from the policy path.

The week is where the damage is, and it has migrated outward. The 3-year is 28 bp cheaper than a week ago and the 2-year 28 bp, against the 5-year 26 bp, the 10-year 19 bp, the 20-year 12 bp and the 30-year 10 bp. A week ago the 2-year was the sole peak at 26 bp with the 3-year at 24; now they are tied at 28 and the 5-year has closed to within two. 2s30s has tightened 18 bp in five sessions and 2s10s 9 bp. More than half the week's cheapening still sits inside three years, but the share held by the very front has stopped growing.

The off-table bills carry the whole of the session's information and the auctions corroborate them to the basis point. The 1-month rose 1 bp to 3.94%, the 1.5-month 1 bp to 4.00%, the 2-month 1 bp to 4.06%, the 4-month 3 bp to 4.18% and the 6-month 6 bp to 4.18%, against the 3-month's 4 bp. Three observations follow. The 4-month and the 6-month are now identical at 4.18%, closing a 3 bp inversion that has run all week — the localised cheapening around the December meeting has finished compressing and the strip is flat from four months to six. The 2-month at 4.06% is 5 bp below the 3-month at 4.11%, a gap that widened from 2 bp, so the instrument that does not span the 16 September decision is falling further behind the one that does. And the 11:30 ET auctions stopped at 3.970% on the 3-month against a 3.800% prior and 4.060% on the 6-month against 3.890% — seventeen basis points of concession at each tenor on the day before the committee convenes. A bill complex conceding seventeen basis points at auction while the 2-year moves two is a financing signal, and it belongs with the Section 9 funding data.

The vendor-versus-official gap, explained. Bloomberg's real-time board closed the 10-year at 4.99%, +2 bp, against the official par 4.97%, +1 bp, and its Treasury board ran to 16:59 ET against the par curve's 15:30 ET strike. Two basis points of level and one of change across ninety minutes on a session whose intraday high was 5.00% is a timing artefact rather than a level dispute, and the direction agrees at every point. The Wall Street Journal real-time quotes normally used as the third check could not be retrieved this session because wsj.com was refused by both browsers available to this run; the cross-check is therefore two-vendor rather than three-vendor, and that is disclosed rather than papered over.

7 · U.S. Macroeconomic Calendar

Source: the TradingEconomics United States calendar, read this session with the timezone selector resolving to Eastern. Its date headers are timezone-shifted, so rows are read by release name rather than by day header. The Wall Street Journal market-data calendar cross-check was unavailable this session. The FOMC date is reconciled against the Federal Reserve's own 2026 schedule, CME FedWatch's meeting label and countdown, and the Investing.com card: all four put the two-day meeting on 15-16 September with the decision and projections at 14:00 ET on Wednesday 16 September.

Current week (Sep 14 - Sep 18) — still to come

Next week (Sep 21 - Sep 25)

8 · Fed Funds Futures & Rate Path

Current target range: 3.50%-3.75%. With no macro release and a 5.9-point jump in the September hike, the front end added certainty while the back end gave up terminal rate.

CME FedWatch headline — 16 September 2026 meeting. Data as of 14 Sep 2026, 05:06:30 p.m. CT (6:06 p.m. ET), read from the FedWatch probability table. Contract ZQU6, mid price 96.2613, prior volume 114,611, prior open interest 248,780.

Provenance of every column, stated, and the live-read correction persists at its usual size. The footer timestamp reads 05:06:30 CT with no meridian; the read was taken at approximately 6:40 p.m. ET, well after the 4:00 p.m. CT ZQ session close, so it resolves as p.m. and is an indicative snapshot rather than a settlement one. The 1 DAY column carries the legend date 11 September and prints 87.3%, against the 86.5% this report published from CME's live column on Friday evening — a +0.8 percentage-point correction. The window now reads +0.8, +0.8, 0.0, +1.0, +1.1, +0.8: five of six corrections sit between +0.8 and +1.1, one is zero, and none is negative. That is a consistent small upward bias in the live read, but it remains inside a single rounding band at its median and no standing adjustment is applied. 1 WEEK (4 September) at 59.4% reproduces exactly what Friday's edition carried for the same reference date, which is a clean confirmation of that column's stability. 1 MONTH carries the reference date 14 August 2026, a genuine one-month-back stamp rather than the 11 August legend Friday's table showed, so the 33.1% is used below rather than marked chart-read.

The CME-versus-Investing.com gap, quantified. CME puts the September hike at 92.4% at 6:06 p.m. ET; Investing.com at 91.8% at 5:55 p.m. ET — a 0.6 percentage-point difference across eleven minutes, against 1.0 on Friday, 0.3 on Thursday and 0.1 on Wednesday. Investing.com publishes the September future at 96.260 against 96.268 on Friday, and CME's own mid at 96.2613, so the two vendors are within 0.13 basis points of contract price and the whole of the probability gap is CME's day-weighting of a mid-month meeting. Because the 16 September meeting falls mid-month, only about 47% of the contract's averaging period is affected, so one basis point of ZQ price maps to roughly ten percentage points of hike probability — and the 0.8 basis points the front contract cheapened maps to about eight points against the 5.1 points the vendor's own columns show. The residual is the same day-weighting effect, and at 0.6 points the CME-Investing gap is the narrowest since Wednesday.

