Closing Briefing — Thursday, September 10, 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.

Two groups green and the worst of them is the one that should have led. Energy fell 0.39% on a session when Brent rose 7.98% and WTI 8.36% — the third consecutive day the equity has refused the barrel, and by far the widest refusal yet. Wednesday bought 0.95% of sector for a 4.43% crude move; Thursday bought minus 39 basis points for an 8.36% one. Either the equity market has decided this level is a war premium with a terminal date, or it is netting the demand destruction that OPEC's forecast cut implied on the same morning. Exxon +0.61% and ConocoPhillips +0.37% were the megacap holds, against Baker Hughes -6.66%, Halliburton -2.85% and Williams -3.10% — services and midstream, not integrateds, took the loss, which is a capital-spending read rather than a price read.

Basic materials at -2.88% is the metals complex and nothing else. With silver -6.68%, copper -5.27% and platinum -7.03%, the group's worst names are exactly the ones you would draw: Freeport-McMoRan -6.59% on 20.48m shares, Mosaic -3.09%, Celanese the outlier at +2.86%. Newmont fell only 2.00% against gold's 2.21%, so the gold miners held their beta while the industrial-metal names did not.

The YTD reconciliation, and one drift persists while another closes. Compounding each group's 9 September YTD by Thursday's one-day move reproduces the published YTD to within 0.04 percentage points for ten of the eleven groups. Worked examples: energy 1.4189 × 0.9961 = 1.41337, or +41.34% against a published +41.34%, deviation zero; technology 1.2560 × 0.9874 = 1.24018 → +24.02% against +24.02%, deviation zero; consumer cyclical 0.9334 × 0.9949 = 0.92864 → -7.14% against -7.10%, deviation 0.04. Financial is the single failure and it fails in the same direction and roughly the same size as yesterday: implied 1.0769 × 0.9974 = 1.07410, or +7.41%, against a published +6.88% — a 0.53 pp excess after Wednesday's 0.55 pp. Technology's 0.47 pp shortfall from Wednesday has closed completely. Two consecutive same-signed deviations in one group while the other resolves in a single session is the signature of a constituent change that has now finished working through technology and is still working through financial. The vendor figure is published, flagged and carried.

The composition traps. Communication services led at +0.24% on Charter +4.98% to $140.56 and Comcast +2.36% to $25.17 — a mechanical bounce off Wednesday's 8.13% and 6.65% cable rout rather than any new information, with Fox +2.07% and Take-Two +2.76% alongside and Alphabet A +0.60% the megacap contributor. Technology at -1.26% understates the semiconductor damage for the first time in five sessions: SOX -2.66% against a group figure diluted by Apple +3.58%, Microsoft +0.16%, Fortinet +1.04% and Accenture +1.20%. Consumer defensive scraped +0.08% on Philip Morris +2.19% and Constellation Brands +1.86%, against Campbell's -4.30% and General Mills -3.06%, which is an eighth consecutive decline for the latter. Healthcare at -0.69% conceals a 19-point spread between Elevance +4.95%, Molina +3.17% and Centene +2.09% on the managed-care side and Labcorp -4.40%, Boston Scientific -4.15% and Cooper -14.67% on the other.

Higher — the Apple complex, managed care and the cable bounce.

Lower — semiconductors, hardware, the metals and everything with a guide.

Analyst and corporate actions, 10 September.

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 an S&P 500 constituent list. 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.

Fri 9/11. BMO: Kroger (KR).

Mon 9/14. No S&P 500 reporter on either bucket.

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.

Changes vs. the prior calendar (9/10 report):

Source: U.S. Department of the Treasury daily par yield curve, 10 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 3 September row.

A belly-led bear flattener, and the diagnostic is policy rather than term premium. The cheapening peaks at the 3-year and 5-year, both +14 bp, falls away monotonically in each direction — 2-year +13, 7-year +13, 10-year +12, 20-year +11, 30-year +9 — and the 3-month bill moves only 5 bp. That shape is a repricing of how many hikes arrive and how soon, not of the compensation for holding long duration: if the market were adding term premium the 30-year would lead, and it lagged the 5-year by five basis points. 2s30s tightened 4 bp to 81 bp and 10 bp on the week, and 2s10s is now 39 bp, the flattest since the front end began cheapening. The week is the more striking frame: the 5-year is 23 bp cheaper than a week ago and the 30-year 12 bp, so the entire move is concentrated where the Federal Open Market Committee's next four decisions live.

The 20s30s inversion is the buyback's fingerprint. The 20-year rose 11 bp to 5.39% and the 30-year 9 bp to 5.37%, putting 20s30s at minus 2 bp after two sessions at zero. Treasury bought $5.19bn of 10-to-20-year debt against a $6bn maximum on Thursday morning, from $10.5bn of offers, and Bloomberg reported that Treasuries cheapened versus interest-rate swaps with the move most pronounced at the 20-year. An operation designed to support a bucket that ends with that bucket the cheapest point on the long end is the clearest evidence available that the market read the undershoot as information rather than as execution. Supply corroborated: the 30-year auction stopped at 5.308% against 5.216% at the prior auction, 9.2 bp of concession in a month, following Wednesday's 10-year at 4.834% against 4.683%.

