The bottom line
- The RBA decides at 14:30 and three of seven house views settle on it, which is a concentration this desk named a week ago and has now failed to reduce twice. A 25bp hike to 4.60% is effectively fully priced — the tracker prints 90.0% but its site-wide stamp is 25 September and its own components sum to 1.54 hikes against a headline of 1.0, so treat it as a soft Friday mark rather than a Monday one; all twenty-nine economists in one survey and all four majors call 4.60%. The decision is therefore the least informative part of the event and the guidance is the whole trade. The street splits two-two: ANZ and CBA read 4.60% as a way-station (ANZ explicitly to 4.85% in November), NAB and Westpac read it as terminal, and Westpac expects a split vote where CBA expects unanimity. The physical curve sides with the hawks — the 3-year ACGB at 5.04% sits 44bp above a post-hike 4.60% cash rate, which is roughly another 1.75 hikes of embedded premium, not the ~1.0 the tracker claims. Cutting the other way, and under-priced: the last substantive signal from a voting member was Iain Ross on 22 September concluding there is “no evidence of the emergence of a wage-price spiral… and recent data suggest such an outcome is unlikely”, and there has been no RBA commentary of any kind in the six days since. A statement that frames 4.60% as the top has ~44bp of 3-year premium to unwind.
- The level this desk built a view on yesterday does not exist, and correcting it is the most important thing in this edition. V028 was opened on Monday around “7,680, where the 200-day (7,679.54) and the dealer gamma flip (7,680) coincide” — support and the damping mechanism failing at one price. Both legs are wrong. The S&P 500's 200-day moving average is approximately 7,205, not 7,679.54 — an error of 474 points, confirmed on two independent sources agreeing to two points and corroborated by the series' own trajectory (the 200-day was 6,286.45 at 31 December 2025, so ~7,205 is the arithmetically plausible value and 7,679.54 is not). And the gamma flip is ~7,700, not 7,680, on three separate Monday reads clustering 7,699 / 7,702 / 7,709. So the index closed Monday 6.6% above its 200-day, not sitting on it — there is no trend support anywhere near — while closing ~17 points BELOW the gamma flip, in negative-gamma territory. The mechanism half of the view is live; the confluence half was an artefact of a bad moving average. The index closed 7,683.26, three and a quarter points above the published 7,680 confirm trigger, so the trigger has not fired — and this desk is not moving it on the day its rationale collapsed. See §08.
- Monday was a real-rate shock, and it is why the regime tag changes today. Crude rose 3.25% and gold fell 3.90% in the same session — that combination is the market pricing Hormuz as an inflation-and-rates event rather than a haven event, and it is the cleanest single observation in the note. The mechanism is visible in the pricing: the probability of a 25bp hike on 28 October went from a settled 66.6% on Friday to 71.2%, and the December modal outcome moved from a coin flip at 51.6% to 58.9% for two cumulative hikes. The 10-year rose ~7bp to 5.24%, a fresh cycle high, and the 2-year has now risen roughly 56bp across September — the largest monthly move since February 2023. Equities fell with it, the dollar rose 0.22%, and the Cleveland Fed's core-PCE nowcast for September at 3.49% y/y against a 3.3% July actual says the disinflation the market needs is not arriving. The previous tag — energy shock plus a demand shock — described the inputs; this one describes the transmission channel that is actually pricing assets.
- Gold's break is a positioning event as much as a macro one, and the position was the most crowded in the entire futures report. Spot closed $4,117.70, down $167.30 or 3.90%, with a session low of $4,110.20 — not merely a fourth consecutive close below the $4,300 level where this desk stopped out of V014, but a decisive break of $4,200 and a seven-week low. Silver fell harder at −5.39% to $60.80 and platinum 3.25%. Managed-money gold was net long 225,853 contracts, 54.71% of open interest, the most crowded long in the COT report, and had been trimmed by only 4,485 contracts on the week. Silver, less crowded at 23.90% of open interest, nonetheless fell more — so this was a broad precious-metals liquidation on a real-rate impulse, not a gold-specific story. Gold is now roughly 12% below its end-August level. The V014 stop, written into the note two editions before it fired, has been right for five consecutive sessions and looks better every one of them.
- Crude rose on a rejection, the Brent contract-month trap fired again, and the high cluster was correct for the second consecutive edition. Front-month Brent (Nov-26) settled $107.71, +$3.39 or 3.25%, and WTI (Nov-26) $96.23, +4.13%, after Trump rejected Iran's seven-day Hormuz roadmap on Saturday — “I'm rejecting their deal… they overplayed their hand” — with Qatari mediators expected to meet both sides again on Monday and Iran saying it will not soften its demands. Two structural reads matter more than the level. The Nov/Dec spread is $8.34 of backwardation in a single month ($107.71 against $99.37), which is the market's statement that the disruption is acute but finite. And Brent–WTI narrowed to $11.48 from $11.91 as WTI outperformed. On method: Investing.com's Brent page tracks the second contract month while its WTI page tracks the front, which is what produced three irreconcilable Brent clusters — and the highest of them was the right one, so the “reject the high cluster” heuristic retired last edition stays retired.
- The ASX rose into a fully-priced hike led by the banks, which is against the house underweight — and Asian semiconductors broke in four separate tapes on one session. The ASX 200 closed 8,679.70, +14.70 or +0.17%, with Financials +1.16% and Health Care +1.49% carrying it against Materials −1.34%: banks were bought into the hike, which is what a market that has already discounted the decision looks like. That leaves V006's payoff entirely in the 15:30 press conference. Underneath, the most-crowded trade in the BofA survey — long global semiconductors at 53% — had a genuinely bad day across four venues at once: KOSPI −2.70% on large-cap tech profit-taking with foreign and institutional investors both net sellers; the Nikkei spiked through 67,000 for the first time and then gave back 1,157 points to close −0.73%; mainland semiconductor and optical-component makers led Shenzhen −3.44% to a nine-month low and CSI 300 −2.22%; and Hang Seng Tech fell 0.31% against a Hang Seng that closed UP 0.54%. Four markets, one direction, no single-market explanation — and Taiwan reopens today having priced none of it.
Overnight recap
Asia — the semiconductor trade broke in four places at once
The session's real signal was not any one index but the fact that the most-crowded trade in the BofA Fund Manager Survey (long global semiconductors, 53% of respondents) sold off in four independent tapes on the same day. Korea was the worst: the KOSPI fell 191.18 points to 6,889.74, −2.70%, back below 7,000 and ending a four-day winning streak, on large-cap technology profit-taking with foreign and institutional investors both net sellers; the open at 7,057.86 and the close both reconstruct exactly to the verified 7,080.92 prior, so the arithmetic is clean even though Monday's Korean single-stock prices are not obtainable. One analyst framed it as easing in both US–China and US–Iran relations being overwhelmed by “a sharp rise in U.S. Treasury yields”, which is the correct read and the theme of the whole session.
Japan produced the reversal day this desk was warned to expect. The Nikkei 225 spiked to an intraday 67,035 — through 67,000 for the first time — and then gave back 1,157 points, 1.73% off the high, to close 65,878, down 486.58 or 0.73%. Semiconductor and AI names drove the early surge and then faded: SoftBank Group, Advantest and Tokyo Electron could not hold the move, and Ibiden was the largest single drag, with Kioxia, Fast Retailing, Chugai and KDDI also negative. Banks and insurers were resilient on dividend-related buying into the 28 September ex-dividend date and on the view that higher rates support financial-sector earnings. The TOPIX fell 0.40% to 4,112.00, reconstructing to the verified 4,128.59 prior exactly — and the direction cross-check passes, which matters: last week a carried Nikkei close was wrong precisely because two vendors served intraday snapshots of a reversal day and the TOPIX contradicted them. Note that a widely-read Australian live blog carried the Nikkei at “−0.2%” on Monday, which is a mid-session snapshot roughly half a percentage point wrong.
Mainland China reopened after the 25–27 September closure and sold off hard: Shanghai Composite −1.67% to 3,823.6 (a near two-month low), CSI 300 −2.22% to 4,340.76, and Shenzhen Component −3.44% to 12,858.8, a more-than-nine-month low, led by semiconductor and optical-component manufacturers. Hong Kong went the other way at the index level and the same way underneath: the Hang Seng rose 0.54% to 24,643 while Hang Seng Tech fell 0.31% to 4,298.49. Anyone reading the headline Hang Seng as “Hong Kong shrugged off the mainland rout” has it backwards — Hong Kong tech participated in the rout and Hong Kong financials, property and energy masked it (Tencent +0.7%, AIA +0.4%, Meituan +0.4%, China Resources Land +2.9%). The A-versus-H split of 2.2 to 4.0 percentage points is the session's China story: offshore money bought the tariff relief from the weekend summit while onshore money sold the absence of stimulus and the absence of progress on AI and Taiwan. China industrial profits rose 15.7% y/y for the first eight months of 2026, described as moderating.
India was the most oil-sensitive market and traded like it: Sensex −1.52% to 72,771.72 and Nifty 50 −1.56% to 22,780.25, both reconstructing exactly to their verified priors, with Bank Nifty −1.99%, all ten sector indices lower, PSU Bank worst at −3.24%, breadth 1,214 advancing against 2,905 declining, and only three of the fifty Nifty constituents higher. FIIs sold a net ₹5,353 crore against DII buying of ₹5,189 crore; USD/INR 95.98. That is week eight of what was already a seven-week losing streak. Taiwan was closed for Teachers' Day, its second consecutive shut session after Friday 25 September — so the TAIEX's last print remains Thursday 24 September at 48,024.60 and it reopens today having priced none of two sessions' accumulated global semiconductor weakness. Indonesia's Jakarta Composite fell 1.45% to a one-month low on 594 decliners against 170 advancers, with GoTo down 14% after the exchange lowered the minimum price cap — a microstructure event that probably explains a chunk of the index move and all of the extreme small-cap dispersion.
Australia and New Zealand — bought into the hike
The ASX 200 closed 8,679.70, up 14.70 points or 0.17%, reconstructing to Friday's verified 8,665.00 exactly. The leadership is the point: Health Care +1.49%, Utilities +1.17% and Financials +1.16% against Materials −1.34% and Information Technology −0.74%. Defensives and banks carried a market on the eve of a rate rise, which is what a fully-discounted decision looks like — and it is uncomfortable for a tactical underweight. Northern Star Resources +6.2% to $23.47 was the session's biggest story, having rejected a $38bn approach from Gold Fields valuing it at $27 a share; that bid premium is most of the reason Materials fell only 1.34% on a day precious metals collapsed. Channel Infrastructure NZ +10.3% and Group 6 Metals +7.1% led; Karoon Energy fell 12.9% to $1.555 after cutting production guidance on a pump failure, and Minerals 260 and Brazilian Rare Earths fell about 10% each. Ingenia Communities rose 5.8% on an improved $2.14bn Warburg Pincus bid at $5.25 a share. Monday's volume, breadth and the A-VIX could not be obtained on any of four routes — with the RBA today, the A-VIX is the single number this desk most wanted and did not get; Friday's 11.53 remains the last verified reading.
The ACGB curve bear-flattened at the front and bear-steepened at the back: 2-year +4bp to 5.05%, 3-year +4bp to 5.04%, 10-year +2bp to 5.41% (independently corroborated at 5.41% by a second source), 20-year +6.2bp to 5.75% and 30-year +6.7bp to 5.82%. 3s10s flattened 2bp to 37bp. That shape — adding near-term tightening while pushing term premium further out the curve — is a market positioning for a hike plus a credible signal of more, not for a terminal announcement. For the first time in several editions Trading Economics' own change column happened to agree with the recomputed values, but only within about half a basis point at the 2-year and 10-year, which is the seventh consecutive edition in which it has not matched exactly. New Zealand's NZX 50 rose 0.14% to 13,830.68, ties to the cent — but note it was up 0.4% at 3pm with breadth dead even at 41 gainers to 41 decliners and gave up roughly two-thirds of the gain in the last 105 minutes, so anyone carrying the 3pm figure overstated the session by 26bp.
Europe — genuinely mixed, four up and three down
There is no single directional characterisation available for Monday in Europe and this note will not manufacture one. Four indices rose — DAX +0.41% to 25,513.30, CAC 40 +0.50% to 8,118.50, FTSE 100 +0.50% to 10,749, Euro Stoxx 50 +0.10% to 6,308.85 — and three fell: IBEX −0.51% to 19,600.30, FTSE MIB −0.16% to 51,782, SMI −0.02% to 13,942.91. All seven reconstruct to their verified Friday priors. Energy led and technology lagged: ASML, BE Semiconductor and Infineon each fell more than 1% after OpenAI paused training and tool-enabled inference on its most capable models following a containment incident — which is the European limb of the same semiconductor break that hit Asia. The SMI is worth a line of its own: it opened +0.64% at 14,034.74, traded a 13,937–14,083 range and faded all day to close essentially unchanged, with Lonza +2.64%, Sika +1.37% and Novartis +1.34% against Roche −1.36% and Logitech −1.36%. UK homebuilders surged on expectations of a revival of Help to Buy. The Stoxx 600 could not be obtained and is reported as not obtained rather than derived from a rounded intraday percentage — see §13, where two new vendor traps are recorded.
European rates sold off with the US but the curves tell different stories. Bunds twist-steepened — 2-year −1.3bp to 3.27% while the 10-year rose 0.8bp to 3.6294% and the 30-year 3.5bp to 3.96% — which is the ECB tone-meter move showing up in the curve, because the Governing Council has spent a week talking October down. Gilts bear-flattened hard from the front: 2-year +8bp to 4.77%, the biggest G4 front-end move of the session, 5-year +6bp, 10-year +4.4bp to 5.4034%, 30-year +4bp to 5.89%. That is Ramsden repricing November, and it is the most coherent rates-to-news link of the day: a dove who had previously penciled in at least two cuts told the Macro Money & Finance Society on Monday that holding since March has “effectively delivered a tightening relative to where I thought we might otherwise have been”, described himself as carrying a “conditional hawkish bias”, and said that “were upside pressures on the inflation outlook to continue to build, there could be a case for increasing Bank Rate”, alongside a multi-year plan to move quantitative tightening “into the background” at £20bn of annual gilt sales through to end-2034. OATs rose 4.5bp to 4.77% against Bunds' 0.8bp, which implies the OAT–Bund spread is testing its 52-week wide — but the dedicated same-page spread series has no Monday observation, so this note publishes no OAT–Bund mark for 28 September rather than mixing legs from two pages, which is the documented artefact that once produced 110.5bp against a true 97bp.
The United States — a bear-flattening yield shock, and the equity damage was underneath
There was no scheduled US data on Monday. The session was driven entirely by Iran and by rates. Fed pricing did the work: the probability of a 25bp hike on 28 October moved from a settled 66.6% on Friday to 71.2%, and the December modal outcome went from 51.6% to 58.9% for two cumulative hikes — a large one-day repricing, on a 16:45 ET post-close read. ⚠ Note the path within the session: a 15:25 ET snapshot had it at 74.6%, so pricing eased 3.4 points into the close — which is exactly why this desk does not quote the mid-afternoon number. Treasuries bear-flattened: the 10-year rose about 7bp to 5.24%, corroborated on an independent source whose implied prior of 5.17% matches the official par curve for 25 September exactly, with an intraday print reported as high as 5.26%; the 30-year is genuinely unresolved between 5.53% and 5.58%; the 2-year sits around 4.91–4.93%. The 2-year has risen roughly 56bp across September, the largest monthly increase since February 2023. Derived from the clearer cluster, 5s30s flattened ~3.5bp to roughly 47–48bp and 2s10s flattened ~5bp to about 31bp — a policy repricing flattening the curve, which is precisely the mechanism that stopped out this desk's 5s30s steepener last week and the reason the re-entry was declined.