One-day, one-week and multi-day momentum. The September hike rose 5.1 points on CME's own columns, 92.4% from 87.3%, and 6.3 points on Investing.com's own columns, 91.8% from 85.5%. ZQU6 cheapened 0.8 bp to 96.260 and ZQZ6 1.0 bp to 95.900 — a ratio of roughly one to one, against one to three on Friday and one to seven on Thursday, so the front and the December leg are now repricing together rather than the back doing the work. The multi-day read: the meeting sat at 59.4% a week ago and 33.1% a month ago, so it is 33.0 points more hawkish than a week ago and 59.3 points more hawkish than a month ago, and the five-observation path runs 59.4 to 69.0 to 72.4 to 87.3 to 92.4 on the vendors' reconciled columns. Further out, October's cumulative-above rose to 95.8% from 92.8% and December's to 98.5% from 97.4%. The probability of a cut at any 2026 meeting remains 0.0%.

(a) Current-year meeting distributions

Investing.com Fed Rate Monitor, updated 14 Sep 2026 05:55 p.m. EDT. Format: current [prior day] [prior week]. Modal range in bold.

All three rows sum to exactly 100.0%. Four observations. First, the vendor's prior-day column reproduced Friday's published figures exactly — 14.5% and 85.5% at September, 7.2%/49.6%/43.2% at October, 2.6%/22.7%/47.3%/27.4% at December — after drifting 2.0 points on Friday and 0.3 on Thursday. The fixed-snapshot caveat is real but it did not bite this session, and that is worth recording as plainly as the failures. Second, October did not flip and instead widened its lead: one hike leads two by 7.2 points, 51.5% against 44.3%, where the lead was 6.4 points on Friday. Four sessions of compression stopped in one. Third, the December distribution keeps shifting mass upward inside an unchanged mode — two hikes to 48.9% from 47.3% and three hikes to 28.3% from 27.4% — while the hold bucket fell to 1.5% from 2.6%. Fourth, the hold bucket is now negligible at every meeting: September 8.2%, October 4.3%, December 1.5%. The current range has effectively been priced out of existence for 2026.

(b) Next-year meeting path

Modal range, its probability, and the cumulative probability above and below the current 3.50%-3.75% range, with the contract price that draws it.

The 2027 strip richened, and that is the single most important change on this page. Friday every contract from June onward moved up a bucket; Monday June moved back down to 4.25%-4.50%, and six of the eight contracts are 1.0 to 2.0 basis points richer than Friday's prints while January is a basis point cheaper and March is unchanged. Set that against the front, where September 2026 cheapened 0.8 bp and December 2026 1.0 bp. The strip is doing two opposite things at once: adding near-term hike certainty and removing terminal rate. That is precisely the shape an insurance hike produces — the committee moves, and the market takes the destination down because the move itself is expected to do the work. December 2027's modal 4.50%-4.75% now leads 4.25%-4.50% by 0.1 points, 28.2% against 28.1%, so the terminal bucket is one session from flipping down. Cumulative-above at December 2027 is 99.5% against 99.4% on Friday and the cut probability at 3.25%-3.50% remains 0.0%.

(c) Year-end probability ladders

Year-end 2026 — the 9 December meeting.

Cumulative above the current range: 98.5%. Cumulative below: 0.0%. Sum: 100.0%.

Year-end 2027 — the 8 December meeting.

Cumulative above the current range: 99.5%. Cumulative below: 0.0%. Sum: 100.0%.

Rounding, transparently. Every figure is reproduced at the vendor's own one-decimal precision. Column sums of 99.9% or 100.1% are rounding artefacts of that precision, not missing probability mass; no cell has been rescaled, and cells shown as 0.0% are ranges the vendor's own card either omits entirely or publishes as zero, which under CME methodology means a probability below the rounding floor. Both year-end ladders and all three 2026 rows sum to exactly 100.0% this session; the January, June, July, September and October 2027 rows sum to 100.1%, 99.9%, 99.9%, 99.9% and 99.9% for exactly this reason.

9 · Credit & Funding

(a) IG and HY credit spreads

ICE BofA option-adjusted spreads via FRED. FRED publishes with a one-business-day lag: the levels below carry the 11 September 2026 as-of date, not the 14 September close. Same-day direction is cross-checked against the cash-market proxies underneath and against Bloomberg credit coverage. 1-Week is versus the 4 September row.