The bill curve says this is not a funding event, and the off-table tenors carry the detail. The 1-month rose 10 bp to 3.91% and the 2-month 8 bp to 4.01%, while the 3-month rose only 5 bp to 4.00%, the 4-month 5 bp to 4.11% and the 6-month 6 bp to 4.07%. Two off-table observations matter. The 2-month at 4.01% is now 1 bp above the 3-month, a fresh inversion at the very front that did not exist on Wednesday, and it sits precisely across the 16 September meeting date. And the four-month-above-six-month inversion narrowed to 4 bp from 5 bp, so the localised cheapening around the December meeting that this report flagged on Wednesday did not extend. A bill complex that moves 5 to 10 bp while the belly moves 14 is a curve pricing meetings, and the 1-month's outsized 10 bp is the single tenor that now fully spans the September decision.

The vendor-versus-official gap, explained. Bloomberg's real-time board closed the 10-year at 4.96%, +12 bp, against the official par 4.95%, and put the 2-year at 4.59%, +16 bp, against the official 4.56%, +13 bp. Investrade's evening review recorded the 10-year at 4.943%, +10.4 bp, and the 30-year at 5.35%. The official par curve is struck from bid-side quotes at approximately 3:30 p.m. ET while the vendor boards run to 5:00 p.m. and beyond, so a session that cheapened steadily into the close produces exactly this pattern — vendors above par, and the gap widest at the tenor that moved most. It is a timing artefact, not a level dispute, and the direction and rank order agree at every point.

Source: Federal Reserve Bank of New York Economic Indicators Calendar for September 2026 (all times Eastern), cross-checked against the TradingEconomics United States calendar and the Wall Street Journal market-data calendar, both read this session. Consensus figures for 11 September and the week of 14 September were independently re-captured this session.

Current week (Sep 7 - Sep 11) — still to come

Next week (Sep 14 - Sep 18)

Current target range: 3.50%-3.75%. A producer-price headline the core did not support moved the September meeting more than ten points in one session.

CME FedWatch headline — 16 September 2026 meeting. Data as of 10 Sep 2026, 05:01:45 p.m. CT (6:02 p.m. ET), read from the FedWatch probability table. Contract ZQU6.

Provenance of every column, stated, and the live-read bias returned. The footer timestamp reads 05:01:45 CT with no meridian; the read was taken at approximately 6:05 p.m. ET, roughly an hour 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 9 September and prints 61.2%, against the 60.2% this report published from CME's live column on Wednesday evening — a +1.0 percentage-point correction, after a zero correction on Wednesday and +0.8 on each of the two sessions before that. Wednesday's clean read is therefore confirmed as the exception rather than the new normal, exactly as this report said at the time: it is a sampling artefact of when the read is taken and no standing adjustment should be built for it. 1 WEEK (3 September) at 49.4% and 1 MONTH (10 August) at 52.2% carry genuine reference dates and are used below. The Investing.com matrix underneath is timestamped 10 Sep 2026, 05:45 p.m. EDT and is the primary source for parts (a), (b) and (c).

The CME-versus-Investing.com gap, quantified. CME puts the September hike at 71.3% at 6:02 p.m. ET; Investing.com at 71.0% at 5:45 p.m. ET — a 0.3 percentage-point difference across seventeen minutes, against 0.1 on Wednesday and 1.0 on Tuesday. Investing.com publishes the September future at 96.288 against 96.300 on Wednesday. Because the 16 September meeting sits mid-month, only about 47% of the contract's averaging period is affected, so one basis point of ZQ price is worth roughly ten percentage points of hike probability — and the 1.2 basis points the front contract cheapened maps to about twelve points of probability against the 10.6 points the vendor's own columns show. The residual is the day-weighting, and the gap is again inside a rounding step.

One-day, one-week and multi-day momentum, and this is the largest single-session move of the reporting window. The September hike rose 10.1 points on CME's own columns, 71.3% from 61.2%, and 10.6 points on Investing.com's own columns, 71.0% from 60.4%. ZQU6 cheapened 1.2 bp to 96.288 and ZQZ6 8.0 bp to 95.965 — the December contract moved nearly seven times the September one, which is the whole story of the session in two numbers. The multi-day read: the meeting sat at 49.4% a week ago and 52.2% a month ago, so it is 21.9 points more hawkish than a week ago and 19.1 points more hawkish than a month ago, and the week's path is 49.4 to 60.4 to 61.2 to 71.3. Further out on the vendor's own columns, October's cumulative-above rose to 82.3% from 72.1% and December's to 94.5% from 87.6%. The probability of a cut at any 2026 meeting remains 0.0%.

(a) Current-year meeting distributions

Investing.com Fed Rate Monitor, updated 10 Sep 2026 05:45 p.m. EDT. Format: current [prior day] [prior week]. Modal range in bold.