Equities fell on that: the S&P 500 closed 7,683.26, −0.77%, which sits within a whisker of the 7,685.8 implied by SPY's settled −0.74% and is the level this note uses; the Dow fell about 0.66% to roughly 51,484; the Nasdaq Composite is unresolved between about 26,763 and 26,783 and is not published as a verified figure — the 26,820 print in circulation has been identified as a 12:16 ET midday snapshot and discarded, and QQQ cannot substitute because it tracks the Nasdaq-100. The Russell 2000 was not obtained at all: the only available read was an IWM capture stamped 3:49 PM EDT, eleven minutes before the bell, and the alternative tracker was still serving Friday's close. Eight of eleven sectors were red, energy and consumer defensives led, and consumer cyclicals and communication services lagged. Single names: Nvidia closed +1.69% at $228.88 on a new $150bn buyback authorisation taking its remaining total to $235bn through FY2028 — note it had been +2.64% at 3:01 ET and faded $2.13 into the bell, which is a useful calibration of how wrong intraday single-stock prints can be; MongoDB fell 16.7% to 20% on its CEO departing to Meta; Roblox −8.25%, Concentrix −7.83%, Boeing −6.2% to 2026 lows; Intel and AMD lower on AI-safety concerns. Breadth was far worse than the index: 422 new 52-week lows against 16 new highs on the NYSE, and 461 against 50 on the Nasdaq, with decliners beating advancers by more than two to one. That is the second consecutive session in which new lows have been wildly skewed against the index move — Friday's session rallied on 308 lows against 46 highs.
Weekend and Monday geopolitics. Trump rejected Iran's seven-day Hormuz roadmap on Saturday 26 September in explicit terms — “I reject their proposal”, Iran wants a deal because it is “losing so badly”, and on resuming strikes, “I am always thinking about it.” The rejected package was substantive: lifting of the US naval blockade, sanctions waivers on Iranian oil, release of $12bn in frozen assets and regional ceasefire observance, in return for reopening the strait and entering final negotiations. Qatar is actively mediating, having met separately with Araghchi and with Witkoff and Kushner around the UN General Assembly, and was expected to meet both sides again as early as Monday on proposed textual changes. Araghchi says Tehran awaits a formal response through mediators — so Iran has not treated the rejection as final — while Pezeshkian said “we have no trust in the American side.” Iran said it would not soften its demands. US officials describe ongoing indirect negotiations on nuclear concessions and sanctions relief but put the chance of agreement as “extremely slim.” Separately and importantly, the Trump–Xi summit did take place on 24–25 September and a tariff deal was announced on Saturday 26 September, cutting tariffs on roughly $30bn of “non-sensitive goods” each way — China lowering on agricultural goods, seafood, wood, cosmetics and medical devices and committing to import 10 million metric tons of US coal in each of 2027 and 2028; the US lowering on small appliances, toys, holiday decorations and children's car seats — with rare earths explicitly unresolved. That is small against total bilateral trade and should not be used to re-rate anything; the unresolved rare-earths limb is the substantive part.
Market dashboard
Monday 28 September — cross-asset change
| Equities | Close | 1d | Note |
|---|---|---|---|
| S&P 500 | 7,683.26 | −0.77% | Corroborated, not verified — AP's tabulation had not posted. Ties to the prior on the percentage (7,683.26 / 7,743.41 = −0.7768%) though the vendor's own point field lagged by 0.43. SPY's settled −0.74% implies 7,685.8; both are published. 3.26 points above V028's 7,680 trigger |
| Nasdaq Composite | ≈26,763–26,783 | ≈−1.1% | ⚠ NOT PUBLISHED AS A FIGURE. Three candidates are all self-consistent against the prior, so arithmetic cannot separate them. The 26,820 / −0.92% read in circulation is a 12:16 ET midday snapshot and is discarded; the index sold off into the bell, so a close below 26,822 is the coherent reading. QQQ cannot substitute — it tracks the Nasdaq-100 |
| Dow Jones | ≈51,484 | −0.66% | Derived range 51,481–51,485. The vendor page contradicts itself (header 51,498 / −0.64%, prose 51,484 / −344 / −0.66%); the prose reconciles to the prior within 3 points, the header does not |
| Russell 2000 | ≈2,818.5 | ≈−0.67% | ⚠ NOT OBTAINED. Derived from an IWM capture stamped 3:49 PM EDT — eleven minutes before the close, on a day the tape sold into the bell. The alternative tracker (VTWO) was still serving Friday's close under a current header |
| VIX / VIX3M | ≈16.3 / n/a | ≈+8% | ⚠ Two intraday reads (15.88 / +6.80%, 16.29 / +9.55%), agreed high 16.43; both reconcile to the verified 14.87 base so the move is corroborated and the level is not. VIX3M unpublished, so no Monday IVTS and no contango day-count — Friday's IVTS 0.8293, day 119 |
| Stoxx 600 | not obtained | — | ⚠ Fourth failure across the desks. The quote page serves Friday's 638.65 under a live “Real-time” header; a post-close vendor prints 638.68 with an uncomputed +0.00% change field. A rounded “+0.2%” from an intraday wrap is not a close and is not published |
| Euro Stoxx 50 · DAX | 6,308.85 · 25,513.30 | +0.10% · +0.41% | Both reconstruct exactly. Semis the drag: ASML, BE Semiconductor and Infineon each −1%+ |
| CAC 40 · FTSE 100 | 8,118.50 · 10,749 | +0.50% · +0.50% | Both tie. ⚠ FTSE 100 is corroborated not verified — single source, and a sibling CFD page failed the same reconciliation test on the same day. UK homebuilders bid on a possible Help to Buy revival |
| IBEX · SMI · FTSE MIB | 19,600.30 · 13,942.91 · 51,782 | −0.51% · −0.02% · −0.16% | ⭐ THE SMI GAP IS RESOLVED BY THE RECONCILIATION TEST ITSELF, a first. Two sources gave two Mondays; the CFD page's −6.41 implies a prior 4.00 points wrong, the cash page's −2.80 ties exactly and states its own prior. SMI opened +0.64% and faded a ~140pt round trip |
| Nikkei 225 | 65,878 | −0.73% | ⚠ Corroborated, NOT exchange-verified — the exchange archive has not posted 28 Sep (its last row is 25 Sep at 66,364.20, matching the prior exactly). A textbook reversal: intraday 67,035, through 67,000 for the first time, then −1,157 points off the high. An Australian live blog carried “−0.2%” — a mid-session snapshot, ~0.5pp wrong |
| TOPIX | 4,112.00 | −0.40% | Ties exactly. ⭐ The direction cross-check PASSES — Nikkei −0.73% against TOPIX −0.40% is internally coherent for a price-weighted, semis-heavy index fading against bid financials. Growth 250 Index 791.47 |
| Hang Seng · Hang Seng Tech | 24,643 · 4,298.49 | +0.54% · −0.31% | ⭐ THE HS TECH GAP IS CLOSED and it is the analytically important number: an 85bp divergence. ⚠⚠ The vendor's QUOTE BOX fails the chain (24,672 − 145.75 = 24,526 against a verified 24,510.09) while its NARRATIVE ties (24,643 − 132.91 = 24,511). Two further sources disagree with both, and one is internally corrupt — it reports the Hang Seng as “3,145.39” |
| CSI 300 · Shanghai · Shenzhen | 4,340.76 · 3,823.6 · 12,858.8 | −2.22% · −1.67% · −3.44% | Mainland reopened after 25–27 Sep and sold off. CSI 300 and Shanghai tie exactly; ⚠ Shenzhen's chain is unverified (no verified prior held). Shenzhen a nine-month low, Shanghai a near two-month low, led by semis and optical components |
| KOSPI · TAIEX | 6,889.74 · 48,024.60 | −2.70% · closed | ⭐ KOSPI's open AND close both reconstruct to the verified prior independently — high confidence. ⚠ Taiwan was SHUT Monday (Teachers' Day, second consecutive closed session); TAIEX is Thursday 24 Sep's close and reopens today with catch-up risk |
| Sensex · Nifty 50 | 72,771.72 · 22,780.25 | −1.52% · −1.56% | Both tie exactly. Bank Nifty −1.99%; all ten sectors lower, only 3 of 50 Nifty names up, breadth 1,214/2,905. FII −₹5,353cr vs DII +₹5,189cr. Week eight after a seven-week losing streak |
| S&P/ASX 200 | 8,679.70 | +0.17% | Ties exactly (+14.70). Financials +1.16% and Health Care +1.49% carried it INTO the hike; Materials −1.34%. NST +6.2% on a rejected $38bn Gold Fields approach at $27/share. ⚠ Volume, breadth and the A-VIX NOT OBTAINED on four routes — the A-VIX is the top gap on RBA day |
| NZX 50 · Jakarta | 13,830.68 · ≈6,151 | +0.14% · −1.45% | NZX ties to the cent — but it was +0.4% at 3pm and gave back two-thirds in 105 minutes. ⚠ Jakarta's level is derived and unverified (no verified prior); the percentage is sourced. GoTo −14% on an exchange price-cap change |
| Rates & credit | Level | 1d | Note |
|---|---|---|---|
| US Treasuries — ⚠ the official par curve has NOT posted for 28 September; its last row is 25 Sep. Every US yield below is a vendor read, labelled accordingly, and the primary is tomorrow's first verification target | |||
| UST 2y | ≈4.91–4.93% | ≈+10bp vs official | ⚠ Unresolved between clusters. The lower read implies a 25 Sep prior of 4.864 against an official 4.81 — a 5.4bp basis too wide to be pure par-versus-on-the-run methodology, so the 2-year is the least trustworthy point on the curve. +56bp across September, the largest monthly rise since Feb 2023 |
| UST 5y | ≈5.05% | ≈+7bp | Single cluster only. The 5-year has never closed above 5% on the official series — if it holds here that changes tomorrow |
| UST 10y | 5.24% | +7bp | ⭐ CORROBORATED and the cluster conflict is settled. An independent source prints 5.24% for 28 Sep with a +0.07 change implying a 5.17% prior — matching the official par 10y for 25 Sep exactly. The competing 5.223 read is an 08:50 ET morning snapshot (a quote page showed 5.22 at 07:50 CDT). A fresh cycle high; an intraday 5.26% is reported single-source |
| UST 30y | ≈5.53–5.58% | ≈+4 to +9bp | ⚠⚠ GENUINELY UNRESOLVED — published as a range. Three sources give 5.525, 5.55 and 5.576 and the middle one matches neither of the others. A “highest since mid-May 2004” claim attaches to the top of the range and is not corroborated |
| 5s30s · 2s10s | ≈47–48bp · ≈31bp | ≈−3.5bp · ≈−5bp | Derived from the internally consistent cluster. A policy-driven bear flattener — which is exactly the mechanism that stopped V003 out, and the reason the re-entry was declined. Official 25 Sep: 5s30s 51bp, 2s10s 36bp |
| JGB 2y · 10y · 30y | 1.97 · 3.09 · 4.17% | +3.0 · +2.2 · +1.0bp | ⭐ All three reconcile to their verified priors. The 2-year is the instrument: at 1.97% it embeds ~72bp over a 1.25% policy rate, up from ~69bp — close to three full hikes, far more than the ~11.5% October probability implies. 2s10s 112bp from 113bp — front-led bear flattening for a second session |
| Bund 2y · 10y · 30y | 3.27 · 3.6294 · 3.96% | −1.3 · +0.8 · +3.5bp | 10y ties exactly and remains the highest since June 2009. ⚠ 2y/30y have no verified prior. A twist-steepener — the front richened while the long end cheapened, which is the ECB dovish tone-shift in curve form. ~100bp of ECB hikes priced by late 2027 |
| OAT 10y · BTP 10y · BTP–Bund | 4.77 · 4.57% · 94.1bp | +4.5 · +2.2 · +1.4bp | Both legs tie to their priors. Italy still trades ~12.9bp inside France on these marks |
| OAT–Bund | no 28 Sep print | — | ⚠⚠ NOT PUBLISHED. The dedicated same-page series' last row is 25 Sep at 105.4bp (which ties to the carried mark exactly, so the series is trustworthy and merely late). OAT +4.5bp against Bund +0.8bp implies ~109bp, testing the 24 Sep 52-week wide of 109.9 — but that is a cross-page inference and mixing legs is the documented artefact that once gave 110.5bp against a true 97bp |
| Gilt 2y · 5y · 10y · 30y | 4.77 · 4.97 · 5.4034 · 5.89% | +8.0 · +6.0 · +4.4 · +4.0bp | ⭐ THE BIGGEST G4 FRONT-END MOVE OF THE SESSION and it has a named cause: Ramsden. 2s10s narrowed 3.6bp to 63.3bp. ⚠ The 10y's stated prior (5.35%) does not tie to the carried 5.3593% — a 0.7bp discrepancy, noted not resolved. 30y described as the highest since 1998 |
| ACGB 2y · 3y · 10y | 5.05 · 5.04 · 5.41% | +4.0 · +4.0 · +2.0bp | All three tie; the 10y independently corroborated at 5.41%. Embedded premium over 4.35% cash: 2y +70bp, 3y +69bp, 10y +106bp. Over a post-hike 4.60%: 3y still +44bp — roughly another 1.75 hikes. 20y 5.75% (+6.2bp), 30y 5.82% (+6.7bp) |
| ACGB 3s10s | 37bp | −2bp | V004 is 6bp in the money against a 43bp entry, having been 4bp on Friday. A pre-hike bear-flattening front end — the view working for the right reason for the first time in several editions. 14:30 is the test |
| Swiss 10y · Canada 10y | 0.665 · 3.965% | +1.6 · +4.0bp | Both unverified — no verified prior held for either |
| US IG · HY · CCC OAS | 79 · 280 · 1,112bp | no new print | ⚠⚠ ALL THREE SERIES STILL LAST-OBSERVED 24 SEPTEMBER. Two trading sessions (25 and 28 Sep) are missing, and they bracket the Hormuz rejection and a 4% crude move — which is precisely when a CCC gap would occur. The staleness is itself the finding. Path: 1,083 / 1,077 / 1,075 / 1,093 / 1,112. HY independently corroborated at 280bp for 24 Sep. V017 is 38bp from the level that confirms it and 62bp inside the level that closes it |
| FX — ⚠ see the caption: the mandated protocol FAILED and no twelve closes are published | Indicative Mon | vs verified Fri | Status |
|---|---|---|---|
| DXY | 101.1973 | +0.2263 (+0.22%) | ⭐ THE ONLY FX FIGURE PUBLISHED AS A CLOSE. Implied prior 100.971 against a verified 100.97 — exact tie, stamped 28 September. MTD +1.78%, YTD +3.36%. ⚠ It does not reconcile to its own legs — see below |
| EUR/USD · GBP/USD | 1.1373 · 1.3263 | −18 pips · +10 pips | Unverified. Two vendor sets differ by 15 and 17 pips; the GBP per-pair page shows current = previous close with a −0.0002 change, internally incoherent and rejected |
| USD/JPY | 157.23 | −5 pips | Unverified, single source. ~2.8 yen below the 160 policy line, which is closer than 72bp of embedded JGB tightening would normally allow |
| AUD/USD · NZD/USD | 0.7026 · 0.5674 | +2 pips · +16 pips | Unverified. ⚠ The AUD per-pair page is stamped 27 September — a Sunday — and was rejected; the NZD leg is disputed by 107 pips between two pages of the same vendor |
| USD/CAD · USD/CHF | 1.4173 · 0.8318 | +29 pips · +33 pips | Unverified. ⚠ The CHF page prints two different levels on one page (0.83147 and 0.8313) |
| USD/CNY · USD/KRW | 6.7133 · ≈1,355.5 | −71 pips · +1.3 | Unverified. ⚠ CNY sets differ by 123 pips; the KRW row is Sunday-stamped and rejected. PBoC fix 6.7399 against a 6.7085 model estimate — ~314 pips weaker-yuan than model and ~266 pips above spot, the third consecutive session in the same direction |
| USD/MXN | 17.9364 | +0.2520 (+1.43%) | The session's standout FX move and the one least trustworthy on magnitude. The vendor's own +2.69% is computed off a 17.47 base that does not tie to the verified 17.6844, so the +1.43% is this desk's derivation. Direction corroborated, magnitude unverified |
| AUD/NZD | 1.2384 | −30.6 pips (−0.25%) | Cross-computed from two legs of the same table, per protocol — and the method validates on Friday (0.7024 / 0.5658 = 1.24143 against a verified 1.2414, exact). ⚠ But the NZD leg is disputed by 107 pips, so the last ~20 pips are not reliable. Substituting the other page would give 1.2407; mixing pages is forbidden, so it is not reported. V023 ≈+0.56% from entry, unverified |
⚠⚠ The FX method failed for Monday and this note is saying so rather than printing twelve plausible numbers. Step 1 passed: Monday's Asia open matched Friday's verified closes within 8 pips on all seven pairs published. Step 2 failed: two mutually contradictory vendor snapshot sets, three pairs on Sunday-stamped or internally incoherent rows, and no dated historical row for 28 September existed for the AUD, NZD or JPY legs. Step 3 failed: weight-summing the legs gives a dollar move of +0.110% against the DXY's own reported +0.224% — a residual that would require USD/SEK +2.7% to close, and the alternative leg-set is worse at +0.043%. Conclusion: the dollar's direction on Monday is verified and its magnitude is not corroborated by any constituent set available. The likeliest explanation is that the DXY print is a later snapshot than the pair quotes.