CDX — the six-step ladder was worked again and only three of the six steps could be executed this session, one fewer than Friday, because a source that has rendered every prior run stopped being reachable. (1) Bloomberg in Chrome: /markets, /markets/rates-bonds and /markets/fixed-income all rendered fully, and a full-text scan of each returns zero occurrences of the index name and zero of "default swap"; the fixed-income tables carry Bloomberg's own aggregate indices and the regional ten-year government boards. (2) WSJ Market Data bonds page: refused by the Chrome extension's domain policy AND by the in-app browser pane's safety restrictions, and blocked at the fetch layer as well — three routes, three refusals. WSJ rendered normally in Chrome on every prior run in this window, so this is a new tooling failure and it removes a step that was previously productive. (3) Cbonds CDX.NA.IG 5Y (index 204395): reached in the in-app pane after Chrome refused the domain, and it carries a live dated record — a current value stamped 11/09/2026 and a previous value stamped 10/09/2026, both one business day later than Friday's capture, with the basis-point figure masked behind the request-access wall. The record is advancing; the number is still not publishable. (4) FT Markets Data and (5) Barchart remain refused by the Chrome extension's domain policy, so those two steps could not be executed at all. CME's credit-index product pages carry specifications rather than levels. (6) Cash-market proxies, labelled as proxies: HYG closed $78.53, -0.09%, and LQD $104.30, -0.02%. No CDX level is published here, and three of the six failures are tooling rather than data.

The tail went through the level and the index went the other way. On the 11 September stamp the CCC credit spread widened 6 bp to 1,076 while HY tightened 5 bp to 265 and IG was unchanged at 80, putting the CCC-minus-HY differential at 811 bp against exactly 800 on the prior update — 11 bp of widening in a single stamp, the largest of the nine-update window, and the eighth widening in nine. It is now 22 bp wider on the week and 188 bp on the year, against an HY series 18 bp tighter than it started 2026. Read the composition rather than the level: the index spread tightened five basis points on the same day the tail widened six, so the eleven-point move in the differential is two-thirds index strength and one-third tail weakness. That is a cleaner decoupling than Friday's, because on Friday both the tail and the index moved in the same direction by a basis point each.

The cash proxies finally broke, and one of them broke a level this report named. HYG closed $78.53, down 0.09%, below the $78.57 one-year low that Section 13 of the 11 September edition flagged as seven cents away — so the high-yield cash proxy is at a fresh one-year low with the index spread inside 270 bp, which is the contradiction investment grade has been living in since August arriving in high yield. LQD closed $104.30, down 0.02%, also a fresh low, with its floor now three sessions in a row of new lows at the close rather than intraday. The base for both series was restated a few cents by the vendor over the weekend and that is reconciled in Data Notes.

(b) Money-market & funding plumbing

New York Fed reference rates, published at approximately 8:00 a.m. ET for the prior business day. The 10 September 2026 row is the latest published to the reference-rates endpoint at capture; no 11 September row had published. The operations figures beneath it are dated 14 September. Rate up = red.

The plumbing reversed direction and the concession moved into the auctions instead. Reverse repo take-up collapsed to $1.420bn on 14 September from $5.255bn on 11 September, a 73% one-session decline that gives back two thirds of the twelve-fold build this report tracked across the prior three sessions. Cash that was returning to the facility at term rates it declined to lend at has come back out — on the same afternoon the bill auctions demanded seventeen basis points more than their prior stops. Those two facts are the same fact: money-fund cash left the facility because the bill market finally paid it enough. SOFR at 3.62% is 3 bp below the 3.65% IORB, the third consecutive sub-administered print and the widest of the three, on $2,921bn of volume with the 99th percentile at 3.70% — the tails are not moving, so this is abundance rather than dispersion. Reserves remain $2.9913tn for the week ended 9 September, unchanged because no new print has published, with quarter-end now ten days away. The standing repo facility result could not be retrieved: the date-scoped endpoint that works for reverse repo returns a 400 error on the repo query shape, exactly as recorded on 10 September, and no figure is asserted.

The off-table bill tenors belong here and they closed the week's last inversion. The 1-month rose 1 bp to 3.94%, the 1.5-month 1 bp to 4.00%, the 2-month 1 bp to 4.06%, the 4-month 3 bp to 4.18% and the 6-month 6 bp to 4.18%, against the 3-month's 4 bp to 4.11%. The 4-month and the 6-month are now identical at 4.18%, closing the 3 bp gap that has run all week, so the localised cheapening around the December meeting has finished expressing itself and the strip is flat from four months out to six. Against that, the 2-month at 4.06% has fallen 5 bp below the 3-month from 2 bp, widening the discount on the instrument that does not span the decision. The shape now says the bill market prices a hike inside the quarter, a second one by the turn of the year, and nothing incremental between four and six months.

(c) Rates volatility & swap spreads

The two surfaces converged for the first time in a week, and the MOVE card failed a different self-check from Friday's. The Investing.com MOVE series now carries an 11/09 date stamp at 82.21 with a day range of 82.09 to 82.21 — the level sits at the top of its own range and the lower bound equals the prior published vintage, which are two internal checks that pass. The card's change field reads 0.00%, which cannot be true when the level has moved from a published 82.09, so the computed change of +0.12 points, or +0.15%, is substituted and the vendor field is named as failing. The "previous close" field still reads 95.74, outside its own day range for a third consecutive session; that field is corrupt and is not used. So the 11 September MOVE rose fractionally rather than extending Thursday's 6.97% jump — rate volatility has plateaued at its high. Against that, VIX rose 7.95% to 17.10 with a high of 18.17, retracing 63% of the 2.00 points it shed on Friday. The MOVE-to-VIX ratio at 4.81, computed on a one-day-stale numerator and flagged, is down from 5.18 on Friday and 4.30 on Thursday. The week's story has been rate volatility making highs while equity volatility collapsed; on Monday the equity surface came back to meet it, which narrows the divergence rather than resolving it — and the meeting is still two days out. Swap spreads at the 2-year, 10-year and 30-year were not obtainable from a primary source this session and are not asserted; the substitute evidence is the cash curve, where 20s30s held at -3 bp for a second session.