All three rows sum to exactly 100.0%. Three observations. First, the vendor's prior-day column no longer reproduces the prior edition: it prints 39.6% and 60.4% at September against the 39.9% and 60.1% published on Wednesday, a 0.3-point drift that ends a three-session clean run and reinstates the fixed-snapshot caveat. Second, the December mode has flipped. Two hikes now lead at 46.3% against 29.1% for one — a 17.2-point lead where one hike led by 1.4 points on Wednesday, by 4.7 on Tuesday and by 9.9 last Thursday. The compression this report tracked across four sessions did not merely continue; it inverted and then ran seventeen points past the crossover in a single day. Third, the tails were bought again and far harder: October's +50 bucket to 27.7% from 16.7% and December's +75 to 19.1% from 9.6%, together 21.5 points of tail mass added against 3.2 on Wednesday and 1.3 on Tuesday. Meanwhile the hold bucket collapsed at every meeting — September to 29.0% from 39.6%, October to 17.7% from 28.7%, December to 5.5% from 12.1%. A December hold at five and a half points is a market that has stopped treating the current range as a live outcome for this year.

(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.

Seven of the eight 2027 meetings are now modal at 4.25%-4.50%. On Wednesday only September and October 2027 had crossed, by four-tenths of a point; on Thursday everything from March onward crossed and January is the sole survivor at 4.00%-4.25%. The eight contracts print 95.905, 95.750, 95.670, 95.560, 95.535, 95.545, 95.500 and 95.550 against Wednesday's 95.995, 95.890, 95.835, 95.755, 95.735, 95.710, 95.705 and 95.715 — 9.0 to 20.5 basis points cheaper, against 1.5 to 4.0 basis points on Wednesday, so the back end cheapened roughly five times as fast as it did the day before. Set that against the front, where September 2026 cheapened 1.2 bp and December 2026 8.0 bp. The route moved, the destination moved five to seventeen times more, and the cheapening is deepest at the October 2027 contract at 20.5 bp rather than at the far end — which is a market adding hikes into 2027 and then holding them, not extending a trend. Cumulative-above at December 2027 rose to 97.4% from 95.7% and the first non-trivial cut probability sits at 0.3% at 3.25-3.50%, up from 0.2%.

(c) Year-end probability ladders

Year-end 2026 — the 9 December meeting.

Cumulative above the current range: 94.5%. Cumulative below: 0.0%. Sum: 100.0%.

Year-end 2027 — the 8 December meeting.

Cumulative above the current range: 97.4%. Cumulative below: 0.3%. Sum: 99.9%.

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. The 2027 year-end ladder sums to 99.9% and the June 2027 row to 100.2% for exactly this reason; the three 2026 rows sum to exactly 100.0%.

(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 9 September 2026 as-of date, not the 10 September close. Same-day direction is cross-checked against the cash-market proxies underneath and against Bloomberg and WSJ credit coverage. 1-Week is versus the 2 September row.

CDX — the six-step ladder was worked again in the local Chrome browser, and four of the six steps could be executed. (1) Bloomberg in Chrome: /markets and /markets/rates-bonds both rendered fully, and a full-text scan of each returns zero occurrences of the index name and zero of "credit default"; the fixed-income tables carry Bloomberg's own aggregate indices and a global ten-year government board, not credit-default-swap levels. (2) WSJ Market Data bonds page in Chrome: rendered with its Treasury, consumer-rate, government-bond and economic-calendar tables populated, and a full-text scan returns zero occurrences. (3) Cbonds CDX.NA.IG 5Y page: reached this session and it does carry a live record — the page shows a quote stamped 08/09/2026 with the level itself masked behind a "you need to request access" wall, printed as asterisks in place of the basis-point figure. That is the closest any step has come in the window and it is still not a publishable level. (4) FT Markets Data and (5) Barchart are both refused by the Chrome extension's domain policy ("Navigation to this domain is not allowed"), so those two steps could not be executed at all — a tooling failure rather than an absence of data, recorded as such rather than reported as a clean failure. CME's credit-index product pages carry specifications rather than levels. (6) Cash-market proxies, labelled as proxies: HYG closed $78.62, -0.46%, and LQD $104.34, -0.93%. No CDX level is published here.

The investment-grade cash market made a new low and the spread series still has not moved. LQD fell 0.93% to $104.34 and closed at a fresh 52-week low — the previous low, set intraday on Wednesday, was $105.08, so the floor moved 74 cents in one session — while the IG credit spread sat unchanged at 81 bp for a fourth consecutive update. Wednesday's configuration has not merely persisted, it has intensified: the price is making new lows at an accelerating rate and the compensation over Treasuries is a flat line, because what is repricing investment-grade credit is a 4.95% ten-year and not the credit itself. HYG fell 0.46% to $78.62 against a 52-week low of $78.57, so the high-yield proxy is now five cents, or six basis points of price, from its own one-year low, having been 0.52% away on Wednesday. On the index series the 9 September stamp finally showed the widening the proxies had been signalling: HY widened 4 bp to 271 and CCC 8 bp to 1,064, taking the CCC-minus-HY differential to 793 bp from 789 and the CCC series 11 bp wider on the week and 176 bp on the year. IG unchanged at 81 bp.