| Commodities & digital assets | Last | 1d | Note |
|---|---|---|---|
| Brent (Nov-26) | $107.71 | +3.25% | ⭐ Implied prior $104.32 — exact tie. ⚠⚠ THE CONTRACT-MONTH TRAP: the vendor's Brent page tracks Dec-26 ($99.37) while its WTI page tracks the front month, which produced three clusters ($99.3–99.7 = Dec-26; $105.5–105.7 = spot/CFD; $107.71 = Nov-26 front, correct). The high cluster was right for the second consecutive edition. Nov/Dec backwardation $8.34 in one month — acute but finite |
| WTI (Nov-26) · Brent–WTI | $96.23 · $11.48 | +4.13% | Implied prior $92.41, exact tie; contract confirmed CL1! November 2026. The spread NARROWED $0.43 from $11.91 — WTI outperformed, consistent with a Hormuz premium partly transmitting to the US grade via freight. Curve: Nov 107.71 / Dec 99.37 / Jan-27 95.49 / Mar-27 91.08 |
| Henry Hub · TTF | $3.1144 · €71.32 | −2.55% · −0.79% | Both tie exactly. Gas FELL on a day crude rose 3–4%, despite Qatar extending its LNG force majeure — that decoupling argues the bid is crude-specific, not an energy-complex risk premium. ⚠ EU storage ~70.9% is UNVERIFIED — a near-identical figure set appears under a September 2024 dateline, so one source may be recycling the other's vintage |
| Gold (spot) | $4,117.70 | −3.90% | ⭐ VERIFIED on five routes, and the intraday path is coherent rather than contradictory. Timestamped 16:12 ET; range $4,110.20–4,279.90; the source's implied prior of $4,284.20 ties to the verified $4,285. Not merely a fourth close below the $4,300 V014 stop but a decisive break of $4,200 — a seven-week low, and ~12% below end-August. ⚠ The CAD cross-check did not post a Monday summary |
| Silver · Platinum · Gold/silver | $60.80 · $1,719.80 · 67.7 | −5.39% · −3.25% | Both tie exactly. Silver UNDERPERFORMED gold by ~150bp — which reverses Friday's “silver leads the physical metals higher” framing. Ratio widening. Note silver is only 23.9% of open interest against gold's 54.7% and still fell more |
| Copper LME cash · 3M | $14,740 · $14,647/t | no new settle | ⚠ The official 28 Sep settle posts today (one-day lag) — the widening-backwardation thesis cannot be updated until then. Cash−3M unchanged at +$93/t, and the 6.6× claim checks out arithmetically (+$14 on 18 Sep to +$93 on the 25th) while stocks fell 255,100 to 251,500t. Monday's move, derived from a $/lb quote, is roughly −1.7% |
| Iron ore | $96.92/t | −0.14% | Ties exactly. A FIFTEENTH consecutive sub-$100 print. China shut 1–7 October, so after tomorrow there is no fresh demand signal for a week |
| Lithium · Aluminium · Uranium | CNY 126,350/t · $3,237/t · $89.50/lb | −5.11% · −1.35% · — | Lithium was the largest single move in the complex — ties exactly, −20.28% on the month, still +71.79% y/y. Aluminium single-sourced, no prior to tie. Uranium has not updated since 25 Sep |
| Bitcoin · Ether · Solana | $83,539 · $2,683.86 · $119.01 | −1.61% · −0.72% · −2.82% | ⚠⚠ THE SIGN INVERSION RECURRED — the aggregator printed every one of five 24h changes POSITIVE while all five were negative. Recomputing from verified priors independently confirms the negatives. Levels from one source, signs and percentages from the other, exactly as the protocol requires — now validated twice. XRP $1.49 (−1.75%); BNB's sign is unverified and not published. Total cap $2.98tn, BTC dominance 56.3%, both single-sourced |
| Spot ETF flows | BTC +$134.5m · ETH +$87.0m (25 Sep) | no 28 Sep row | ⚠ The 25 Sep BTC row is INCOMPLETE — only three of twelve issuers carry values and the other nine print 0.0 placeholders, not zeros. The true figure will settle higher. Precedent with a measured magnitude: ETH's 23 Sep row, once carried as “incomplete at $2.5m”, settled at $104.5m — 42×. Weekly context: BTC ETFs took +$2.4bn in the week to 25 Sep, the largest weekly inflow in nearly a year, flipping 2026 net flows positive to +$934.1mn (single-sourced) |
| Crypto derivatives | OI: BTC $27.0bn · ETH $17.9bn | funding ≈flat | Funding +0.0025% / +0.0001% / +0.0048% (BTC/ETH/SOL) — positive but negligible. 24h liquidations $54.1m / $42.8m / $12.7m, $252.7m page-wide — a mild-decline tape with no forced deleveraging. OI figures are per-asset, not market-wide. No page timestamp |
Conventions: 1d = change on Monday 28 September's session. Yields in per cent, changes in basis points; commodities in USD per unit unless marked; gold and silver are spot. “≈” marks a derived or approximate value; “not obtained” means exactly that and is preferred to an estimate. Evidence states used throughout: verified = read from the primary issuer's own page; corroborated = two independent secondary sources agree; derived = computed by this desk from verified inputs; unverified = single source, disputed or stale. Vendor change columns are ignored on principle — levels are taken and changes recomputed against a verified prior, a rule that has now earned its keep for seven consecutive editions.
What is driving markets
1. The energy shock has stopped being an inflation story and started being a real-rate story — and that is why the regime tag changed
The single most informative observation on Monday was a correlation, not a level. Crude rose 3.25% and gold fell 3.90% in the same session. A haven-driven energy shock does not do that; a shock the market expects central banks to lean against does. The transmission is visible at every step. Hike pricing for 28 October moved from a settled 66.6% on Friday to 71.2%, and the December modal outcome from a coin flip at 51.6% to 58.9% for two cumulative 25bp increases. The 10-year rose ~7bp to 5.24%, a fresh cycle high, on a day with no US data at all. The 2-year has risen roughly 56bp across September, the largest monthly move since February 2023. The dollar gained 0.22%. And the thing that makes it a regime statement rather than a session: the Cleveland Fed's September core-PCE nowcast sits at 3.49% y/y against a 3.3% July actual, so the direction of travel on the Fed's own target variable is upward, and Wednesday's August print lands into that. Reading it as a demand shock — the previous tag's second leg — is still true of the household data (German consumer climate at −30.6, a near-two-year low on an income-expectations shock; US Michigan at 48.1 with one-year expectations at 4.6%), but demand weakness is not what priced assets on Monday. Real rates did.
2. The long end keeps rising while Treasury buys it — and this week the buying and the supply collide on the same day
This theme carries, and the calendar has sharpened it. Treasury has been running doubled long-end liquidity buybacks since 9 September — from $2bn to a $4bn minimum per operation in the 10–20y and 20–30y buckets, framed in some press as up to $6bn or “triple the normal level” — and the long end has gone up anyway: the 30-year printed a new cycle high on Friday and is somewhere between 5.53% and 5.58% now. The structural test is 8 October, and it is now a genuine collision: Treasury buys back at least $4bn of 20Y–30Y paper on the same day it auctions the 30-year reopening. The near-term sequencing matters too. The 2-year, 5-year and 7-year all settle tomorrow, 30 September — the same morning core PCE prints — which puts a front-end duration add into the market in the same minute as the inflation number, and there is then no coupon supply until 6 October. Last week's 7-year auction is the cautionary detail: the headline cleared in line but indirect bidders took 57.2% against a 64.6% average, and per this desk's own post-mortem on V003, strong cover is evidence about clearing, not about direction.
3. The crowded trades are unwinding one at a time, and the order is legible
Two of the three most crowded positions in the market took damage on Monday, and the data on both is unusually clean. Gold was net long 225,853 contracts, 54.71% of open interest — the most crowded long in the entire Commitments of Traders report — and it fell 3.90%, with silver at a less-crowded 23.90% of open interest falling harder at 5.39%. That is a liquidation on a real-rate impulse rather than a gold-specific story, and the crowding explains the violence rather than the direction. Separately, the most-crowded trade named in the September Fund Manager Survey — buying global semiconductors, 53% of a 190-respondent, $512bn panel — sold off in four independent tapes on one session: Korea −2.70%, a 1,157-point Nikkei reversal off a 67,000 print, Shenzhen −3.44% led by semis and optical components, and Hang Seng Tech −0.31% against a Hang Seng that rose. Europe's limb was ASML, BE Semiconductor and Infineon each down more than 1% after OpenAI paused training and tool-enabled inference on its most capable models. The third crowded position is the one that has not moved: leveraged funds are short 6,517,822 CBOT Treasury contracts across all six tenors as of 22 September, a figure this desk has now reproduced from the raw category columns, and net short 375,574 E-mini S&P contracts — the latter consistent with substantial cash-futures basis exposure rather than a directional bearish bet, and the distinction matters because the two unwind differently.
4. The RBA's problem today is not the hike, it is that the curve wants more than the Bank has signalled — and its last voting-member signal pointed the other way
Every input says 4.60% at 14:30. All four majors, all twenty-nine economists in one survey, and a tracker at 90.0% — though that tracker's site-wide stamp is 25 September and its own per-meeting components sum to 1.54 hikes against a headline claim of 1.0, so it is a soft Friday mark. The interesting disagreement is entirely about what comes after, and the bond market and the economists are on opposite sides of it. The 3-year ACGB at 5.04% sits 44bp above a post-hike 4.60% cash rate — call it another 1.75 hikes of embedded premium, pointing to a terminal nearer 4.85–5.00%. Money markets are reported as pricing at least two further increases with better-than-even odds on a third. Against that, most economists think today is the last one: ANZ is the only major officially forecasting a November rise, joined by HSBC and UBS, and AMP's Shane Oliver argues that by November there will be “more evidence of a cooling economy, falling home prices, a softer jobs market and rising recession risks, so we don't think a second hike let alone a third will be necessary.” The household evidence is on Oliver's side: eighteen lenders repriced fixed rates in September and all four majors moved up to 0.48pp — roughly double the Bank's own step, already delivered — Westpac consumer sentiment fell 5.2% to 84.40, NAB business conditions are at a six-year low, and the weekend's auction clearance printed a ten-week-low 50.3% on volumes down 22.4% week-on-week. And the last substantive signal from inside the Board was not hawkish: Iain Ross, a voting Monetary Policy Board member, concluded on 22 September that “there is no evidence of the emergence of a wage-price spiral in the present circumstances and recent data suggest such an outcome is unlikely”, with no RBA commentary of any kind in the six days since.
5. Credit's silence is now the loudest thing in this note
The IG, HY and CCC option-adjusted spread series are all still last-observed on 24 September, at 79bp, 280bp and 1,112bp. Two trading sessions — 25 and 28 September — are missing, and the gap is not neutral: those two sessions bracket Trump's rejection of the Hormuz roadmap, a 4% move in crude, a near five-point jump in October hike pricing and a 7bp rise in the 10-year. If a low-quality credit gap were going to occur, it would occur there. What the last available observation showed is the shape the house view requires and it was emphatic: over 23–24 September, IG widened 2bp, HY 12bp and CCC 37bp — CCC moved eighteen times as much as IG, which is credit risk repricing at the low-quality end rather than a rates pass-through. The daily path is accelerating: +2, +5, +7, +19bp. And the named live test resolved for the view rather than against it — SoftBank's $11.1bn BB+ deal, the largest high-yield corporate bond sale ever printed, priced through talk on 23–24 September, and over exactly those two pricing sessions the secondary index cheapened 12bp at HY and 37bp at CCC. The issuer got a good outcome and the secondary market paid for it. Note precisely what that claim is: the literal question — did the tranches hold reoffer — is unanswerable, because a sixth consecutive attempt to obtain secondary levels or a new-issue concession failed across five sources. The index move is a derived inference, not reported secondary colour, and should be read as such.