(d) Issuance, leveraged loans & private credit

The financing market spent Monday on the equity side of the capital structure and on the borrower most exposed to the day's story. Larry Ellison cancelled a planned $7.5bn sale of Oracle stock, which is a controlling shareholder declining to monetise into a market that has taken the equity down 7.3% in two sessions while the company carries $90-95bn of guided capital expenditure, negative free cash flow expected through 2029 and a rating S&P Global cut to the lowest investment grade in July. Sysco is seeking to raise $1bn in a share sale for its Jetro acquisition and the Kentucky Derby operator has tapped the loan market to refinance, while Canadian insurer Sagen plans to raise around C$300m — small, and all of it equity or loan rather than dollar investment-grade primary. That remains the pattern: post-Labor-Day investment-grade issuance is still running at its weakest pace since 2020 after an August near a record $130-145bn and year-to-date supply above $1.68tn, up 27% on 2025, with the 10-year having touched 5.00% intraday. Apollo used $101m of a tariff refund to cut debt at Michaels, a reminder that the sponsor complex is deleveraging where it can rather than refinancing where it must. On the loan side no updated Morningstar LSTA print was obtainable this session and none is asserted. The named private-credit watch item is unchanged — Broadcom's contingent residual-value guarantees to two artificial-intelligence laboratories, vendor financing no published spread series captures, with Broadcom -4.77% — and the day added a second: Rumble rose on a report of a $13.7bn Anthropic computing deal, signed by the counterparty whose chief executive spent the weekend arguing the industry should slow down. That contract is not on any published credit curve either.

10 · FX

Source: TradingEconomics currency board, read after the U.S. close on the vendor's Sep/14 stamp. Quote basis: EUR, GBP, AUD and NZD are quoted as units of dollar per unit of foreign currency, so a fall is a weaker foreign currency; every other pair is quoted as units of foreign currency per dollar, so a rise is a weaker foreign currency. The %Chg column is the vendor's own daily change on its Sep/14 boundary and covers the full U.S. session.

The take: the dollar finally moved, and it moved a third of what the rate story would justify. DXY rose 0.35% to 99.468, which is the first change of any size in three sessions, on a day the September hike went to 92.4% and the 10-year touched 5.00%. Two things follow. First, the level dispute from Friday is resolved in the vendor's favour: Investing.com's independently derived DXY closed at 99.47 against TradingEconomics' 99.468 — agreement to two decimal places on a pair of series that were 0.39 apart on Friday's capture, which retrospectively confirms that Friday's "unchanged to three decimals" print was a stale field rather than a stationary currency. Second, 35 basis points of dollar strength against a 33-point one-week increase in the September hike probability is a very poor exchange rate, and the composition says why: ten of the thirteen crosses moved in the dollar's favour but the largest of them, NZD at -0.58% and USD/KRW at +0.43%, are both risk-proxy currencies rather than rate-differential ones. The dollar rallied because equities fell, not because the front end cheapened.

The haven cross failed for a third consecutive session and this time the test was the cleanest available. USD/CHF rose only 0.08% — the franc was essentially unchanged — on a session with a 5.86% semiconductor decline, a 7.95% rise in VIX and a global growth narrative under public attack from the people building it. Thursday the franc would not bid on an 8% crude rally; Friday it would not bid on an inflation surprise; Monday it would not bid on an equity-risk event. Three different risk signals, three refusals. A haven competing against a 4.11% three-month bill that just conceded seventeen basis points at auction is being outbid by cash, and that is the structural read on this table. The mirror image is USD/NOK +0.40%: the krone weakened on a session when Brent rose 1.79%, having weakened on a session when Brent fell 2.94% and again when it rose 7.98%. Three directions, one outcome — the Norwegian currency has stopped trading terms of trade entirely. USD/CAD +0.19% repeated the pattern more mildly on the same crude tape.

Asia's divergence closed and the closing is the finding. USD/KRW rose 0.43% to 1,347.32 — the won was the weakest Asian major — on a session when the Kospi fell 3.26%, its worst of the window. On Friday the won strengthened 0.65% while the Kospi fell 1.76%, and this report read that as a domestic rate story rather than foreign flow. Monday inverted both legs: the equity decline doubled, the currency weakened, and the Korean ten-year richened a basis point to 4.53% rather than cheapening. The rate leg and the equity leg now agree, which means Friday's won strength was positioning being unwound rather than a structural bid — and it is the semiconductor concentration, not the curve, that sets the won. Against that, USD/CNY moved 0.00002 in a full session and is 0.02% lower on the week, so two weeks of the largest rate repricing of the year have left the managed currency exactly where it started. The yen gave back the entire week in one day: USD/JPY rose 0.47% to 154.339, leaving the week at -0.01%, with ten-year JGBs a basis point cheaper at 2.97% and the Bank of Japan four days out. A currency that erases a 1.69% weekly gain on a single risk-off session was carrying a crowded long, and the Bank of Japan meeting now lands on a flat position rather than a profitable one.