(b) Money-market & funding plumbing

New York Fed reference rates, published at approximately 8:00 a.m. ET for the prior business day. The 8 September 2026 row published to the reference-rates endpoint this session, so the table advances four business days from the prior edition's 4 September basis; no 9 September row had published at capture. The operations figures beneath it are dated 10 September. Rate up = red.

The plumbing did two things and they point opposite ways. Reverse repo take-up jumped to $4.736bn on 10 September from $432m on 9 September, an eleven-fold move and the largest single-session change of the reporting window, on the day the entire coupon curve cheapened 9 to 14 basis points. Cash going back into the facility while term rates rise is money-fund cash declining to term at the new levels — a duration judgment expressed at the front end, not a scarcity signal. Against that, reserve balances printed $2.9913tn for the week ended 9 September, up $96.8bn and fully reversing the $30.4bn drain this report flagged last week, which removes the tightening framing entirely with quarter-end fourteen days away. SOFR at 3.64% is 1 bp below the 3.65% IORB on the 8 September row — the first sub-administered print of the window — with the 99th percentile unchanged at 3.73% on $2,904bn of volume. The 10 September standing repo facility result had not published to the endpoint at capture and is not asserted; the last known figure is $5m on 9 September, a third consecutive non-zero operation.

The off-table bill tenors belong here, and this session they carry a new inversion. The 1-month rose 10 bp to 3.91%, the 1.5-month 5 bp to 3.93%, the 2-month 8 bp to 4.01%, the 4-month 5 bp to 4.11% and the 6-month 6 bp to 4.07%, against the 3-month's 5 bp. The 2-month is now 1 bp above the 3-month for the first time in the window, and the 2-month bill is the shortest instrument that fully spans the 16 September decision — so the front of the bill curve has begun pricing the meeting directly rather than through the coupon complex. The four-month-above-six-month gap narrowed to 4 bp from 5 bp, so the December-meeting localisation did not extend. A bill complex that moves 5 to 10 bp on a session when the 5-year moves 14 is still saying the pressure is policy expectation rather than funding, and the NY Fed bought $2.122bn of 4-to-12-month bills during the session without disturbing it.

(c) Rates volatility & swap spreads

The MOVE vintage advanced again and the internal consistency check passed for the first time in eleven sessions. The Investing.com series now carries an 09/09 date stamp at 76.74, with a day range of 76.14 to 76.74 that contains its own level — the "previous close" that had been printing outside the day range since late August has resolved. The 9 September value is 0.60 points, or 0.79%, above the 76.14 that carried an 8 September vintage, so rate volatility rose modestly into Wednesday's auction and buyback news. What has not appeared is any 10 September value: the card's change field reads 0.00, and on the session that produced the largest one-day 2-year move since April 2025 the rate-volatility series is silent. The level is published with its vintage; the change is withheld, and the MOVE-versus-VIX ratio is computed on a one-day-stale numerator and flagged. Alongside it, VIX rose 8.38% to 17.84 with an intraday high of 18.17, breaking above 17 after 28 consecutive sessions inside a 14-to-17 band. Two internals corroborate the breadth deterioration behind it: the S&P 500 McClellan Oscillator fell below -72, its lowest in more than five months, and only about 36% of S&P 500 members trade above their fifty-day moving average. 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 Bloomberg's report that Treasuries cheapened versus swaps with the move most pronounced at the 20-year, the tenor the buyback targeted, which is corroborated on the cash curve by 20s30s inverting to -2 bp.

(d) Issuance, leveraged loans & private credit

The primary market is shut and the level that shut it got 12 basis points worse. The constraint this report has named for a fortnight is the yield rather than the spread, and Thursday moved the yield decisively: the 10-year closed at 4.95%, on the cusp of its late-2023 peak, the 30-year at 5.37%, a 19-year high, and LQD set a fresh 52-week low at $104.34. The muni market is the visible casualty — long-dated municipal yields rose 10 to 15 bp on the session and more than 50 bp since end-June, on over $15bn of weekly supply led by an Alabama toll-road deal of $3.82bn, the year's largest — and it is the market where issuers cannot defer. Post-Labor-Day investment-grade issuance is running at its weakest since 2020, after an August that set a record near $130-145bn against a post-2019 August average of roughly $95bn, and year-to-date supply through August above $1.68tn, up 27% on 2025, against a full-year projection of $2tn or more that the current pace makes a stretch. The concrete prints remain Dell Technologies seeking $4bn for refinancing and Enbridge's $2.55bn Tallgrass acquisition, which has to be funded somewhere and which lifted the whole midstream complex on Morgan Stanley's read-across upgrades of DT Midstream and TC Energy. 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 are vendor financing that no published spread series captures, and Broadcom closed -0.97%. Alongside it, Oracle's after-hours record backlog disclosure is the other side of the same ledger, and S&P Global cut Oracle to the lowest investment-grade rating in July with free cash flow expected negative through 2029.

Source: TradingEconomics currency board, read after the U.S. close on the vendor's Sep/10 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/10 boundary and covers the full U.S. session.