Central bank watch
Fed funds pricing — Monday's repricing, by meeting
| Bank | Policy rate | Last move / vote | Next decision (Sydney) | Market pricing | Bias |
|---|---|---|---|---|---|
| RBA | 4.35% | ⚠ “Hold since 11 August” is NOT confirmable from the issuer — the RBA's releases archive ends 18 May and its last listed decision is 5 May (+25bp to 4.35%, 8–1). The calendar is current; the archive is four months stale | TODAY, TUE 29 SEP 14:30 AEST · presser 15:30 · then 3 Nov with a full SMP | ⭐ +25bp to 4.60% effectively fully priced. Tracker 90.0% but stamped 25 Sep and internally inconsistent (components sum 1.54 hikes vs a 1.0 headline). All four majors and 29 of 29 economists call 4.60%. The 3y at 5.04% embeds +44bp over a post-hike 4.60% — ~1.75 more hikes | Hiking today; guidance is the trade |
| Fed | 3.75–4.00% | +25bp 16 Sep, 12–0 — the first hike since 2023. Chair Kevin Warsh. ⭐ Governor Cook's status is CLOSED: she is listed as a sitting Governor and a current FOMC voter on the Fed's own pages | Two-day meeting 27–28 Oct, decision Wed 28 Oct · 05:00 Thu 29th AEDT · no SEP | ⚠ POST-CLOSE READ, 16:45 ET — not the 17:00 settled refresh. Oct hike 71.2% from a settled 66.6% Friday and 59.7% a week ago. Dec modal 4.25–4.50% at 58.9% from 51.6%. ⚠ A 15:25 ET snapshot read 74.6% / 60.3%, so the strip eased into the close. Polymarket 66%; other venues 66–68%. ⚠ The page mislabels its own buckets — re-bucket every time | Hawkish, repricing fast |
| BoJ | 1.25% | +25bp 18 Sep, 7–2 — a 31-year high | Meeting 29–30 Oct, decision Fri 30 Oct — the Bank's own schedule, eleventh consecutive edition. Outlook Report the same day; Summary of Opinions Thu 1 Oct | ⚠ ~11.5% October — and that figure is attached to a WRONG DATE by its source, which lists the meeting as 28 Oct. The 2y at 1.97% embeds ~72bp over policy, close to three hikes, which is far more than 11.5% implies | Hiking; Dec the base case |
| ECB | DFR 2.50% | +25bp 10 Sep (Lane: a “no-brainer”) | Thu 29 Oct · 00:15 Fri 30th AEDT · then 17 Dec | ⚠ 38.6% October / 61.4% hold, and 93.3% for December — but stamped 25 Sep. ⭐ The carried ~48% could not be confirmed from any source and the 29% is confirmed stale (10 Sep). 38.6% is LESS hawkish than the figure this desk carried — a direction corroborated by the tone meter. ~99bp priced over nine meetings | Dec over Oct |
| BoE | 3.75% | Held 17 Sep, 6–3 (⚠ vote not verified from the Bank) | Thu 5 Nov · 23:00 AEDT · with an MPR | 85.7% November as of 25 Sep — higher than the ~81% carried. ⭐ Ramsden delivered Monday and the gilt 2y moved +8bp, the biggest G4 front-end move of the session. Taylor's NIESR Dow Lecture is tonight, squarely on inflation persistence; Mann speaks three times this week | Hawkish; a dove has turned |
| RBNZ | 2.75% | +25bp September 2026 | ⭐ DATE SETTLED FROM THE ISSUER after two editions: Wed 28 Oct, 14:00 NZ = 12:00 AEDT. It is a Monetary Policy REVIEW, not a full MPS; presser 15:00 NZ | ⚠ UNCONFIRMED — the tracker carries no RBNZ probability data at all. The carried ~31% for October is not reproducible | Tightening |
| BoC | 2.25% | Held 2 Sep — seventh consecutive | Wed 28 Oct · 09:45 ET = 00:45 Thu 29th AEDT | ⚠ Unconfirmed on the tracker. Two banks now forecast an October HIKE. Canadian GDP today | Hawkish hold |
| Norges · Riksbank | 4.50% · 1.75% | Norges +25bp 24 Sep with an MPR, all members supported; Riksbank held 24 Sep | Norges Thu 5 Nov 10:00 CET = 20:00 AEDT, issuer-confirmed; then 17 Dec with an MPR. ⚠ Riksbank 4 Nov still NOT confirmed — its calendar would not render on a third attempt | Norges: rate “remains close to the current level for a period ahead”, ~40% for another hike within six months. Riksbank: a 2026 hike is in the BASELINE | Hawkish hold · Hiking this year |
| SNB | 0.00% | Held 24 Sep, fifth consecutive | Thu 10 Dec · 09:30 CET = 19:30 AEDT | 39.9% hike / 60.1% hold (25 Sep) — a live tail. Inflation forecast raised across the whole horizon; Swiss 10y +28bp in a month | Hold, hawkish tilt |
| PBoC | LPR 3.00% / 3.50% | Unchanged 21 Sep, 16th month | ⚠ 19–20 Oct not confirmed from the PBoC | Monday's fix 6.7399 against a 6.7085 model estimate — ~314 pips weaker-yuan than model, ~266 pips above spot, a third consecutive session in the same direction. ⚠ The source's own headline (“tightens”) contradicts its numbers, and it is internally inconsistent on which session the fix applies to | Leaning against yuan strength |
| Banxico · Brazil | 6.50% · 13.75% | Banxico held 24 Sep unanimous 5–0; Brazil CUT 25bp on 16 Sep — a fifth straight reduction, real policy rate ~9.6% | ⚠ Both unconfirmed | Banxico is under real pressure: MXN was Monday's worst major at ~−1.43% after softening its language to say policy would not mechanically follow the Fed. Brazil is the G20 outlier, easing into a hiking G10 | Decoupling · Easing |
| India (RBI) | 5.25% | Held August, neutral stance retained | Wed 7 Oct | ⚠⚠ A HIKE TO 5.50% IS CONSENSUS, not merely priced — a poll has the MPC raising “as inflation broadens”, and commentary is now framing the risk that an October hike is too late given oil. India was Monday's most oil-sensitive major market. The first hike after a cutting cycle | Turning hawkish |
| Korea · Taiwan | 3.00% · 2.00% | Korea +25bp 27 Aug; Taiwan held 17 Sep | Korea Thu 22 Oct — issuer-confirmed (then 26 Nov). ⚠ Taiwan 17 Dec not confirmed | Unconfirmed. OECD raised Korea's 2026 growth forecast to 3.7% on the semiconductor cycle — against a KOSPI that fell 2.70% on semis | Tightening · Hold |
| South Africa · Indonesia · Czech · Philippines · Hungary · Turkey | 7.25% · 5.75% · 3.75% · 5.00% · 6.50% · 37.00% | SARB +25bp 23 Sep unanimous; BI held a third month; others held | ⚠ All six dates unconfirmed this edition; Czech and Hungarian policy rates also not re-retrieved and are carried | Indonesia stays the watch item — on hold three months while the IDX fell 3.09% and then another 1.45% on Monday, the region's worst | Mixed |
Fed detail — and one standing item closes today. The October repricing is the story: 71.2% for a hike against a settled 66.6% on Friday and 59.7% a week ago, with December's modal outcome now two cumulative hikes at 58.9%. ⚠ Quote the timestamp: this is a 16:45 ET post-close read, not the post-17:00 settled refresh. The page was stamped 15:25 ET at 74.6% when first checked and again at 17:10 ET, and then refreshed to 16:45 ET at 71.2% — so the strip eased 3.4 points into the close, and the mid-afternoon number that would otherwise have been published was the wrong one by more than three points. That is the protocol earning its keep rather than a data problem. There was no Fed speech between 25 and 28 September on any of three independent checks, but the recent record is uniformly hawkish: Hammack (Cleveland, a voter) on 24 September said inflation risks are tilted to the upside and that the longer above-target inflation persists the harder the return to target — paraphrased by the reporting source, so not quotable verbatim — with Paulson, Barr and Barkin all hawkish in the same week and Barr explicitly endorsing the September hike. This week is unusually heavy: Bowman, Barr, Goolsbee, Musalem, Williams and Waller all speak on Tuesday US time, with Waller and Williams the ones that matter. ⭐ And a standing daily check since edition 001 is retired: the Fed's own Board-members and FOMC pages list Lisa D. Cook as a sitting Governor and a current voting member. A search result suggesting a 28 September Cook speech was spurious — the URL 404s and the 2026 speeches listing carries no September Cook entry — but the membership pages settle the substantive question outright. Note also from those pages that Powell is listed as a Governor and Warsh as Chairman.
RBA detail. Timing is issuer-verified from a page last updated 23 September: statement 14:30 AEST, press conference 15:30, after a two-day 28–29 September meeting. Follow-ons are the Chart Pack tomorrow at 11:30, the Financial Stability Review on 1 October at 11:30, and the September minutes on 13 October. There is a live data-integrity problem worth knowing: the Bank's media-releases archive ends 18 May 2026 and its most recent listed monetary policy decision is 5 May, so the August hold and its vote split cannot be confirmed from the issuer at all, and today's 14:30 statement may not appear on the site promptly — plan for the presser and wire copy as primary. The guidance split is two-two: ANZ reads 4.60% as the first of two to a 4.85% terminal in November; CBA expects hawkish language and flags 4.85% if Q3 trimmed mean prints at or above 1.0% q/q; NAB and Westpac both read 4.60% as terminal, with Westpac expecting a split vote on differing views about trend productivity and labour-market slack where CBA expects unanimity. The 5 May statement went 8–1, so a dissent has precedent on this Board.
ECB detail — the tone has moved and this desk's carried figure was too hawkish. The weekly tone meter on 28 September read “Schnabel drives sharp drop as urgency about further hikes fades”, assessed as markedly dovish with policymakers showing little urgency to raise again while retaining a tightening bias. The individual record supports it: Lagarde said there are no signs yet of second-round effects from energy prices and that a “measured response” is appropriate because the shock is not yet embedded; Lane said future decisions may be “less straightforward” after September's “no-brainer” and pointed to December rather than October; Kocher said it is preferable to reach 2% without further hikes. That Schnabel — the hawk — is the one driving the dovish shift is what makes it a signal rather than noise. The Bund twist-steepener on Monday is the consistent trade. Carry 38.6% for October with its 25 September date attached, and note it is below the figure this desk previously published.
BoJ detail, and a correction to this desk's own copy. The decision date is 30 October from the Bank's own schedule — eleventh consecutive edition of resolving this against the issuer and nothing else, which continues to be worth doing because a widely-used tracker lists the meeting as 28 October, neither the 29th start nor the 30th decision, and its 11.5% October probability is therefore attached to a wrong date. The correction: the “hikes roughly once every three months, to 2% by around June 2027” signal came from Makoto Sakurai, a FORMER board member, not a serving official. Yesterday's edition and the ledger both called him an official and that materially overstates the source. His actual base case is December, with October live only on a large upward revision to the Bank's inflation forecasts — and since the Outlook Report lands the same day as the decision, 30 October is a single binary. The market's own read is in the 2-year at 1.97%, embedding ~72bp over policy. On the yen: no dated MoF or Finance Minister comment could be found for 25–28 September and none is carried. What is corroborated is that the 3 August intervention was a JOINT US–Japan operation, which is the material asymmetry — a repeat above 160 would not be Japan acting alone — and USD/JPY at ~157.2 sits about 2.8 yen below that line.
Regional briefs
United States
A rate shock with no data in it. There was no scheduled US release on Monday and the market repriced the Fed by nearly five percentage points anyway — October to 71.2%, December's modal outcome to two cumulative hikes at 58.9% — on Iran and on nothing else. That is a market whose reaction function has changed: the 10-year at 5.24% is a fresh cycle high, the 2-year has added ~56bp in a month, and equities and gold both lost to the same input. The week's data then arrives all at once and every print is asymmetric the wrong way. Conference Board confidence (consensus 89.2) and JOLTS (7.23m) today; core PCE tomorrow at 22:30 AEST with consensus +0.3% m/m against a +0.2% prior and personal spending +0.8% against +0.2% — a nominal-demand signal landing in the same minute as the 2-year, 5-year and 7-year settlements; ISM manufacturing Thursday; payrolls Friday, where the consensus is disputed between +90K and +98K and a strong print is the bearish outcome. The Cleveland Fed's September core-PCE nowcast at 3.49% y/y against a 3.3% July actual says the trend is going the wrong way. ⭐ One multi-edition gap closes: BLS has published the September CPI date — Wednesday 14 October, 08:30 ET / 22:30 AEST — which puts the last major inflation print of the cycle two weeks before the FOMC and three days before the blackout begins on 17 October. Politics: funding runs to 11 December with no October cliff, and the US–China tariff deal announced 26 September is real but small — ~$30bn each way, rare earths explicitly unresolved.
Euro area
The Council has talked October down and the curve has listened. October pricing is 38.6% against 93.3% for December, and the weekly tone meter moved markedly dovish with Schnabel — the hawk — named as the driver. Lagarde's “measured response” framing and Lane's “less straightforward” both point past October. The Bund twist-steepener on Monday (2y −1.3bp, 30y +3.5bp) is the consistent expression. The German signal of the week is a divergence: business climate at a three-year high of 89.9 against consumer climate at −30.6, a near-two-year low. Note a correction — −27.4 was the consensus and the prior was −26.8 revised, so the fall was 3.8 points not 3.2 — and the internals matter: a significant drop in income expectations with economic expectations virtually unchanged, so it is an income shock, not a growth scare. On the fiscal side the two bases must stay apart: ~€119bn of net new borrowing on the core federal budget and ~€200bn all-in including special funds are the same budget on two perimeters and must never be summed; readings are the week of 23–27 November with the Bundesrat on 18 December. The euro-area flash CPI is Friday 2 October at 19:00 AEST with consensus 3.7% headline against 3.2% prior — a +50bp jump that is the most consequential number on the forward calendar, and ⚠ the national flashes for Germany, France, Italy and Spain could not be dated at all across three routes.
United Kingdom
A dove turned, and it was the largest marginal shift available on the MPC. Ramsden had previously expected at least two cuts; on Monday he said holding since March has “effectively delivered a tightening relative to where I thought we might otherwise have been”, described a “conditional hawkish bias” watching energy, weather, AI supply-chain pressures, food and wage-setting, and allowed that “were upside pressures on the inflation outlook to continue to build, there could be a case for increasing Bank Rate.” Alongside it, a QT plan to reach end-2034 at £20bn of annual gilt sales, explicitly designed to move the programme “into the background.” The gilt 2-year rose 8bp, the biggest G4 front-end move of the session, and November pricing sits at 85.7% — above the ~81% this desk carried. Taylor's NIESR Dow Lecture tonight is on inflation persistence and is the next catalyst; Mann speaks three times this week (today, Thursday and 6 October), the heaviest load on the Committee and the best read on the hawkish wing; Bailey opens the LSE Future of Money conference Thursday; and ⭐ the Financial Policy Committee Record publishes tomorrow — not on this desk's carried list and a live financial-stability event with the 30-year near 5.89%, described as the highest since 1998. ⚠ Correction to carry: the Chancellor is John Healey under PM Andy Burnham since 20 July 2026. Budget 28 October; headroom estimates span £7bn to £17bn and this desk's £8–11bn working range sits at the conservative end, not the consensus.
Japan
A reversal day, and the front end doing the work for a second session. The Nikkei printed 67,000 for the first time and closed 0.73% lower, giving back 1,157 points; semiconductor and AI names led both legs. The 2-year JGB rose 3.0bp to 1.97%, more than the 10-year (+2.2bp) or the 30-year (+1.0bp), so the curve bear-flattened from the front again — 2s10s 112bp from 113bp. At 1.97% the 2-year embeds ~72bp over the 1.25% policy rate, close to three full hikes, which is a long way from the ~11.5% October probability circulating. The near calendar is dense and all of it lands Thursday 1 October, 08:50 JST: the Summary of Opinions for the 17–18 September meeting — the read on whether the two dissents were dovish or hawkish — plus the September Tankan, where the consensus dispute is now settled at +26 for large manufacturers (from +22), against 36 for non-manufacturers (from 37), on a twelve-institution panel confirmed by two independent outlets. ⚠ Note the tension: that +26 is built on an AI and semiconductor demand story that Monday's tape was actively repudiating across four markets. Tokyo CPI is Friday 09:30 AEST — Thursday in ET, which is a date this desk had on the wrong side of the dateline — and its 2.4% consensus against a 1.8% prior is a 0.6pp step far outside the series' 2026 range of 1.3–2.0%, so sanity-check it before trading it. Japanese single-stock prices remain unsourceable for a seventeenth consecutive edition, so no Japanese relative value is opened.
China & Hong Kong
The A/H split is the story and the closure is the structure. Mainland reopened and sold off — Shanghai −1.67%, CSI 300 −2.22%, Shenzhen −3.44% to a nine-month low — while Hong Kong rose 0.54% and Hong Kong Tech fell 0.31%. Offshore bought the weekend tariff relief; onshore sold the absence of stimulus and the absence of progress on AI and Taiwan. Read the HSI headline carefully: HK tech participated in the regional semiconductor break and HK financials, property and energy masked it. Industrial profits rose 15.7% y/y for the first eight months, described as moderating. The tradeable structure of the week is mechanical: the mainland has exactly two sessions left — today and tomorrow — before Golden Week shuts it from 1 to 7 October, and the official PMIs print tomorrow on the last one. Manufacturing consensus is 50.1 against 49.8, which would be the first expansion reading since June, with non-manufacturing at 49.2 against 49.0; ⚠ the release time is standardly 11:30 AEST but could not be confirmed from the NBS itself. Any repositioning has to clear in a single session or wait a week, which mechanically raises the odds of an outsized Wednesday move and pushes a week of gap risk onto the 8 October reopen. Hong Kong closes only 1 October and trades from the 2nd, so it becomes the sole venue for China price discovery through the break — expect displaced flow and do not read HSI or HS Tech as clean Hong Kong signals in that window. Property, credit and stimulus news: largely a gap this edition.