11 · Commodities

Settlement basis, stated, and reconciled to the prior edition. The Investing.com per-contract historical board remains the settle series of record, unchanged from the prior four editions; the basis has not been altered. Rows were captured at approximately 18:30 ET. Changes are computed against the 11 September finalised rows, not against the Sunday 13 September electronic rows the board interleaves — the vendor's own %Chg column measures against the Sunday row and is therefore not reproduced. All eight of the prior edition's rows have finalised and are restated below with every derived figure recomputed. Five rows are forming and each is corroborated against TradingEconomics spot on the same capture and against dated third-party settlement snapshots, quantified below. WTI, RBOB, heating oil and natural gas are on the October contract; gold, silver and copper on December; Brent on the November contract. No front-month roll occurred this session. Weekly and year-to-date columns are TradingEconomics spot returns on the vendor's Sep/14 stamp.

*\YTD figures marked with an asterisk are TradingEconomics spot year-to-date returns, not futures returns on the contracts quoted above. Weekly columns are TradingEconomics one-week changes on the same caveat. Mixing the two would be a basis error; they are presented in separate columns for exactly that reason.

The 11 September restatement, and it is the smallest of the window on energy and the largest on the metals. Published against finalised: WTI $99.99 against $100.05, Brent $104.47 against $104.61, heating oil $4.9936 against $4.9593, RBOB $3.3192 against $3.3072, natural gas $2.820 against $2.831, gold $4,390.00 against $4,408.90, silver $65.020 against $65.188, copper $6.5570 against $6.5480. WTI's gap of six cents is an order of magnitude better than the $1.24 of 8 September and the $1.60 of 10 September, and the corroboration procedure is why: Friday's rows were published at volumes between 42% and 99% of the prior session rather than at the 0.1% levels that produced the big errors. The exception is gold, $18.90 higher on the restatement, and it changes a published number's sign — Friday's gold move restates from -0.39% to +0.04%, so the metal did not fall on the hot core inflation print, it was unchanged. Silver's Friday move restates from +0.14% to +0.40%. The restated 11 September cracks are distillate $108.24 and gasoline $38.85 for a differential of $69.39, against the $109.74, $39.42 and $70.32 published — the published differential was 93 cents too wide. Restated Brent-WTI is $4.56 and the restated gold-silver ratio 67.63.

The forming-row check, quantified, and this is the thinnest capture of the window. Volumes this session against the prior session: natural gas 184.59K against 128.32K (144%), Brent 544.15K against 459.78K (118%), RBOB 18.60K against 32.52K (57%) — and then a cliff: gold 0.12K against 187.98K (0.06%), WTI 0.11K against 399.14K (0.03%), copper 0.06K against 42.94K (0.14%), silver 0.05K against 53.69K (0.09%) and heating oil 0.01K against 58.86K (0.02%). Five rows are three orders of magnitude below their predecessors, which is the clearest forming-row signal this report has recorded. Natural gas and Brent are completed settles needing no qualification. Corroboration on the same capture against TradingEconomics spot: crude 102.014 against the board's 101.98 — four hundredths apart on a $102 barrel — natural gas 2.8804 against 2.882, gasoline 3.3548 against 3.3566 and heating oil 4.9789 against 4.9830 all agree to the third decimal, and Brent 106.277 against 106.48 differs by twenty cents. The metals agree in direction with the expected futures-over-spot basis: gold spot 4,296.67 against Comex December 4,337.76 (a $41.09 or 0.96% carry), silver spot 63.211 against 63.720 (0.80%) and copper spot 6.3250 against 6.4008 (1.20%) — all three inside the 0.84%-to-1.35% band the prior four editions recorded.

The dated third-party snapshots disagree with the thin rows, and they are named rather than withheld. Bloomberg's board and the Investrade review both record WTI at $101.39, +$1.34, and both record Comex December gold at $4,351.90, -$57.00; Investrade additionally records Brent at $105.68, +$1.07 and silver at $64.14, -$1.05. Every one of those figures is computed off the same finalised 11 September base this report uses, so the vendors agree on the starting point and disagree on the settlement snapshot by $0.59 on WTI, $0.80 on Brent, $14.14 on gold and $0.42 on silver. The rule set on 11 September applies: where a dated third-party review disagrees with a thin historical row, the third party is the better estimate of where the row will finalise, and on that reading gold's true settle is nearer $4,351.90, -1.29% than the -1.61% tabled, and WTI's nearer $101.39, +1.34%. The table holds the historical-board basis because changing it mid-window would make five editions non-comparable; the direction and the magnitude of the expected revision are stated here instead, and Tuesday's edition will restate against whatever prints.