The take: two currencies did the opposite of what the day's news said, and both are the same trade. The euro fell 0.21% in the hours after the European Central Bank raised its deposit rate to 2.50% from 2.25% and its president called the decision a "no brainer." A currency that loses ground on a confirmed hike is a currency whose rate differential is moving against it faster than its own central bank can move, and the number that explains it is on the U.S. curve: the American 2-year cheapened 13 bp against the Bund's 6 bp at ten years. The second and stranger one is the Norwegian krone, down 0.88%, the worst move on the board, on a session when Brent rose 7.98% to $109.29. The Canadian dollar did the same thing more mildly at -0.20%. Two petro-currencies weakening into the largest single-day crude rally of the war is a market pricing the dollar's rate story ahead of every terms-of-trade story, and it is the cleanest read available that Thursday was a U.S. rates event wearing an energy costume.

The haven cross confirms it and the yen breaks a pattern. USD/CHF rose 0.33% — the franc weakened — on a day of an 8% crude rally, a 2-year selloff and a VIX breakout above a 28-session range. Havens that will not bid when three separate risk signals fire together are havens whose bid is being outcompeted by cash, and at 4.00% on the 3-month bill it is. USD/JPY rose 0.61% to 154.494, the yen's first meaningful weakening in the reporting window, on a session when 10-year JGBs cheapened 3 bp rather than richening. This report has flagged a repatriation signature for three sessions — a Japanese curve that rallied while every other developed curve was sold, with a currency that would not give the move back. Both halves broke at once, which resolves the question in favour of rate differentials rather than flow, with the Bank of Japan on 18 September eight days out. The yen is still 0.84% stronger on the week.

Asia moved as one and the mainland did not move at all. USD/KRW +0.72%, USD/TWD +0.61% and USD/INR +0.52% cluster within twenty basis points of each other, which is dollar strength rather than any local story — the won weakened on a session the Kospi fell only 0.25%, and the Taiwan dollar weakened on a session Taiwan Semiconductor reported 53.3% revenue growth. Against that, USD/CNY moved 0.12% for the second consecutive session, and the yuan is 3.76% stronger year to date with a week-on-week change of five basis points. A managed currency holding still through an 8% oil move and a 13 bp U.S. front-end selloff is the single most deliberate price on this table.

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 two editions. Rows were captured at approximately 18:15 ET and every row except Brent carried a volume far below the prior session's, which marks it as still forming under the rule this report set on 9 September; each was therefore corroborated against the TradingEconomics spot board on the same capture and against Bloomberg's board before publication, and the corroboration is quantified below. The prior edition's 9 September rows have since finalised and are restated here, with every daily change and crack spread recomputed on the finalised base. 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/10 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 forming-row check, quantified, because this is the trap that cost the 8 September edition. Brent's row carried 518.27K of volume against 392.07K on 9 September — above the prior session, so it is a completed settle and needs no qualification. Every other row carried volume between 0.12K and 20.42K against six-figure prior sessions, so each was checked against TradingEconomics spot on the same capture: crude 103.997 against the board's 104.08, Brent 109.108 against 109.29, heating oil 5.1343 against 5.1353, gasoline 3.4319 against 3.4327, natural gas 2.8378 against 2.838 — five energy rows agreeing to within nine cents on a $104 barrel and one-thousandth of a cent on the products. The metals agree in direction and magnitude with the expected futures-over-spot basis: gold spot 4,320.15 against Comex December 4,362.01 (a $41.86 or 0.97% carry) and silver spot 63.549 against 64.060 (0.80%), with spot day changes of -1.77% and -5.53% against the board's -2.21% and -6.68%. Every published row is therefore corroborated, and the two energy contracts that drive the crack arithmetic are corroborated to within a tenth of a percent. Investrade's evening review recorded WTI at $102.48 and Brent at $107.63; those are earlier intraday electronic prints from before the settlement window, they agree in direction and are not adopted, and under the rule this report set on 9 September a third-party figure that disagrees with a corroborated row is evidence about the timestamp rather than about the row.

The 9 September restatement. Published against finalised: WTI $97.15 against $96.05, Brent $101.75 against $101.21, gold $4,447.35 against $4,460.70, silver $67.925 against $68.646, copper $6.8583 against $6.8885, RBOB $3.2236 against $3.2211, heating oil $4.8023 against $4.8010, natural gas $2.805 against $2.822. The gaps are far smaller than the 8 September restatement — the largest is $1.10 on WTI against $1.24 a session earlier — which is what a one-hour-earlier capture buys. Every daily change above is computed against the finalised 9 September row. The restated crack spreads for 9 September are distillate $105.59 and gasoline $39.24 for a differential of $66.35, against the $104.55, $38.24 and $66.31 published — so the differential moves by four cents and the two legs by about a dollar each.