Emerging Asia & LatAm
India is where the oil shock is landing hardest and the central bank is now expected to respond. Sensex −1.52% and Nifty −1.56% on Monday, all ten sectors lower, only three of fifty Nifty names higher, FIIs selling ₹5,353 crore — week eight after a seven-week losing streak, and moving against the US tape on local drivers. The RBI meets 7 October and a hike to 5.50% is consensus rather than merely priced, with commentary now framing the risk that October is too late given oil. That would be the first increase after a cutting cycle. Korea fell 2.70% on semiconductor profit-taking even as the OECD raised its 2026 growth forecast to 3.7% citing the chip cycle — the survey and the tape disagreeing in the same week. Taiwan was shut Monday for Teachers' Day, its second consecutive closed session, and reopens today with two sessions of unpriced global semiconductor weakness; TSMC is more than 40% of market capitalisation. Indonesia is the region's worst and remains the watch item: −1.45% Monday to a one-month low on 594 decliners to 170 advancers, with Bank Indonesia on hold for a third month; a 14% fall in GoTo on an exchange price-cap change explains part of the index move and all of the small-cap dispersion. Mexico had the worst major currency move of the session at roughly −1.43%, days after Banxico held unanimously and softened its language to say policy would not mechanically follow the Fed — the peso is testing that proposition. Brazil cut 25bp on 16 September, a fifth straight reduction, and is the G20's easing outlier against a hiking G10.
Australia & New Zealand
The RBA: 4.60% is not the question
Statement 14:30 AEST, press conference 15:30, both issuer-verified, after a two-day 28–29 September meeting. A 25bp move to 4.60% would be the highest cash rate since November 2011. It is effectively fully discounted: all four majors, all twenty-nine economists in one survey, and a tracker at 90.0% whose site-wide stamp is 25 September and whose components sum to 1.54 hikes against a 1.0 headline — a soft Friday mark, not a Monday one. So the decision is the least informative part of the event and the guidance is the whole trade.
The disagreement about what follows is unusually sharp and the bond market is on the hawkish side of it. The 3-year at 5.04% sits 44bp above a post-hike 4.60% cash rate — roughly another 1.75 hikes — pointing to a terminal nearer 4.85–5.00%, and money markets are reported pricing at least two more with better-than-even odds on a third. Against that, most economists think today is the last one: ANZ alone among the majors officially forecasts November (joined by HSBC and UBS), and AMP's Shane Oliver argues that by November there will be “more evidence of a cooling economy, falling home prices, a softer jobs market and rising recession risks, so we don't think a second hike let alone a third will be necessary.”
Two things cut against a hawkish statement and both are under-weighted. First, the last substantive signal from inside the Board was dovish: Iain Ross, a voting Monetary Policy Board member, concluded on 22 September that “there is no evidence of the emergence of a wage-price spiral in the present circumstances and recent data suggest such an outcome is unlikely” — and there has been no RBA commentary of any kind in the six days since, on the Bank's own speeches list. Second, the household channel has already delivered roughly double the Bank's step without the Bank: eighteen lenders raised at least one fixed term in September and all four majors moved up to 0.48pp. Watch the statement for whether the Bank softens its upside-risk language or leans on “restrictive conditions” — historically the transition marker to a hold — and for any shift onto the employment side of the mandate, which is the November tell. Westpac expects a split vote where CBA expects unanimity; the 5 May decision went 8–1, so a dissent has precedent. ⚠ A data-integrity caveat that affects today: the Bank's releases archive ends 18 May, so the 14:30 statement may not post promptly — plan for the presser and wire copy.
Markets: bought into the hike, which is the uncomfortable part
Equities. The ASX 200 closed 8,679.70, +14.70 or +0.17%, and the leadership is what matters: Health Care +1.49%, Utilities +1.17%, Financials +1.16% against Materials −1.34% and IT −0.74%. Banks and defensives carried the market on the eve of a rate rise — a fully-discounted decision, and a tape that ran against the house underweight. Northern Star +6.2% to $23.47 on a rejected $38bn Gold Fields approach at $27 a share is most of the reason Materials fell only 1.34% on a day gold lost 3.9%; Channel Infrastructure NZ +10.3% and Group 6 Metals +7.1% led, while Karoon Energy fell 12.9% to $1.555 on cut production guidance after a pump failure (a second source has −12.3%; the 0.6pp gap is unreconciled), with Minerals 260 and Brazilian Rare Earths each about −10%. Ingenia Communities +5.8% on an improved $2.14bn bid at $5.25. ⚠ Volume, breadth and the A-VIX could not be obtained on four separate routes — on decision day the A-VIX is the number this desk most wanted; Friday's 11.53, falling into the event, is the last verified reading and says the local market is not paying for it.
Rates. 2-year +4bp to 5.05%, 3-year +4bp to 5.04%, 10-year +2bp to 5.41% (independently corroborated), 20-year +6.2bp to 5.75%, 30-year +6.7bp to 5.82%. 3s10s flattened 2bp to 37bp — V004 is now 6bp in the money against a 43bp entry, and for the first time in several editions it is working for the right reason. The shape is a bear-flattening front end with a bear-steepening long end: adding near-term tightening while pushing term premium out the curve, which is a market positioning for a hike plus more, not for a terminal announcement. Embedded premium over 4.35% cash: 2y +70bp, 3y +69bp, 10y +106bp.
The SPI, after eleven editions. ⭐ The December contract is quotable again — S&P/ASX 200 futures, front month December 2026 (APZ2026), last 8,727.5, +16.5, prior settlement 8,711 — which resolves the sourcing failure. But the quote is stamped 26/09, so it is not Monday's settle, and there is a new trap replacing the old one: the +62.5 point gap to cash is the December contract's carry and dividend basis over an index three months away, NOT a directional signal. Anyone comparing APZ2026 to spot will manufacture a 62-point bullish read that does not exist. A live blog noted futures up 3 points at Monday's open, which is consistent with the flat-to-firm session that transpired.
Currency. AUD/USD's Monday close is not established — the per-pair page is Sunday-stamped and was rejected, and no dated historical row for 28 September exists for the AUD leg. The indicative aggregate read is 0.7026, up ~2 pips. AUD/NZD is cross-computed from two legs of the same table at 1.2384, −30.6 pips, with the method validated on Friday to the fourth decimal — but the NZD leg is disputed by 107 pips between two pages of one vendor, so the last ~20 pips are unreliable.
Property and the household channel — the transmission the Bank is not doing itself
Weekend auctions cleared a preliminary 50.3% across the combined capitals on 1,428 auctions — a ten-week low — with volumes down 22.4% week-on-week and 17.7% year-on-year. Sydney 53.6% on 790 auctions (its busiest week since late May), Melbourne 48.8% on 286 — the weakest since September 2021 — Canberra 53.8%, Brisbane 42.6% (below 50% in eighteen of the past nineteen weeks), Adelaide 39.3%, down 18.9pp and only the second sub-40% print this year. ⭐ Unusually, the published national figure reconciles with its own city rows this week: Sydney and Melbourne account for 1,076 of 1,428 auctions at a weighted 52.3%, leaving an implied 44.1% across the remaining 352, which sits between Adelaide and Canberra and next to Brisbane. No arithmetic contradiction — a first in recent editions, and only Cotality rows were used as the basis given the known reconciliation failures at two other providers. But do not read 50.3% as a clean signal: Melbourne's volume collapsed ~70% on the AFL Grand Final long weekend, so Melbourne fell to ~20% of the national count against a normal ~40% while Sydney — the strongest city — carried 55% of it. The mix flatters the headline; on a normal distribution the national rate prints lower. (The same holiday cut Friday's ASX volume by 19%.) On values, the verified recent read is national dwelling values down 3.1% over the three months to August, with high-end homes leading the downturn; ⚠ the −0.9% m/m, five-consecutive-falls and −3.6%-from-March-peak figures were not re-verified this edition and are carried, and the next Cotality HVI release date could not be confirmed.
Mortgage repricing is the mechanism. Eighteen lenders raised at least one fixed term in September; all four majors moved up to 0.48pp; Macquarie went twice in three weeks (+0.30pp on the 8th, +0.20pp on the 24th) to 6.49% at one year and 6.59% at two and three, which now matches the cheapest offers from NAB and ANZ — the majors and the leading non-major have converged at the front of the fixed curve. Aussie +0.25pp, ubank +0.30pp. Lenders have delivered roughly 48bp of fixed-rate tightening ahead of a 25bp cash move, which is the mechanism behind a ~9% fall in borrowing capacity this year and the auction weakness above. Confidence confirms it: Westpac–Melbourne Institute consumer sentiment fell 5.2% to 84.40 on fuel prices and rate-hike fears, and NAB business confidence at −8 sits alongside business conditions at a six-year low — conditions usually lead confidence, which makes that the more alarming of the two. Both readings predate the September fuel compounding and today's hike.
New Zealand and the China link
New Zealand. The NZX 50 rose 0.14% to 13,830.68, ties to the cent, on consumer staples, utilities and real estate — but it was up 0.4% at 3pm with breadth dead even at 41–41 and surrendered two-thirds of the gain in the final 105 minutes. ⭐ The RBNZ date is settled from the issuer after two editions of carrying it unconfirmed: Wednesday 28 October, 14:00 NZ = 12:00 AEDT, and it is a Monetary Policy REVIEW rather than a full MPS, with a 15:00 NZ press conference. The OCR is 2.75% after a September increase. ⚠ NZD/USD and any RBNZ commentary could not be obtained this edition, and the tracker carries no RBNZ probability data at all, so the carried ~31% for October is not reproducible.
The China and commodity link. Iron ore at $96.92/t, a fifteenth consecutive sub-$100 print, in a four-session band. The demand signal goes dark after tomorrow: the official PMIs print Wednesday on the last mainland session before a seven-day closure, manufacturing consensus 50.1 against 49.8 — which would be the first expansion since June — and then nothing until 8 October. Copper's official 28 September LME settle posts today on the usual one-day lag, so the widening-backwardation thesis cannot be updated yet; cash−3M remains +$93/t, and the 6.6× widening over five sessions checks out arithmetically (+$14 on 18 September to +$93 on the 25th) against stocks falling 255,100 to 251,500 tonnes. Monday's derived copper move is roughly −1.7% on weakening China signals. Lithium was the complex's largest single move at −5.11% to CNY 126,350/t, down 20.28% on the month though still +71.79% y/y. Australia's own data: nothing from the ABS on Monday; today at 11:30 the Monthly Household Spending Indicator lands three hours before the decision — not decision-relevant, since the Board has already met, but it will shape the tape into 14:30.
| Australia — key data trail | Latest | Prior / context | Next release (AEST) |
|---|---|---|---|
| Cash rate | 4.35% | ⚠ Last issuer-confirmable decision 5 May (+25bp, 8–1) | TODAY 14:30 · presser 15:30; then 3 Nov 14:30 AEDT with a full SMP; 8 Dec |
| Monthly CPI indicator | 3.5% headline · 3.6% trimmed mean | 3.8% · 3.6% | Tomorrow, Wed 30 Sep 11:30 — time issuer-confirmed. ⚠ Consensus not obtained from a forecaster survey; a 4.1% y/y figure circulating on one calendar is single-sourced and a very large step from 3.5% |
| Unemployment (Aug) | 4.6% | ABS itself caveated the print for Labour Force Modernisation effects | September: late October. Bullock's own CEDA test — unemployment “between 4.5 and 5” would “probably take enough heat out of the labour market” — is satisfied |
| Dwelling values | −3.1% over the three months to August | High-end homes leading the downturn; ⚠ the −0.9% m/m and −3.6%-from-peak figures are carried, not re-verified | ⚠ Cotality HVI date NOT confirmed — “around 1 October” remains an expectation |
| Auction clearance | 50.3% preliminary, 1,428 auctions | A ten-week low; volumes −22.4% w/w, −17.7% y/y. ⭐ Reconciles to its own city rows | Next weekend. ⚠ Melbourne volume −70% on the AFL long weekend flatters the headline |
| Fixed-rate repricing | 18 lenders in September | All four majors up to 0.48pp — roughly double the Bank's step, already delivered | Sub-6% owner-occupier fixed rates “likely to become scarce” post-decision |
| Consumer sentiment · business confidence | 84.40 (−5.2% m/m) · −8 | 88.90 · −6; business CONDITIONS at a six-year low | Both predate the September fuel compounding and today's hike |
| ACGB 3y · 10y · 3s10s | 5.04% · 5.41% · 37bp | +4bp · +2bp · −2bp. 3y embeds +44bp over a post-hike 4.60% | 14:30 is the test. V004 6bp in the money from a 43bp entry |
| Iron ore · copper cash−3M | $96.92/t · +$93/t | 15th sub-$100 print · backwardation 6.6× wider in five sessions on falling stocks | China PMIs Wed 30 Sep, then the mainland shuts 1–7 October. LME copper's 28 Sep settle posts today |
House views & tactical framework
V028 was opened yesterday on a level that does not exist, and both legs of it are wrong. The view was built around “7,680, where the 200-day (7,679.54) and the dealer gamma flip (7,680) coincide” — the claim being that trend support and the damping mechanism would fail at one price, so moves would go from damped to amplified there.
The S&P 500's 200-day moving average is approximately 7,205, not 7,679.54 — an error of 474 points, or 6.6%. Two independent sources agree at 7,205.20 and 7,207.26, and three separate checks say they are right rather than the outlier: the 200-day's own year-end series (6,286.45 at 31 December 2025, 5,551.62 at 31 December 2024) makes ~7,205 the arithmetically plausible value for late September; the outlier source prints its MA100 below its MA200 while price sits 1.8% off the high, which is not a coherent ordering for an index in an uptrend; and a 200-day of 7,679.54 would require the index to have averaged within 1% of its record for two hundred consecutive sessions while still printing a 52-week low 18% lower. The index closed Monday 6.6% ABOVE its 200-day. There is no trend support anywhere near current levels and there will not be one.
And the gamma flip is ~7,700, not 7,680 — three independent Monday reads cluster at 7,699, 7,702 and 7,709, while the single source printing 7,680.77 is internally inconsistent (it places spot above its own flip while reporting net gamma exposure negative, which cannot both be true). So the flip migrated up ~20 points, exactly as the re-fetch discipline anticipated.