The crack spreads on a consistent October basis against $101.98 WTI:

The crack trade failed its first genuine test, and the invalidation clause missed by seven hundredths of a percentage point. The written invalidation is "a crude session above 2% in either direction in which heating oil underperforms gasoline." Monday delivered a crude session of +1.93% — seven hundredths short of the threshold — and heating oil did underperform, badly: +0.48% against gasoline's +1.49%, a 1.01-point margin the wrong way. The clause did not fire on its literal terms and the mechanism failed anyway. Two directional tests inside forty-eight hours last week produced two wins by margins of 0.39 and 2.74 points; the third produced a loss, on a smaller crude move, which is the configuration the thesis is least able to survive — physical distillate tightness is supposed to be indifferent to the size of the crude move. Section 12 marks it accordingly and does not hide behind the technicality.

Crude rose on transit risk and the barrel did not carry the products with it. Brent settled $106.48, up 1.79% on a completed settle, and WTI $101.98, up 1.93%, against a tape carrying Saudi efforts to route more crude around the Strait of Hormuz and a Bloomberg report that oil tankers are now earning $1m a day as the conflict leaves a ship shortage. Crude is +9.66% on the week and Brent +9.38% on the spot basis. The Brent-WTI differential narrowed to $4.50 from a restated $4.56 — six cents of narrowing on a 1.8% up day, against 67 cents of narrowing on a 2.4% down day and five cents of widening on an 8% up day. Three sessions, three different crude directions, and the seaborne premium has moved a total of 73 cents: the transit story is in the flat price, not in the spread, which continues to say less of the level is Hormuz than the headlines imply. Energy equities fell 0.79% on the same tape — a fifth consecutive session of refusing the barrel — with SLB -4.89%, Halliburton -2.32%, Valero -1.91% and Phillips 66 -0.93% all lower.

The metals broke and the break has the wrong sign for a haven. Gold fell 1.61% to $4,337.76 and silver 2.25% to $63.720, with spot gold -1.23% and spot silver -1.64%, on a session when SOX fell 5.86%, VIX rose 7.95% and the dollar rose only 0.35%. A metal that sells off on an equity-risk event, with the 30-year richening a basis point underneath it, is not being driven by real rates or by the currency. The gold-silver ratio widened to 68.07 from a restated 67.63, and gold's spot year-to-date return is back to -0.53% after Friday's restatement had put it at +0.71%. Copper fell 2.25% to $6.4008 and is -6.16% on the week, the worst weekly number in the complex, with platinum -1.51%. Set against natural gas +1.80% to $2.882 on a completed settle — the best percentage move on the board and its third consecutive gain — the complex has separated cleanly into an energy bid and a metals liquidation. The uncomfortable read on gold is that it traded with the artificial-intelligence complex rather than against it, which would mean the debasement trade and the buildout trade have been held by the same people.

12 · Trading Views

Desk-style ideas for institutional investors. Each carries an explicit expression, catalyst and invalidation. These are not personalized investment advice; verify independently and size to your own mandate before acting.

1. The rates trade — long ZQZ6 against short ZQZ7 lost money on exactly the move it was built to catch; hold the quarter

Mark first. Long ZQZ6 (December 2026) against short ZQZ7 (December 2027), DV01-matched one-for-one at $41.67 per basis point per contract, entered on 11 September at 95.910 / 95.450 for a spread of 46.0 bp, quarter size. Monday's mark: ZQZ6 95.900, ZQZ7 95.470 — a spread of 43.0 bp. That is -3.0 bp, worth -$125.01 per contract pair before costs on a quarter, in one session.

The honest reading, and it is uncomfortable. The trade gains as the market adds tightening to 2027 faster than to the rest of 2026. Monday did the opposite with unusual purity: ZQZ6 cheapened 1.0 bp while ZQZ7 richened 2.0 bp, and six of the eight 2027 contracts richened 1.0 to 2.0 bp while June 2027 dropped back a full bucket to 4.25%-4.50%. The market added 5.1 points of September hike probability and simultaneously took terminal rate out. That is the insurance-hike interpretation, and it is precisely tail one of the two this position wrote down on Friday — eToro's Bret Kenwell's "dovish hike" framing — arriving two days before the statement that would confirm it. So why hold? Because none of the three written invalidations fired and the catalyst that justified the trade has not happened yet. The spread is 43.0 bp against a 36.0 bp trigger, December 2026's hold probability is 1.5% against a 10% trigger, and the 2027 modal range sits at 4.25%-4.50% at three of eight meetings against a five-meeting trigger. Retail sales at 08:30 on 16 September and the projections at 14:00 are the events; closing a position 44 hours before its own catalyst because the pre-positioning went against it is the error this report criticised on Friday in the opposite direction. Catalyst: the 20-year auction 9/15 at 13:00 against a 5.204% prior; retail sales 9/16 at 08:30, consensus +0.9% against -0.6%; the decision, projections and press conference 9/16 at 14:00. Invalidation, unchanged: the spread through 36.0 bp; or December 2026's hold probability back above 10%; or the 2027 modal range at 4.25%-4.50% at five or more of the eight meetings. Sizing: a quarter, at $41.67 per basis point per pair. Mark to date: -3.0 bp.