The crack spreads on a consistent October basis against $104.08 WTI:

Gasoline finally joined and it still lost the race by six dollars. For two sessions the gasoline crack fell while the distillate crack rose; on Thursday both rose, but heating oil's 6.96% against gasoline's 6.57% was enough to widen the differential another $5.16 to a window high of $71.51. The mechanism has not changed and the data confirmed it in the same twenty-four hours: the August PPI showed diesel prices up 24.1% in a single month, more than a third of the entire final-demand goods increase, and that is the war arriving in an official statistic rather than a futures screen. Distillate stocks built 2.087m barrels against a -0.7m consensus on the same morning and the crack widened anyway, which says the physical build is not in the places the shortage is.

Eight per cent on a triple inventory miss is the number to carry. Brent settled $109.29 and WTI $104.08, taking the Brent-WTI differential to $5.21 from a restated $5.16 — five cents of widening on an eight-per-cent move, so the seaborne-versus-landlocked risk premium did not reprice at all even though the catalyst was attacks around the Strait of Hormuz. Against that, the EIA report cut the other way in every line: crude -0.391m against -1.6m expected, gasoline +1.269m against -1.4m, distillate +2.087m against -0.7m, and OPEC cut its 2026 demand-growth forecast to 380,000 barrels a day from 580,000. A market that adds eight per cent through all of that is pricing transit risk, not balances — and WTI is now +13.92% on the week and +81.13% on the year.

The metals broke on the real rate and the break was the cleanest signal of the session. Silver fell 6.68% to $64.060 and gold 2.21% to $4,362.01, with platinum -7.03% and copper -5.27% alongside. Gold's failure is now a level: it has turned back from roughly $4,480 on three consecutive sessions and Thursday's high of $4,476.05 came within four dollars of Wednesday's before selling $121 to a $4,355.00 low. Gold's spot year-to-date return is now -0.01% — the entire 2026 gain is gone. The transmission is explicit: a PPI headline whose core undershot pushed nominal 5-year yields up 14 bp without moving realised inflation, which raises the real rate, and the metal that had been trading the war premium had to give it back. The gold-silver ratio jumped to 68.09 from a restated 64.98, a 3.11-point move in one session that unwinds five consecutive sessions of silver outperformance. Silver and copper falling five to seven per cent while crude rises eight is the same divergence this report flagged on Wednesday, running at three times the speed and in the opposite direction — the metals are not trading the shock the barrel is trading, and on Thursday they traded the rates market instead. Natural gas rose 0.57% to $2.838, its first gain in three sessions, and remains -23.01% on the year, the only major commodity deeply negative in a complex where two are above eighty.

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 made its best money on the day its own invalidation fired; close it

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 3 September at 96.085 / 95.820 for a spread of 26.5 bp, quarter size. Thursday's mark: ZQZ6 95.965, ZQZ7 95.550 — a spread of 41.5 bp. That is +8.5 bp on the session, worth +$354.20 per contract pair before costs on a quarter, and the position is +15.0 bp from entry, or +$625.05 per pair. It is by a factor of three the best session the trade has had.

And the invalidation fired on the same session.* The written invalidation included "the December 2026 mode reverting to two hikes," which Wednesday's edition flagged as 1.4 points away and the closest of the three to firing. On Thursday it did not merely fire, it inverted: two hikes at 46.3% against one hike at 29.1%, a 17.2-point lead. The position is therefore closed at the gain, not held. That is the discipline the invalidation exists to enforce, and a trade that pays on the day its thesis breaks is exactly the case where the rule earns its keep — the front leg cheapened only 1.2 bp while the back cheapened 16.5 bp, so the spread widened because the market added hikes into 2027, which is the opposite of the front-end-anchoring the trade was built on. **Closed. Final mark +15.0 bp, +$625.05 per pair.*

2. New — the rates trade, long ZQU6 against short ZQZ6, quarter size

Expression: long ZQU6 (September 2026) against short ZQZ6 (December 2026), DV01-matched one-for-one at $41.67 per basis point per contract, quarter size, entered at 96.288 / 95.965 for a spread of 32.3 bp. The position gains as the spread widens — that is, as the September meeting is repriced down relative to the cumulative path priced into December.

The modal path, base case and tails. Modal path: a 25 bp hike on 16 September at 71.3% on CME and 71.0% on Investing.com, with ease at 0.0%; one hike is still modal at October (54.6%); and December's mode has flipped to two hikes at 46.3%, with seven of eight 2027 meetings now modal at 4.25%-4.50%. Base case: the committee hikes in September and the strip's cumulative-above at December stays near 94.5%, so the destination is right and only the schedule is in question. Tail one, and it is the whole trade: CPI at 08:30 on 11 September, consensus +0.4% headline and +0.2% core. August PPI split cleanly — headline +5.4% year on year, core month-on-month +0.2% against +0.3% expected, with goods +1.1% against services +0.1% — and if CPI reproduces that split, a 71.3% September is the expensive leg, because the argument that this is a relative-price shock lands on the meeting six days away and not on the 2027 strip. Tail two: a core CPI at 0.3% or above, which takes September toward certainty and cheapens the front leg faster than December, costing this spread. Practical implication: at roughly ten percentage points of probability per basis point of ZQ price for a mid-month meeting, the September leg carries about four times the probability sensitivity per basis point that December does, so the spread is a genuinely asymmetric expression of the CPI split rather than a directional rates view — which is why it is a quarter and not a half. Catalyst: CPI 9/11 at 08:30; Michigan one-year inflation expectations at 10:00, prior 4.0%; retail sales 9/16; the 16-17 September FOMC with projections. Invalidation: the spread through 25.0 bp; or the September cumulative hike through 85% on either vendor; or a December 2026 hold probability back above 15%, which would say the whole path is being unwound rather than the schedule. Sizing: a quarter, at $41.67 per basis point per pair.