What survives and what this desk is doing about it. The mechanism half of the view is live and arguably stronger than published: net gamma exposure flipped negative on Monday (three sources agree on the sign; the magnitude is unreconcilable between −$2.2bn and −$6.0bn), and the index closed ~17 points BELOW the flip, in short-gamma territory, where dealer hedging amplifies rather than damps. The confluence half was an artefact. The index closed 7,683.26, three and a quarter points above the published 7,680 confirm trigger, so the trigger has NOT fired — and this desk is not moving it on the day its rationale collapsed. Conviction stays Low. If the trigger is to be restated around the gamma flip rather than the moving average, that restatement must be published in advance of a session, not after a move — which is the whole point of the pre-commitment convention, and the reason V003's stop was allowed to fire at a local extreme last week rather than being argued away.
| Asset | Bias | Conv. | Horizon | Rationale | What changes the view |
|---|---|---|---|---|---|
| Rates | |||||
| ACGB 3s10s (V004) | Flattener | Med | 1–2 mo | Working for the right reason for the first time in several editions. 37bp against a 43bp entry — 6bp in the money, up from 4bp on Friday, on a clean pre-hike bear-flattening front end (2y and 3y both +4bp, 10y +2bp). All three tenors reconcile to verified priors, so the mark is sound rather than contested for the first time in four editions. The 3y at 5.04% embeds +44bp over a post-hike 4.60% — the market is not marking this as near-terminal any more, which reverses last Friday's read | 14:30 today is the live test and the presser matters more than the statement. Guidance framing 4.60% as terminal has ~44bp of 3y premium to unwind, which steepens this and should close it. A dovish RBA with a sticky 10y, or a China stimulus impulse steepening the long end |
| OAT–Bund (V025) | Widener | Low | 1–3 mo | Still the best view in the book, and its catalyst date is now settled — into a collision this desk had not seen. ⚠ No 28 September mark: the dedicated same-page series' last row is 25 Sep at 105.4bp (which ties to the carried figure exactly, so the series is trustworthy and merely late). OAT +4.5bp against Bund +0.8bp implies ~109bp, testing the 24 Sep 52-week wide of 109.9 — published as an inference only, because mixing legs across pages is the documented artefact that once gave 110.5bp against a true 97bp. Entry ≈94bp, so ~11bp of widening with ~25bp to the stop. BTP–Bund 94.1bp: Italy still trades ~13bp inside France | A compression inside 80bp. ⭐ The catalyst date is RESOLVED, discharging the first verification action named yesterday: PLF 2027 goes to the Conseil des ministres THURSDAY 1 OCTOBER, on two sources including one published Monday morning; the 30 September variant traces to an undated page. And it lands the same day as the AFT's long-term OAT auction — a sequencing collision that does not exist under the 30 September hypothesis. €54bn sought against a 5.0% deficit target and 121.7% debt; Assembly deposit 6 October. ⚠ The censure arithmetic (LFI/Greens/PCF committed, PS leaning, the RN's 122 votes decisive at a price of no new taxes) is carried, not verified this edition |
| Equities | |||||
| S&P 500 (V028) | Own downside convexity, not delta | Low | 2–4 wk | Direction right, reason wrong — see the box above. Reference 7,743.41; Monday closed 7,683.26, −0.78% in the money in one session. What holds: net gamma flipped negative and the index closed ~17 points below a ~7,700 flip, so dealer hedging now amplifies moves; the corporate bid goes from 10% of index weight in blackout to 61% by 30 September with no reopening until 1 November; CTAs are asymmetric 21:1; vol-control sits at ~86% equity exposure. Breadth is worse than the index and getting worse — 422 NYSE new 52-week lows against 16 new highs, and 461 against 50 on the Nasdaq, the second consecutive session where lows are wildly skewed against the index move. What does not hold: the 200-day is ~7,205, not 7,679.54, and the index is 6.6% above it | UNCHANGED AS PUBLISHED, deliberately: two consecutive closes above 7,800 closes this view; a close below 7,680 confirms it and conviction rises to Med. It closed 3.26 points above that line and the line's rationale is void — and the trigger is still not being moved today. ⚠ Cost caveat stands and is now stale as well as adverse: SKEW is 144.9 and the page did NOT print a Monday read, so the carried level is two sessions old. Its 54th percentile against a trailing year is the misleading number; its 93rd percentile against all history is the real one — deep protection is expensive |
| ASX 200 (V006) | Underweight tactically | Med | 2–4 wk | Still well in the money at −3.62% from a 9,005.9 entry, but Monday went against it and the way it went against it is the point. The index rose 0.17% to 8,679.70 with Financials +1.16% and Health Care +1.49% carrying it: banks were bought into a hike that is fully priced. The structural case is intact — 18 lenders repriced fixed rates in September, all four majors up to 0.48pp, roughly double the Bank's step already delivered; values −3.1% over three months to August; auctions at a ten-week-low 50.3% on volumes −22.4% w/w, and flattered by a Melbourne volume collapse; consumer sentiment −5.2% to 84.40; business conditions at a six-year low. ⚠ Volume, breadth and the A-VIX all unobtainable Monday | An RBA statement framing 4.60% as terminal — which is the majority economist view and would hit hardest in exactly the financials and real estate this is short. Iron ore reclaiming $100; banks stabilising as a trend. ⚠ The concentration is unchanged and resolves in under eight hours: this, V004 and V023's short leg all settle on one press conference |
| FX | |||||
| AUD/NZD (V023) | Long | Low | 1–2 mo | ≈1.2384, ≈+0.56% from a 1.2315 entry — and the mark is explicitly unverified. Cross-computed from two legs of the same dated table per protocol, with the method validated on Friday to the fourth decimal (0.7024 / 0.5658 = 1.24143 against a verified 1.2414). ⚠ But the NZD leg is disputed by 107 pips between two pages of one vendor, so the last ~20 pips are not reliable, and substituting the other page would give 1.2407 — which mixing pages forbids. This is the third consecutive edition in which this view's mark has been contested, and all three times the problem was a LEG rather than the thesis | An RBA hold today; a hawkish RBNZ on 28 October (⭐ date now settled from the issuer — and it is a REVIEW, not a full MPS); a China shock hitting Australia harder. ⚠⚠ Do not mark this off spec positioning: AUD remains the single largest disagreement in the COT report — leveraged funds net LONG 58,726 against legacy non-commercials net SHORT 46,814, a 105,540-contract spread on open interest of 306,488 that matches on both pages, so the divergence is real classification, not error |
| Commodities | |||||
| Brent (V024) | Residual call spread only — no new risk above $100 | Low | 1–3 mo | $107.71, +7.1% above the $100.60 reference and well outside the re-own band. Front-month Nov-26 verified with an exact tie to the prior. ⚠⚠ The contract-month trap fired again and the high cluster was right for the second consecutive edition — the vendor's Brent page tracks Dec-26, which produced three clusters and would have understated the settle by $8. Two structural reads matter more than the level: Nov/Dec backwardation of $8.34 in a single month says the market reads the disruption as acute but finite, and Brent–WTI NARROWED to $11.48 from $11.91. Physical remains poor — observed Hormuz transits ~1/day on a seven-day AIS average against 6–14 across providers, VLCC rates above $1.2m/day, Qatar's LNG force majeure extended, and the East–West pipeline running at 100% of its 7.0 mbd capacity with zero redundancy | Unchanged as a CONDITION: two consecutive settles in the $92–95 band re-owns this outright. Nowhere near. Also a confirmed physical restart at scale, or a Hormuz reopening that Oman confirms rather than Tehran asserts. ⚠ Corrections to carry: the “$1.27m/day record” VLCC rate cannot be tied to a named assessor — the verifiable statement is “above $1.2m/day, Gibson Shipbrokers, 21 September”; and a US official's claim of 40 protected ships a day is the fourth consecutive government volume claim to fail against vessel tracking |
| Credit & digital assets | |||||
| US credit (V017) | UW HY/CCC; prefer 3–5y IG | Med | 1–3 mo | Unmarkable, and the reason it is unmarkable is itself the finding. All three series are still last-observed 24 September at IG 79, HY 280, CCC 1,112bp — HY independently corroborated. Two trading sessions are missing and they bracket the Hormuz rejection, a 4% crude move, a near five-point jump in October hike pricing and a 7bp rise in the 10-year. The last available observation is emphatic in the view's favour: over 23–24 September IG +2bp, HY +12bp, CCC +37bp — CCC moved eighteen times as much as IG, with a daily path of +2, +5, +7, +19bp. 38bp from the level that confirms it; 62bp inside the level that closes it — roughly two sessions at the recent run rate | Unchanged: CCC inside 1,050bp with IG unchanged or tighter closes this; CCC through 1,150bp with IG flat confirms it and conviction goes to High. ⚠ The SoftBank read-through is a derived index inference, not reported secondary colour — a sixth consecutive attempt at secondary levels or a new-issue concession failed across five sources, and that item should now go to a dealer run or be dropped. Process note that stands: this view's trigger is an index level, so index silence is not evidence against it — the lesson from reversing Friday's premature conviction cut |
No view opened today, and that is a decision rather than an omission. Three candidates presented themselves and all three were declined. A Taiwan catch-up short — the TAIEX reopens today having priced none of two sessions of global semiconductor weakness, and it is the cleanest identifiable setup on the desk after the RBA — was passed because a one-session gap trade in an index this desk cannot mark intraday is not a view, it is a bet on an opening print. A gold re-entry in either direction was declined on the standing convention: V014 was stopped out at $4,300 and re-entering an instrument days after being stopped out of it is how one loss becomes two, regardless of how far it has since travelled. And a 5s30s steepener on the Treasury buyback theme was declined for the second consecutive edition, on Monday's own evidence: the curve flattened another ~3.5bp as the front end repriced, which is precisely the mechanism that killed V003. That is the fifth, sixth and seventh candidates named in advance and declined on evidence this month. Given that the newest view in the book has just had its central level invalidated and three of seven resolve in under eight hours, adding risk this morning would be the wrong instinct.
Portfolio-level read. Seven views, none closed, none opened, scorecard unchanged at 3 right / 12 wrong / 6 scratch across 21 decided. Better than yesterday: V004 is markable on clean data and working for the right reason for the first time in four editions, its sourcing blocker fully discharged; V025's catalyst date is resolved and turns out to collide with a French long-end auction on the same day; V017's underlying shape is intact and 38bp from confirmation. Worse than yesterday, and it is worth stating plainly: the newest view in the book was built on a moving average that was wrong by 474 points, which is a research failure rather than a market one, and it happened because a single vendor's technical block was taken without a plausibility check against the arithmetic of its own window — the same check that caught a 50-day moving average moving 32.87 points in a session last week. The correction was available at the time and this desk did not run it. Separately, the Australian concentration has now been named for a third consecutive edition without being reduced, and it resolves at 15:30 today — at which point restating the observation a fourth time will be pointless, so it is recorded here as a decision to accept the concentration explicitly rather than as an outstanding action.
These are analytical framings for a professional reader, expressed in the vernacular of a macro desk. They are not personalised investment advice, do not consider any individual's objectives or circumstances, and carry no position sizing. The “what changes the view” column is the accountability mechanism — every view is logged, marked and scored in the project's views ledger, including the ones that go wrong.
Positioning, flows & sentiment
| Indicator | Latest | Context | Read |
|---|---|---|---|
| CFTC Commitments of Traders — data as of Tuesday 22 September, released Friday 25 September. ⚠ NO NEW REPORT: nothing below is fresh. The next release is Friday 2 October with 29 September data | |||
| CBOT Treasuries — leveraged funds (TFF), all six contracts | net short 6,517,822 | Open interest 23,498,338 | ⭐ Reproduced from the raw category columns and it validates exactly (long 2,307,380 / short 8,825,202). Ultra-bond −824,343, bond −162,052, ultra-10y −395,678, 10y −1,926,947, 5y −1,858,062, 2y −1,350,740. 6.5 million contracts of short duration sitting under Friday's payrolls is the reflexivity in the whole note |
| Gold — non-commercial (COMEX) | net long 225,853 = 54.71% of OI | Long 253,982 / short 28,129 / spreading 48,923; OI 412,800 | ⭐ Reproduced from raw columns — the most crowded long in the entire report, and it fell 3.90% on Monday. Trimmed only 4,485 on the week. The crowding explains the violence, not the direction |
| Silver — non-commercial (COMEX) | net long 25,444 = 23.90% of OI | Long 34,701 / short 9,257; OI 106,474 | Less crowded than gold and it fell HARDER (−5.39% vs −3.90%) — so Monday was a broad precious-metals liquidation on a real-rate impulse, not a positioning-specific unwind |
| AUD — the largest disagreement in the report | LF net LONG 58,726 vs legacy net SHORT 46,814 | A 105,540-contract spread; open interest 306,488 on BOTH pages | ⚠⚠ Cross-venue validation PASSES, which means the divergence is real trader classification and not a data error. Same contract, same report, opposite signs. One of the two cuts is the wrong map for V023 and there is no way to tell which — so neither is used to mark it |
| NYMEX WTI — non-commercial | net LONG 141,106 | Long 360,810 / short 219,704; OI 1,841,811 | ⭐ The venue trap was avoided by reading the exchange header rather than taking the first crude block, and the figure reproduces exactly. Last edition's correction stands: escalation has far less short-squeeze fuel than this desk once published, and de-escalation carries real long-liquidation risk |
| Dollar index — ICE Futures U.S., legacy only | net LONG 10,330 = 22.3% of OI | Long 27,768 / short 17,438; OI 46,328 | Confirmed again: the dollar index is NOT on the TFF page — it is an ICE contract, so only the legacy cut exists. A small and unremarkable long |
| E-mini S&P 500 — leveraged funds (TFF) | net short 375,574 | Long 120,133 / short 495,707; OI 1,890,653 | Consistent with substantial cash-futures basis exposure rather than a directional bearish bet — and the distinction matters because the two unwind completely differently. Do not read it as a view |
| G10 FX — legacy non-commercial (CME) | JPY +71,982 · MXN +75,167 · GBP −82,568 · CAD −53,210 · EUR −52,334 · CHF −26,752 · NZD −11,380 | Nets derived from printed columns | Always name the series AND the venue — the leveraged-fund and legacy cuts have disagreed in sign on four of eight majors, and last month that cost a published error on the yen. Note the peso was net LONG 75,167 going into a ~1.43% depreciation |
| Options surface and market structure — re-fetched, not carried, and the whole sentiment stack failed to update for Monday | |||
| Net gamma exposure · gamma flip | NEGATIVE · flip ≈7,700 | Friday closed long gamma at +$67.0bn | ⭐ The SIGN is corroborated on three Monday sources and the MAGNITUDE is not (−$2.22bn against −$6.0bn, with a third source using per-1% normalisation and a fourth printing evidently broken units). Flip cluster 7,699 / 7,702 / 7,709 — it migrated UP ~20 points, which is why re-fetching rather than carrying was correct. With the close ~17 points below, dealer hedging now amplifies moves |
| Call wall · put wall · max pain | ⚠ 7,800 or 8,000 · ~7,600–7,675 · 7,725 or 7,450 | Five sources, three camps | ⚠⚠ V028's 7,800 close-out level is supported by exactly ONE source, and two others say 8,000. Two sources putting both walls within ±0.2% of spot were discounted as near-dated concentrations rather than structural walls — a judgement, not a finding. Max pain is genuinely two-camped |