2. Protection on the CCC cohort funded in IG — the watch level cleared; hold the quarter, do not add

Mark. CCC 1,076 bp, +6 bp; HY 265 bp, -5 bp; IG 80 bp, unchanged on the 11 September FRED update, taking the CCC-minus-HY differential to 811 bp from exactly 800 — an 11 bp gain, the largest single-update gain of the position's life, for a cumulative +45 bp across nine updates, paid on eight of them. The honest reading. Friday's composition was a tail widening 6 while the index tightened 1. Monday's is a tail widening 6 while the index tightened 5 — two-thirds of the move now comes from strength above rather than weakness below, which is a better decoupling signal, not a worse one, because it cannot be explained by a common risk factor. The 825 bp threshold this report named as the line between two prints and a trend is 14 bp away. Action: hold the quarter; do not add. The position has now made money on eight of nine updates and the temptation to size up at the moment a thesis is confirmed is exactly what the Friday discipline note was written against. Catalyst: the 16 September decision; the first large investment-grade deal to clear against a 4.97% ten-year; quarter-end funding from the 22nd. Invalidation, unchanged: the differential back through 750 bp, or IG widening beyond 90 bp. Sizing: a quarter, duration-hedged.

3. Long the October distillate crack against the gasoline crack — the mechanism failed its test; close it at the gain

Mark. Entered on 9 September at a $66.79 differential on twice-finalised historical-board settles. Monday: distillate crack $107.31, gasoline crack $39.00, differential $68.31 — a $1.52 gain from entry, and $1.08 worse than Friday's restated $69.39. Why this is a close rather than a hold. The written invalidation is "a crude session above 2% in either direction in which heating oil underperforms gasoline." Crude rose 1.93% — seven hundredths of a point short — and heating oil underperformed gasoline by 1.01 points, +0.48% against +1.49%. The clause did not fire on its literal terms. But the thesis was physical distillate tightness, and physical tightness is supposed to be indifferent to whether crude moves 1.9% or 2.4%; a mechanism that wins by 0.39 points on an 8% rally and 2.74 points on a 2.4% decline and then loses by 1.01 on a 1.9% rally has stopped being a mechanism and started being noise. Closing on the spirit of a clause that missed by seven hundredths is the same discipline as closing on the letter of one that hit. Action: closed at +$1.52 per barrel-for-barrel pair. No replacement crack expression is proposed into an EIA report on 9/16 and a decision the same afternoon.

4. Short the credit-bureau complex against long the S&P 500 financials — the five-session clock expired; closed at a loss

Mark. Fair Isaac +1.63% to $1,001.50 and Equifax +1.95% to $172.29; TransUnion did not appear in the 494-line component capture for a third consecutive session and no independently sourced close is asserted, so the basket is marked on two of three names at +1.79%, against S&P 500 financials at -0.41% — a 2.20-point loss on the pair in one session, taking the cumulative from a 0.48-point gain to a 1.72-point loss across five sessions. The honest reading. Friday's instruction was explicit: "hold the quarter through Monday's final session of the window and close the position on the deadline regardless of the mark — the thesis was a five-session reaction-function trade and it expires on its own terms." The deadline is today and the mark is bad. The recovery invalidation never fired — the basket has retraced 6.45% of the 4 September decline against an 8.34% trigger — so the trade was not stopped out; it simply did not work, and the final session took back everything the first four had earned. The specific failure is that financials fell 0.41% on a day the banks inside it fell 1.7% to 5.1%, so the long leg was carried by insurers and data vendors while the short leg rallied with every other business whose costs are people. Action: closed at a 1.72-point loss on the pair. Recorded as a loss, not as an expiry.

5. New — long the power and electrical tier against short the artificial-intelligence security complex, quarter size

Expression: long an equal-weight basket of GE Vernova, Eaton, Constellation Energy, Vistra and Quanta Services against short an equal-weight basket of CrowdStrike, Palo Alto Networks and Fortinet, dollar-neutral, quarter size, entered at Monday's closes. The long basket fell an average of 6.57% on the session; the short basket rose an average of 11.99% — an 18.6-point single-session dispersion between two baskets whose cash flows are barely related. The thesis. Monday's move priced an essay, not an order book. Nothing in Amodei's argument or Altman's warning changes a signed interconnect agreement, a turbine delivery slot or a power-purchase contract, and the one hard datum on the same tape ran the other way: Rumble rose on a report of a $13.7bn Anthropic computing deal — the laboratory calling for a slowdown signing for compute at scale on the day its chief executive's essay took the suppliers down 6% to 14%. Meanwhile the short leg has just been re-rated 12% on a narrative with no revenue attached to it yet. Base case: the dispersion mean-reverts by a third to a half over two to four weeks as the capital-expenditure evidence arrives and the security re-rating is digested. Tail one: a named hyperscaler actually defers a data-centre programme, in which case the long leg has much further to fall and this trade is wrong for the right reason — watch for it in the 9/22-9/24 reporting block and in any capital-expenditure guidance revision. Tail two: a major security incident that validates the short leg's re-rating on fundamentals rather than narrative. Catalyst: the 16 September decision and its effect on long-duration infrastructure equity; any capital-expenditure confirmation from the hyperscalers; Lennar on 9/16 as the housing-side read on the same rate backdrop. Invalidation: the spread widening a further 8 points from entry against the position; or any credible report of a deferred or cancelled data-centre programme at a named operator. Sizing: a quarter, dollar-neutral, deliberately small because the catalyst is the absence of news rather than the arrival of it.