3. Long the 20-year against the 30-year — the operation printed and the result was the thesis inverted; close it

Expression and mark.* Long the 20-year bond against short the 30-year, DV01-neutral, half size. 20s30s at -2 bp, with the 20-year at 5.39%, +11 bp and the 30-year at 5.37%, +9 bp — so the 20-year underperformed by 2 bp on the session and the position is -2 bp against entry. The honest reading. Wednesday's note said the trade had received "a lucky right answer to a wrong question" and instructed holding the half until the operation itself printed. It printed. Treasury bought $5.19bn of 10-to-20-year debt against the $6bn maximum, out of $10.5bn of offers, and Bloomberg reported the cheapening versus swaps was most pronounced at the 20-year — the bucket the operation targeted became the cheapest point on the long end on the day it was targeted. The written invalidation of 20s30s through -3 bp is one basis point away and did not technically fire, but the event the trade was waiting for has now happened and delivered the opposite of what the trade needed. Action: close the half at -2 bp rather than hold for a mechanical trigger on a thesis that the operation has already answered. **Closed. Final mark -2 bp.*

4. Protection on the CCC cohort funded in IG — paid a third time while the funding leg got expensive; cut to a quarter and hedge the duration

Mark. CCC 1,064 bp, +8 bp; HY 271 bp, +4 bp; IG 81 bp, unchanged on the 9 September FRED update, taking the CCC-minus-HY differential to 793 bp from 789 — a 4 bp gain, for a cumulative +27 bp across seven updates, on six of which the pair has been paid. The honest reading, and the deterioration Wednesday flagged has arrived. The thesis is working better than ever: this is the first update in the window where both HY and CCC widened by multiple basis points at once rather than the tail moving alone, and IG has now not moved for four updates. The problem is the funding leg. LQD fell 0.93% and closed at a fresh 52-week low of $104.34, against -0.16% on Wednesday — the duration bleed on a long-IG-cash leg is now running at nearly six times the prior session's rate and it exceeds the spread gain in dollar terms. Action: cut from a half to a quarter and express the funding leg duration-hedged, as flagged on Wednesday. A structure whose funding leg loses more to a 4.95% ten-year than the thesis earns from the tail is not the trade that was written, whatever the spread series says. Catalyst: the September IG calendar at its weakest post-Labor-Day pace since 2020, which now has to clear against a 4.95% ten-year; CPI 9/11; quarter-end funding from the 15th. Invalidation, unchanged for the residual: the differential back through 750 bp, or IG widening beyond 90 bp. Sizing: cut from a half to a quarter, duration-hedged.

5. Long volatility on the semiconductor complex — the best session of the position; suspend the time-decay override and hold through CPI

Mark. October volatility on the semiconductor proxy, a quarter, expressed in premium rather than delta. SOX fell 2.66% to 11,614.2 against the Nasdaq 100's 1.08% and the S&P 500's 0.58%, with VIX +8.38% to 17.84 and an intraday high of 18.17 — the first print above 17 in 28 sessions. Every leg of the position paid at once for the first time since entry. The clock, precisely, and the override. The written invalidation is "SOX outperforming the Nasdaq 100 on two consecutive up sessions." Thursday was not one on either count — the Nasdaq 100 fell and SOX underperformed it by 1.58 points — so the counter still cannot advance, and the Nasdaq 100 has not had an up session since 4 September, four sessions ago. Wednesday's edition added a time-decay override: close on decay if the condition cannot evaluate for two more sessions. That override is suspended, for cause, and the reason is stated rather than assumed. It was written to stop a position bleeding theta while nothing happened; on Thursday something happened, the surface repriced 8.38%, and the position is no longer decaying. Closing a long-volatility position the session before the only Very-high release on the calendar because a stale mechanical clock says so would be following the letter of a rule against its purpose. Action: hold the quarter through CPI; mark it out on 11 September regardless of outcome, and replace the invalidation with one that can evaluate — SOX outperforming the Nasdaq 100 by more than a point on any two of five sessions, up or down. Catalyst: CPI 9/11 at 08:30; the 16-17 September FOMC. Sizing: a quarter, unchanged.