| CBOE SKEW | 144.9 — 25 Sep, NO Monday print | 54th percentile vs 1 year · 93rd percentile all-time, both as the page states them simultaneously | ⚠ The page did not update, so the carried level is now two sessions stale and the apparent continuation of the 154.5 → 152.1 → 146.6 → 144.9 downtrend is staleness, not a new print. ⭐ The percentile to quote is the all-time one: at the 93rd percentile of all history, deep downside protection is NOT cheap, and the 54th-percentile one-year read only says tails are priced normally relative to a year in which they were persistently elevated |
| Equity put/call · VIX term structure | 0.55 (24 Sep) · IVTS 0.8293 (25 Sep) | Page-stated 14th percentile; contango day 119 | ⚠ Both stale — put/call is four sessions old, and ⚠⚠ its 14th percentile is computed over roughly 20 months of free history beginning in 2025, NOT a market cycle, so it must not sit beside SKEW's all-time basis as though comparable. No Monday IVTS exists because VIX3M is unpublished; a derived figure assuming VIX3M unchanged is not published |
| Technicals and breadth | |||
| S&P 500 200-day · 50-day | ≈7,205 · ≈7,636 | Price is +6.6% above the 200-day and +0.6% above the 50-day | ⭐⭐ THE CORRECTION OF THE EDITION — see §08. The carried 7,679.54 was wrong by 474 points. Two sources agree at 7,205.20 / 7,207.26 and at 7,636.09 for the 50-day. Neither moving average is in play at current levels, and the 50-day was NOT broken on Monday — using the outlier's 7,699.99 would have produced exactly that false conclusion |
| Daily pivots · RSI | not obtained · ≈low-to-mid 50s | RSI 56.69 at Friday's close | ⚠ The intraday-pivot trap fired again and the check caught it: the published “daily” block has R1 − S1 = 10.50 points, which asserts a 10.5-point daily range on the S&P. For classic pivots R1 − S1 = High − Low, so that is an intraday block mislabelled — and it is centred on Friday's close. The entire block is discarded and no daily pivots are published from any source |
| Breadth — Monday, exchange level | NYSE 422 new lows vs 16 new highs; Nasdaq 461 vs 50 | Decliners beat advancers by more than 2:1 | Severe internal deterioration on a −0.77% day, and the second consecutive session where lows are wildly skewed against the index move — Friday's session rallied on 308 lows to 46 highs. ⚠ A different universe from the 4,733-stock series below; do not compare the two |
| % of S&P 500 above the 200-day | 49.10% — 25 Sep, no Monday read | Long-run mean 65.34%; prior year-ends 62.10% (2025), 60.30% (2024) | ⚠ The page did not update — unchanged because it is stale, not because it was re-confirmed. Still below 50% for the first time in the carried series. The % above the 50-day remains formally dropped as a structural sourcing failure and is not carried as a gap |
| New highs/lows · McClellan · Hindenburg | 46 / 308 · −22.59 · 3 of 4, signal INACTIVE | 25 Sep; fails condition 1 (new highs 0.97% < 2.2% threshold) | ⚠ Stale at 25 Sep. ⚠ Correction: the universe is 4,733 eligible US common stocks, not the 4,751 carried — it is recomputed daily and must never be carried. ~2 sessions left in the warning window from a 17 August trigger |
| Flows — three universes, three vintages, labelled separately and NOT netted | |||
| ICI combined long-term flows | −$10.10bn, week to 16 Sep | Released 23 Sep — 12 days stale | Equity −$13.30bn (domestic −$12.15bn), hybrid −$2.17bn, bond +$3.24bn (decelerating sharply from +$11.55bn), commodity +$2.12bn. Mutual-fund outflows $36.70bn against ETF issuance $26.61bn — the conversion still dominates the gross figures. Label this ICI, never BofA or EPFR |
| ICI money market funds | $7.94tn, +$15.00bn, week to 23 Sep | A different and fresher vintage than the flows above — do not present as one week | The entire weekly build is institutional (+$15.26bn against retail −$256mn), and prime institutional did $12.91bn of it. With the 10-year at 5.24% that is yield-grabbing, not de-risking |
| BofA / EPFR Flow Show | Week to ~24 Sep, published 25 Sep | Bull & Bear Indicator 9.3 — an active contrarian SELL signal | Global equities −$10.2bn, US equities −$21.1bn, bonds +$17.3bn — a 74th consecutive week of inflows. ⭐ BofA private clients: $4.6tn AUM, 66.1% in equities, cash at a RECORD LOW 9.4%. Hartnett's base case is “Debasement & Duration”. ⚠ The 9.7 (August) and 8.8 (June) readings are stale and must not be quoted as current |
| Sign comparison | ICI equity −$13.3bn · BofA US equity −$21.1bn | Different weeks, different universes | They agree in sign this time — and that is not confirmation. Do not add them and do not treat the agreement as corroboration. ⚠ Lipper: DECLARED FAILED for a fifth consecutive edition across both routes; recommend formally dropping it as was done with the % above the 50-day |
| Crypto ETF flows | BTC +$2.4bn, ETH +$689.9mn, week to 25 Sep | Single-sourced | The largest weekly bitcoin ETF inflow in nearly a year, flipping 2026 net flows positive to +$934.1mn from roughly −$5.8bn in mid-July. IBIT $1.2bn, FBTC $701.7mn, ARKB $294.7mn. ETH reversed the prior week's −$140mn |
| Systematic / corporate | CTAs 21:1 asymmetric · vol-control ~86% | $84bn of selling capacity in a down market against $4bn of buying in an up one | The corporate bid goes from 10% of index weight in blackout to 61% by 30 September and does not reopen until 1 November. ⚠ No fresh CTA or vol-target estimate was obtainable this edition — these are carried from the 28 September edition |
| Sentiment, valuation and earnings | |||
| AAII | Bulls 32.7% · neutral 19.2% · bears 48.1% | Week of 23 Sep; next reading 30 Sep | Bull–bear spread −15.4, improved from −24.5 but a third consecutive week of bearish plurality. The long-run averages are NOT on the dated results table and are not quoted from memory. ⭐ The divergence is the signal: retail is outright bearish at 48.1% while BofA private-client cash sits at a record-low 9.4% and the Bull & Bear is on a sell. Survey sentiment and actual allocation point opposite ways — and allocation is the one that has to unwind |
| BofA Fund Manager Survey | September remains current | Published 18 Sep; survey 4–10 Sep, n=190, $512bn AUM. October FMS due ~mid-October | Cash 3.9% of AUM (a Cash Rule sell signal); net 49% overweight global equities; net 48% underweight bonds, the most since May 2022; net 25% say policy is too stimulative; most-crowded trade “buying global semiconductors” at 53%. ⭐⭐ The panel's own number-one tail risk — a disorderly rise in bond yields — is now the realised market, against a panel positioned 48% underweight bonds |
| Valuation — FactSet, 25 Sep | Forward P/E 19.2× | 5-year average 19.8× · 10-year average 19.0× | Below its own five-year mean and only marginally above the ten-year — on a forward-multiple basis this is NOT a stretched market. The valuation risk is in the discount rate at 5.24%, not the multiple. Bottom-up target 9,275.04, +20.4% above the 24 Sep close — an aggregated analyst figure, not a strategist target, and it habitually runs well above them |
| Earnings | Q3 2026 estimated +29.1% y/y; revenue +12.1% | Guidance 72 positive / 44 negative | ⚠ +29.1% is extraordinarily high for a blended S&P quarter, is single-sourced, and is reported as printed with a plausibility flag. It leaves minimal room for disappointment, though the 72:44 guidance ratio is a cushion. The season's front edge is this week; banks and the bulk of the index from mid-October |
| Sell-side targets | HSBC 8,100 — Nicole Inui, 13 Sep | 2026 EPS $360 at a 22.5× multiple — 3.3 turns above FactSet's actual 19.2× | ⚠⚠ The dateline trap fired FIVE times this edition and all five were quarantined — items dated 13 June, 15 September, 9 June and 12 May 2026, and one from 28 September 2023 whose matching day-and-month is exactly how the trap works. No target revision later than 13 September exists, which is a notable silence given the 2-year has moved 56bp this month |
The week ahead
| Day | AEST | ET | Event | Cons. | Prior | Imp. |
|---|---|---|---|---|---|---|
| TUESDAY 29 SEPTEMBER — RBA day. Taiwan reopens after two closed sessions; mainland China trades the second of three before Golden Week | ||||||
| Tue | 11:30 | Mon 21:30 | AU Monthly Household Spending Indicator (Aug) — lands three hours before the decision | M | ||
| Tue | 14:30 | 00:30 | RBA CASH RATE DECISION + statement — issuer-verified | 4.60% | 4.35% | H |
| Tue | 15:30 | 01:30 | RBA press conference — the actual event. Watch for “restrictive conditions” language and any shift onto the employment mandate | H | ||
| Tue | 18:30 | 04:30 | UK M4 money supply m/m · mortgage approvals · net lending to individuals | +0.1% · 56K · £6.2bn | −0.3% · 56K · £6.3bn | L |
| Tue | 22:30 | 08:30 | Canada GDP m/m — two banks now forecast an October BoC hike | 0.0% | +0.3% | M |
| Tue | 00:00 Wed | 10:00 | US Conference Board consumer confidence · JOLTS job openings | 89.2 · 7.23M | 89.4 · 7.27M | H |
| Tue | 01:00 · 01:30 Wed | 11:00 · 11:30 | BoE Mann (McKinsey Global Institute panel) · BoE Taylor — NIESR Dow Lecture, on expectations and inflation persistence | H | ||
| Tue | US day | — | Fed speakers: Bowman, Barr, Goolsbee, Musalem, Williams, WALLER — the heaviest slate of the week (single-sourced list) | H | ||
| Tue | — | — | US Treasury buyback: TIPS 1Y–10Y, up to $750m | L | ||
| WEDNESDAY 30 SEPTEMBER — the densest session of the week, and the last mainland China session before a seven-day closure | ||||||
| Wed | 11:30 | Tue 21:30 | AU Monthly CPI indicator y/y · m/m — time issuer-confirmed · also building approvals m/m | ⚠ see note | 3.5% · 1.0% | H |
| Wed | 11:30 | Tue 21:30 | CHINA official manufacturing PMI · non-manufacturing PMI — the last fresh mainland signal for a week ⚠ time not confirmed from the NBS | 50.1 · 49.2 | 49.8 · 49.0 | H |
| Wed | 11:30 | Tue 21:30 | RBA Chart Pack release | L | ||
| Wed | ? | ? | ⚠ German, French, Italian and Spanish national flash CPIs — DATE, TIME AND CONSENSUS ALL UNCONFIRMED across three routes. Conventionally 30 Sep–1 Oct | ⚠ n/a | M | |
| Wed | 20:25 · 23:50 | 06:25 · 09:50 | BoE Benjamin (ISDA) · BoE Evans (ISDA, Basel III) | L | ||
| Wed | — | — | ⭐ BoE Financial Policy Committee RECORD published — a live financial-stability event with the 30-year gilt near 5.89% | M | ||
| Wed | 22:30 | 08:30 | US CORE PCE m/m · headline PCE · personal spending m/m · final Q2 GDP q/q | +0.3% · · +0.8% · 1.5% | +0.2% · 3.3% y/y · +0.2% · 1.5% | H |
| Wed | 22:30 | 08:30 | UST 2-year, 5-year AND 7-year all SETTLE — a front-end duration add in the same minute as core PCE. No coupon supply then until 6 October | M | ||
| Wed | 23:45 | 09:45 | US Chicago PMI | 51.2 | 47.1 | L |
| Wed | US day | — | Fed speakers: Barkin, Cook, Goolsbee, Kashkari · US Treasury buyback 10Y–20Y ≥$4bn is Thursday | M | ||
| THURSDAY 1 OCTOBER — the day everything lands at once. Mainland China SHUT (to 7 Oct); Hong Kong shut for the day only | ||||||
| Thu | 09:50 | Wed 19:50 | JAPAN TANKAN large manufacturers DI · large non-manufacturers DI — ⭐ consensus dispute settled at +26 on a 12-institution panel | +26 · +36 | +22 · +37 | H |
| Thu | 09:50 | Wed 19:50 | BoJ Summary of Opinions, 17–18 September meeting — the read on whether the two dissents were dovish or hawkish | H | ||
| Thu | 11:30 | Wed 21:30 | AU goods trade balance · RBA Financial Stability Review 11:30 | A$2.00bn | A$1.92bn | M |
| Thu | — | — | ⭐⭐ FRANCE: PLF 2027 to the Conseil des ministres (date resolved this edition; a 30 Sep variant traces to an undated source) AND the AFT's long-term OAT auction — the same day. €54bn sought, 5.0% deficit target, 121.7% debt | H | ||
| Thu | 18:00 · 18:30 · 19:00 | 04:00 · 04:30 · 05:00 | EZ final manufacturing PMI · UK final manufacturing PMI · EZ unemployment rate | 52.7 · 52.0 · 6.4% | 52.7 · 52.0 · 6.4% | L |
| Thu | 18:00 · 22:00 · 22:30 | 04:00 · 08:00 · 08:30 | BoE Bailey (LSE/BoE Future of Money, opening remarks) · BoE Mann (Nomura) · BoE Pill | M | ||
| Thu | 22:30 | 08:30 | US initial jobless claims | 201K | 197K | M |
| Thu | 00:00 Fri | 10:00 | US ISM MANUFACTURING PMI — and prices-paid is the component that matters in this regime | 54.8 | 54.6 | H |
| Thu | ≈06:30 | Wed 16:30 | ⚠ MICRON fiscal Q4 results and call — CORRECTED: Thursday 1 October ~06:30 AEST, NOT Wednesday 05:00. Verified from Micron IR (call 2:30pm Mountain, Wed 30 Sep). The AI-memory read-through for the crowded semis trade | EPS ≈$31.56 | H | |
| Thu | — | — | US Treasury buyback 10Y–20Y, ≥$4bn · Fed speakers: Barkin, Collins, Schmid, Waller, Jefferson, Bowman, Cook, Logan | M | ||
| FRIDAY 2 OCTOBER — payrolls, and the COT report. India SHUT (Gandhi Jayanti); mainland China shut; Hong Kong TRADES | ||||||
| Fri | 09:30 | Thu 19:30 | TOKYO CORE CPI y/y — ⚠ date corrected: Thursday in ET is FRIDAY 09:30 in AEST. ⚠⚠ A 2.4% consensus against a 1.8% prior is a 0.6pp step far outside the series' 2026 range of 1.3–2.0% — sanity-check before trading it | 2.4% | 1.8% | H |
| Fri | 19:00 | 05:00 | EURO-AREA FLASH CPI y/y · core — a +50bp jump in headline is the most consequential number on the forward calendar | 3.7% · 2.5% | 3.2% · 2.4% | H |
| Fri | 22:30 | 08:30 | US NON-FARM PAYROLLS · unemployment rate · average hourly earnings m/m — ⚠ consensus DISPUTED between +90K and +98K across two calendars; resolve before the print, it changes the surprise threshold | +90K to +98K · 4.1% · +0.3% | +162K · 4.1% · +0.3% | H |
| Fri | 00:00 Sat | 10:00 | US factory orders m/m · CFTC Commitments of Traders release (29 September data) — the first new positioning read in a week | +0.1% | +0.9% | M |
| The sessions after — the next fortnight of scheduled events | ||||||
| Mon 5 | Mainland China shut (to 7 Oct). BoE Sales. US Treasury 3-year auction announced 1 Oct | L | ||||
| Tue 6 | 13:00 | UST 3-year auction (announced 1 Oct, settles 15 Oct) · Treasury buyback 2Y–3Y $4bn · BoE Mann (TS Lombard), her third appearance | M | |||
| Wed 7 | 13:00 | RBI MPC decision — a HIKE to 5.50% is consensus, not merely priced, with commentary framing October as possibly too late given oil · UST 10-year reopening | 5.50% | 5.25% | H | |
| Thu 8 | 13:00 | ⭐ UST 30-YEAR REOPENING — and Treasury buys back ≥$4bn of 20Y–30Y paper the SAME DAY. The cleanest test of whether official demand is the marginal price-setter at these levels · mainland China REOPENS with a week of accumulated gap risk | H | |||
| Mon 13 | 11:30 | RBA September minutes (AEDT) | M | |||
| Wed 14 | 22:30 | 08:30 | ⭐ US CPI (September) — DATE NOW PUBLISHED BY BLS, closing a multi-edition gap. The last major inflation print before the FOMC, two weeks ahead of it and three days before the 17 October blackout | H | ||
| Thu 15 | 08:30 | US PPI (September) · Treasury buyback 10Y–20Y ≥$4bn · 3y/10y/30y all settle | M | |||
| Fri 17 | Fed blackout begins · FOMC minutes of the 16 September meeting due ~6–7 Oct — re-tasked target for the named voting roster, which is structurally unavailable from the statement | M | ||||
| 19–22 | China LPR 19–20 Oct (⚠ unconfirmed) · Bank of Korea Thu 22 Oct, issuer-confirmed · Philippines 22 Oct (⚠ unconfirmed) | M | ||||
| 27–29 | FOMC 27–28 October, decision Wed 28th (05:00 Thu 29th AEDT), NO SEP · RBNZ Wed 28 Oct 12:00 AEDT — a REVIEW not an MPS, issuer-confirmed · BoC Wed 28 Oct · UK Budget Wed 28 Oct · ECB Thu 29 Oct (00:15 Fri 30th AEDT) · Treasury buyback 20Y–30Y 27 Oct | Fed 71.2% hike | 3.75–4.00% | H | ||
| Fri 30 | BoJ decision Fri 30 October with the Outlook Report the same day — a single binary · US ECI Q3 | ⚠ ~11.5% | 1.25% | H | ||
| 3–5 Nov | 14:30 | RBA Tue 3 Nov 14:30 AEDT with a full Statement on Monetary Policy · Riksbank 4 Nov (⚠ unconfirmed) · BoE Thu 5 Nov with an MPR (85.7% priced) · Norges Thu 5 Nov 20:00 AEDT, issuer-confirmed | H | |||
⚠ Every H/M/L importance rating in this table is this desk's own judgement and is labelled as such. The calendar aggregator used as the base has a permanently broken impact column — for a fourth logged occasion every row parsed as “Low”, including central-bank decisions — while its event list, times, consensus and priors have cross-checked against issuer pages without conflict. Consensus figures are as of Monday 28 September and shift; where two calendars disagree the range is shown (see payrolls). Times are AEST (UTC+10); Sydney moves to AEDT on 4 October, after which AEST+1 applies to every Sydney time above and AEDT = EDT + 15h until US daylight saving ends on 1 November. Unconfirmed items are marked rather than dropped.