6. New — long October volatility on the semiconductor complex, quarter size, expressed in premium

Expression: October volatility on the semiconductor proxy, a quarter, expressed in premium rather than delta, entered at Monday's close with VIX at 17.10 and SOX at 11,131.3. Mark the predecessor honestly first, because it is the same trade. This report closed exactly this position on Friday at a loss in premium, as instructed, on the session VIX fell 11.21% to 15.84 and SOX rose 1.81%. Monday VIX rose 7.95% to 17.10 and SOX fell 5.86% — the largest semiconductor decline of the reporting window, one session after the close. Holding through Monday would have recovered the entire round trip and more. The decision to close on Friday cost real money and it is recorded as an error of process, not of luck: the position was closed because a scheduled binary had passed, but the binary that mattered was never the inflation print — it was the meeting, which is still two days away. The thesis for re-entry. The surface has retraced only 63% of Friday's collapse, MOVE has plateaued at 82.21 on its 11 September vintage while VIX rose 7.95%, and the decision on 16 September at 14:00 is 92.4% priced as to the move and entirely unpriced as to the statement. Realised single-name dispersion ran 27.6 points on Monday, from CrowdStrike +13.85% to Corning -13.70%, on an index that moved 0.48%. Catalyst: retail sales 9/16 at 08:30; the projections and press conference 9/16 at 14:00, which is the instrument that confirms or destroys the insurance-hike reading the ZQ strip bought on Monday. Invalidation: VIX through 15.50; or a semiconductor session in which SOX outperforms the Nasdaq 100 by more than 2 points on an up day. Mechanical stop: mark it out on 18 September regardless of outcome — and this time the stop sits after the event, not before it. Sizing: a quarter, in premium.

Prior closes, marked forward. The long ZQU6 against short ZQZ6 spread, closed on Friday at +3.5 bp when its own probability trigger fired, would have gained a further 0.2 bp: ZQU6 96.260 against ZQZ6 95.900 is a spread of 36.0 bp against the 35.8 bp it was closed at. The decision cost almost nothing. The long-volatility semiconductor position, closed on Friday at a loss in premium, would have recovered the entire loss and more — VIX +7.95% and SOX -5.86% in one session — and it is re-entered above as idea 6 at a 17.10 VIX against the 15.84 it was closed at, which is recorded plainly as buying back higher. The long 20-year against short 30-year, closed on Thursday at -2 bp, was flat: 20s30s held at -3 bp. The short-debasement basket against long dollar, closed on 3 September, would have gained: gold -1.61% against a dollar +0.35%.

The vol note. VIX closed 17.10, up 1.26 points or 7.95%, with a session range of 16.58 to 18.17, on a day the index fell 0.48% — a 16.6-to-1 ratio of volatility gain to index loss, against Friday's 13.0-to-1 in the other direction. The five-observation path is 16.46 → 17.84 → 15.84 → 17.10, so 63% of Friday's 2.00-point collapse has been retraced in one session. A 17.10 handle asks for roughly a 1.07% daily move against realised index moves of 0.58%, 0.86% and 0.48% — index volatility is rich to realised at about two-to-one, unchanged from Friday, so the premium has not re-expanded even though the level has. That is the case for owning premium rather than selling it: the surface has re-priced the level without re-pricing the relationship. What is genuinely extraordinary is the dispersion underneath. A 494-name distribution running from CrowdStrike +13.85% to Corning -13.70% is a 27.6-point spread on a 0.48% index day, against 18.1 points on Friday's 0.86% day — half again as wide, on a smaller index move, with 282 advancers against 211 decliners. Own dispersion and single-name convexity into the meeting; do not sell the index level here, and do not buy it either.

13 · Risk Map

The crowded consensuses worth stress-testing, with the numbers that would break them.

What VIX is and is not pricing. At 17.10, up 7.95% and back above the 14-to-17 band it re-entered only one session earlier, the index option surface is asking for roughly a 1.07% daily move against three consecutive realised moves of 0.58%, 0.86% and 0.48% — rich to realised at about two-to-one, which is exactly where it stood on Friday at a 15.84 handle. The level has re-priced and the relationship has not. It has stopped mis-pricing the meeting, which it did on Friday by giving back 2.00 points 44 hours before a decision. It is not pricing the statement and the projections, which are the live variable and the instrument that adjudicates Monday's 2027 richening. It is not pricing the rate market's own view: MOVE plateaued at 82.21 on its 11 September vintage while VIX rose 7.95%, so the two surfaces converged rather than resolved. And above all it is not pricing the dispersion, which has gone from wide to extraordinary: a 494-name distribution running 27.6 points from CrowdStrike +13.85% to Corning -13.70%, with 282 advancers against 211 decliners, on an index that moved 0.48%. A market where more than half the names rise, the index falls, and the range between the best and the worst is fifty-seven times the index move is not a market with a level problem. It is a market with a composition problem, and the composition is the risk.