6. Short the credit-bureau complex against long the S&P 500 financials — the invalidation retreated; hold the quarter

Mark. Fair Isaac -2.29% to $960.67 and Equifax -0.03% to $168.04; TransUnion again did not appear in the 494-line component capture and no independently sourced close is asserted, so the basket is marked on two of three names at -1.16%, against S&P 500 financials at -0.26% — a 0.90-point gain on the pair, and cumulatively -1.23 points across three sessions. The honest reading. Wednesday's cut to a quarter was made because Fair Isaac's 5.35% bounce had taken it within 2.88 points of the invalidation trigger. It gave back 2.29% on Thursday, so the cumulative recovery from the post-rout level is +2.94% against an 8.34% trigger — the invalidation is now 5.40 points away, having been 2.88. Cutting the size ahead of a trigger that then retreated cost performance on the residual, and that is recorded as a sizing error rather than a thesis error. Action: hold the quarter; do not re-add to a half on one favourable session in a basket that can only be marked on two of three legs. Catalyst: any FHFA follow-through on bi-merge; sell-side mortgage-channel revisions. Invalidation, unchanged: the basket recovering more than half of the 4 September decline within five sessions — two sessions remain — or an explicit FHFA statement retreating from bi-merge. Sizing: a quarter, dollar-neutral.

7. Long the October distillate crack against the gasoline crack — the mechanism was tested and held; hold the quarter, do not add

Mark. Entered on 9 September at a $66.31 differential on the then-published basis, which restates to $66.35 on the finalised historical-board settles. Thursday: distillate crack $111.60, gasoline crack $40.09, differential $71.51 — a $5.16 gain on the restated basis and a window high. Why the test matters more than the gain. The written invalidation included "a crude session above 2% in which heating oil underperforms gasoline," which would break the mechanism. Thursday delivered a +8.36% crude session, the largest of the war, and heating oil outperformed gasoline 6.96% to 6.57% — a narrow win, but on the largest possible test. Two independent corroborations arrived with it: August PPI showed diesel up 24.1% in a single month, over a third of the entire final-demand goods increase, and distillate stocks built 2.087m barrels against a -0.7m consensus while the crack widened regardless. A paper crack that widens through a reported build is a crack pricing where the barrels are rather than how many. Action: hold the quarter, do not add at a window high. The entry was already a re-entry $4.83 above the level this report closed the same pair at on 3 September; adding another $5.16 higher would be chasing. Catalyst: next week's EIA report; any Gulf de-escalation headline; CPI 9/11, which will carry the retail energy pass-through. Invalidation: the differential back through $55; or a crude session above 2% in which heating oil underperforms gasoline. Sizing: a quarter, barrel-for-barrel.

Prior closes, marked forward. The AI-halo basket against long Nvidia, closed on 1 September, would have gained for the first time: marked on the two legs with independently captured closes, Fortinet +1.04% and Adobe -2.32% average -0.64%, against Nvidia -2.26% — a 1.62-point gain on the pair, after two consecutive losing sessions. The short-debasement basket against long dollar, closed on 3 September, would have won on both legs and won large: gold fell 2.21% and DXY rose 0.28%, the best session either leg has produced since the idea was closed, which is a reminder that the close was a discipline decision rather than a view.

The vol note. VIX closed 17.84, up 1.38 points or 8.38%, with a session range of 16.29 to 18.17, on a day the index fell 0.58% — a 14.4-to-1 ratio of volatility gain to index decline, and the first close above 17 in 28 sessions. The five-observation path is 14.32 → 14.53 → 15.72 → 16.46 → 17.84, so the surface has added 3.52 points, or 24.6%, in four sessions while the S&P 500 has fallen a cumulative 1.64% across the same span. A 17.84 handle asks for roughly a 1.12% daily move against realised index moves of 0.38%, 0.58%, 0.48% and 0.58% — so index vol is now emphatically rich to realised, and the case that was for dispersion over index vol has become a case for selling index gamma into the CPI print rather than owning it. Two internals argue the other way and are stated: the McClellan Oscillator below -72, its lowest in five months, and only 36% of S&P 500 members above their fifty-day average, both of which say the breadth deterioration under the index is real even where the index level is not moving. The instruction is therefore split rather than clean: own the 11 September inflation block outright — it is the one event the realised-versus-implied argument does not cover — and finance it in the 17-18 September meeting expiry, where a 71.3% meeting probability still carries the richest premium on the curve and where a decision that is now three-quarters priced has less left to deliver than a print that is not priced at all.

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

What VIX is and is not pricing. At 17.84, up 8.38% and above 17 for the first time in 28 sessions, the index option surface is asking for roughly a 1.12% daily move against four consecutive realised moves of 0.38%, 0.58%, 0.48% and 0.58%. It is pricing the CPI print and the meeting. It is not pricing what sits underneath the index: breadth of 1.89-to-1 negative on a session when the loss was concentrated in SOX -2.66%, a McClellan Oscillator below -72 at a five-month low, only 36% of members above their fifty-day average, and a single-name distribution running from Skyworks +9.79% to Cooper -14.67% — a 24.5-point spread inside one index on a 0.58% day. And it is not pricing the cross-asset tail at all: the largest one-day 2-year move since April 2025, an 8% crude session, and a rate-volatility series that did not publish a value. A market whose index vol is rich to its own realised while its internals are at five-month extremes is a market where the dispersion is the risk and the level is the distraction.

Full Data Notes & Conflicts, the Overnight / Asia & Europe read-through and the complete categorized source links are in the companion file US_CrossAsset_Daily_2026-09-10_DataNotes.txt.