Risk radar
| # | Risk | Trigger / timing | Probability | Hedge / expression |
|---|---|---|---|---|
| 1 | The RBA statement resolves two views at once and a third is the same currency | TODAY 14:30, presser 15:30 | Hike ~90% (soft, 25 Sep stamp) | The hike is not the risk; “terminal” versus “first of several” is. The 3y embeds +44bp over a post-hike 4.60% — a terminal framing has that to unwind, which steepens 3s10s against V004 and rallies exactly the financials and real estate V006 is short. Named three editions running and not reduced; now explicitly accepted rather than restated |
| 2 | Payrolls as a yield accelerant, with the asymmetry inverted from normal | Fri 2 Oct 22:30 AEST | ⚠ Consensus disputed +90K to +98K | A STRONG print is the bearish outcome — it pushes the 10-year further through 5.24% and raises October hike odds already at 71.2%. But the mirror is violent: 6,517,822 contracts of leveraged-fund Treasury shorts is the squeeze fuel if the data turns. Own convexity, not direction |
| 3 | Core PCE lands in the same minute as a three-tenor settlement | Wed 30 Sep 22:30 AEST | Oct 71.2% · Dec two hikes 58.9% | Consensus core +0.3% with spending +0.8% against a +0.2% prior — a nominal-demand signal. The 2y, 5y and 7y all settle that minute, then no coupon supply until 6 October. The Cleveland Fed's September core-PCE nowcast at 3.49% y/y says the trend is going the wrong way |
| 4 | Short gamma with no trend support below | Continuous; ~7,700 flip | CTA asymmetry 21:1 | ⭐ The correction changes this risk's shape rather than removing it. Net gamma is negative and the index sits ~17 points below the flip, so hedging amplifies — but the 200-day is ~7,205, so there is no trend support for 6.6%. That is worse for a drawdown scenario, not better. Blackout peaks at 61% of index weight on 30 Sep, no reopening until 1 Nov; vol-control ~86% |
| 5 | Credit's two dark sessions bracket the largest news of the week | Any observation after 24 Sep | CCC 1,112bp; 38bp from confirm | The staleness IS the risk. IG/HY/CCC all last-observed 24 September; 25 and 28 September are missing and they bracket the Hormuz rejection, a 4% crude move and a 7bp rise in the 10-year. The last print showed CCC widening 18× IG on an accelerating path — the risk is that it extends far enough to matter for equity multiples |
| 6 | The French budget and a long-end OAT auction on the same day | Thu 1 Oct — both | — | ⭐ The date is resolved this edition and the collision was not previously visible. €54bn buys only ~0.4pp of deficit; LFI, the Greens and the PCF are committed to censure and the RN's 122 votes are decisive at a published price of no new taxes (⚠ carried, not verified). Both branches are OAT-negative, and per the V003 post-mortem strong auction cover is evidence about clearing, not direction |
| 7 | Crude gaps in either direction on a rejection that is not yet final | Qatari-mediated talks “as early as Monday” | US officials: agreement “extremely slim” | Trump rejected the seven-day roadmap Saturday; Araghchi says Tehran awaits a formal response through mediators, so Iran has not treated it as final. ⚠⚠ The asymmetry is NOT a short squeeze: NYMEX WTI non-commercials are net LONG 141,106. De-escalation carries real long-liquidation risk. Nov/Dec backwardation of $8.34 says the market reads it as finite |
| 8 | China's PMIs print on the last session before a seven-day shutdown | Wed 30 Sep; mainland shut 1–7 Oct | Mfg consensus 50.1 vs 49.8 | Repositioning must clear in one session or wait a week, which mechanically raises the odds of an outsized Wednesday move and pushes a week of gap risk onto the 8 October reopen. Hong Kong trades from 2 October and becomes the sole venue for China price discovery — do not read HSI or HS Tech as clean HK signals in that window |
| 9 | The crowded semiconductor trade meets an earnings test it has already started failing | Micron Thu 1 Oct ~06:30 AEST (date corrected) | Most-crowded trade 53% | Monday broke it in four tapes at once — Korea −2.70%, a 1,157-point Nikkei reversal, Shenzhen −3.44%, HS Tech −0.31% against a rising HSI, plus ASML/BESI/Infineon in Europe. Taiwan reopens today having priced none of it and TSMC is >40% of its market cap. ⚠ Note the Tankan's +26 consensus rests on the AI demand story the tape is repudiating |
| 10 | The long end rises while Treasury buys it, and on 8 October it does both at once | 8 Oct 30-year reopening + ≥$4bn buyback, same day | — | Doubled buybacks since 9 September and the 30-year is at a cycle high anyway. Last week's 7-year took indirects of 57.2% against a 64.6% average. Own it as an observation — a tail into a same-day buyback is the cleanest evidence official demand is not the marginal price-setter. A 5s30s steepener has been declined twice on the V003 post-mortem |
| 11 | Gold and duration are now the same trade, and a book holding both is not diversified | Continuous | Gold net long 54.71% of OI | Monday is the demonstration: crude +3.25% with gold −3.90%. In a real-rate regime the geopolitical hedge and the growth hedge lose together. Gold is ~12% below end-August and made a seven-week low; silver at 23.9% of OI fell harder at −5.39%, so this is broad liquidation. V014 is not re-entered in either direction |
| 12 | The RBI is expected to hike into the oil shock and may already be late | Wed 7 Oct | 5.50% is consensus | India is the most oil-sensitive major Asian market and traded like it on Monday (−1.5%, all ten sectors lower, three of fifty Nifty names up, FIIs −₹5,353cr). Week eight after a seven-week losing streak. The first hike after a cutting cycle |
| 13 | The household-versus-institution gap is the widest thing in the note | Confidence today; claims Thursday | AAII bears 48.1% vs private-client cash 9.4% | Retail is outright bearish while actual allocation is at a record-low cash weight and the Bull & Bear is on a sell at 9.3. Survey sentiment and allocation point opposite ways and allocation is the one that has to unwind. Germany's −30.6 and Michigan's 48.1 say the same on the household side of both economies |
| 14 | The Australian household channel is tightening faster than the Bank is | Continuous; auctions each weekend | — | Eighteen lenders repriced in September, all four majors up to 0.48pp — roughly double today's step, already delivered. Auctions at a ten-week-low 50.3% and flattered by a ~70% Melbourne volume collapse; values −3.1% over three months; business conditions at a six-year low. This is the mechanism under V006 and the argument against a hawkish presser |
| 15 | Euro-area flash CPI is consensus to jump 50bp into a Council that has talked October down | Fri 2 Oct 19:00 AEST | Oct 38.6% · Dec 93.3% | Consensus 3.7% headline against 3.2%. The tone meter moved markedly dovish with Schnabel — the hawk — named as the driver, and Lane pointed to December. A 3.7% print against that guidance is the sharpest available test of whether “measured response” survives contact with the data. ⚠ The national flashes could not even be dated |
Key levels
| Instrument | Last | Support | Resistance | Comment |
|---|---|---|---|---|
| S&P 500 | 7,683.26 | 7,680 (V028 confirm) · ~7,636 (50d) · ~7,205 (200d) | ~7,700 (gamma flip) · 7,800 or 8,000 (call wall) · 7,816.70 (record) | ⭐⭐ THE FOUR-WAY CONFLUENCE PUBLISHED YESTERDAY DOES NOT EXIST. The 200-day is ~7,205, not 7,679.54 — the index is 6.6% above it and there is no trend support anywhere near. What is real: the close is ~17 points BELOW a ~7,700 gamma flip, so hedging amplifies. It closed 3.26 points above the published 7,680 trigger, which has not fired and is not being moved. ⚠ The 50-day was NOT broken. ⚠ No daily pivots obtained |
| Nasdaq Comp · Russell 2000 | ≈26,763–26,783 · ≈2,818.5 | 26,500 · 2,800 | 27,068.72 · 2,837.55 | ⚠ Neither is a published close. The Nasdaq's 26,820 read is a 12:16 ET midday snapshot and is discarded; the Russell rests on a 3:49 PM capture on a day the tape sold into the bell. Both prior closes are where resistance now sits |
| UST 10y · 30y | 5.24% · ≈5.53–5.58% | 5.17 (25 Sep official) · 5.49 | 5.26 (intraday, single source) · 5.58 | The 10y is corroborated at a fresh cycle high and the 30y is genuinely unresolved across three sources. ⚠ Treasury's official 28 Sep par row has NOT posted — it is tomorrow's first verification target |
| UST 2y · 5y · 5s30s · 2s10s | ≈4.91–4.93 · ≈5.05% · ≈47–48 · ≈31bp | 4.81 · 4.98 · 44 · 31 | 4.93 · 5.00 · 51 · 36 | The 2-year has added ~56bp across September, the largest monthly rise since Feb 2023. The 5-year has never closed above 5% on the official series and is testing it. Both curve measures FLATTENED — the mechanism that killed V003 and the reason the re-entry was declined twice |
| JGB 2y · 10y | 1.97 · 3.09% | 1.90 · 3.00 | 2.00 · 3.11 (30-yr high) | The 2-year is the instrument: +72bp over a 1.25% policy rate, close to three hikes, against a circulating ~11.5% October probability that is itself attached to a wrong meeting date |
| Bund 10y · OAT–Bund · BTP–Bund | 3.629% · 105.4bp (25 Sep) · 94.1bp | 3.50 · 80 (V025 closes) | 3.70 · 109.9 (52-wk wide) | Bund 10y still the highest since June 2009. ⚠ No 28 Sep OAT–Bund print exists; a cross-page derivation implies ~109bp and is NOT published as a mark. Italy trades ~13bp inside France |
| Gilt 10y · 30y · 2y | 5.403 · 5.89 · 4.77% | 5.35 · 5.80 · 4.69 | 5.45 · 6.00 · 4.85 | The 2y's +8bp was the biggest G4 front-end move of the session and it has a named cause. 30y described as the highest since 1998; 2s10s narrowed 3.6bp to 63.3bp |
| ACGB 3y · 10y · 3s10s | 5.04 · 5.41% · 37bp | 4.95 · 5.30 · 35 | 5.10 · 5.50 · 43 (V004 entry) | Clean marks on all three for the first time in four editions. 3s10s 6bp in the money. The 3y embeds +44bp over a post-hike 4.60% — that is what 14:30 and 15:30 are trading against |
| DXY · EUR/USD · USD/JPY | 101.20 · ≈1.1373 · ≈157.23 | 100.97 · 1.1350 · 156.94 | 101.50 · 1.1450 · 160 (policy line) | The DXY is the ONLY FX close published this edition and it ties exactly — though it does not reconcile to its own legs. USD/JPY sits ~2.8 yen below a line the 3 August joint US–Japan intervention defended |
| AUD/USD · AUD/NZD | ≈0.7026 · ≈1.2384 | 0.6990 · 1.2315 (entry) | 0.7115 · 1.2450 | ⚠ Neither is a verified close. The AUD per-pair page was Sunday-stamped and rejected; the AUD/NZD cross is same-table but its NZD leg is disputed by 107 pips. Dovish guidance at 15:30 puts 0.6990 in reach |
| Brent · WTI | $107.71 · $96.23 | 104.32 · 92–95 (V024 re-own) | 110 · 100 | Nov-26 front month, verified with an exact tie. Nov/Dec backwardation $8.34 in one month; Brent–WTI narrowed to $11.48. ⚠ The vendor's Brent page tracks Dec-26 at $99.37 — read the contract label before trusting any Brent history table |
| Gold · Silver | $4,117.70 · $60.80 | 4,110.20 (Mon low) · 60.00 | 4,200 · 4,300 (V014 stop) · 4,279.90 | ⭐ Verified on five routes. A decisive break of $4,200 and a seven-week low, ~12% below end-August — the fourth consecutive close below the V014 stop and the largest. Silver underperformed gold by ~150bp, reversing Friday's framing |
| Copper · Iron ore | $14,740/t (25 Sep) · $96.92/t | 14,500 · 95 | 14,900 · 100 | Cash−3M +$93/t, 6.6× wider in five sessions on stocks falling 3,600t — the 28 Sep official settle posts today. Iron ore's FIFTEENTH sub-$100 print, and no fresh China demand signal after tomorrow until 8 October |
| ASX 200 | 8,679.70 | 8,639.90 · 8,600 | 8,702 · 9,005.9 (V006 entry) | Up 0.17% INTO the hike on Financials +1.16% — the uncomfortable detail for the underweight, which is still −3.62% in the money. ⚠ Volume, breadth and the A-VIX all unobtainable; Friday's A-VIX 11.53 and falling is the last verified read. SPI (APZ2026) 8,727.5 — the +62.5 basis to cash is December carry, NOT a signal |
| Nikkei · TOPIX · Hang Seng Tech | 65,878 · 4,112.00 · 4,298.49 | 65,000 · 4,050 · 4,250 | 67,035 (Mon intraday, first time through 67,000) · 4,150 · 4,350 | A textbook reversal: through 67,000 for the first time, then −1,157 points. ⚠ Nikkei corroborated not exchange-verified — the archive has not posted. HS Tech −0.31% against an HSI +0.54% is the 85bp tell |
| IG / HY / CCC OAS | 79 / 280 / 1,112bp (24 Sep) | CCC 1,050 (V017 closes) | CCC 1,150 (V017 confirms) | 38bp from confirmation, and two dark sessions bracketing the week's biggest news. CCC widened 18× IG on the last available print, path +2/+5/+7/+19bp |
| VIX · SKEW | ≈16.3 · 144.9 (25 Sep) | 14.87 · 140 | 16.43 (Mon high) · 150 | ⚠ The VIX move (+7–8%) is corroborated and the level is not; VIX3M is unpublished so there is no Monday IVTS or contango count. ⚠⚠ SKEW did not print Monday — the carried level is two sessions stale. Quote the 93rd all-time percentile, not the 54th one-year: deep protection is expensive |
Data notes & sources
How this edition was built, and why the US block is the weakest part of it
Every figure was retrieved on 29 September 2026; nothing is recalled from memory. The structural timing problem the framework documents for Tuesday-to-Friday runs applied in full. Asia, Australia and Europe closed hours before filing and were drafted first, and those sections are the firm part of this note — twenty-one index closes were reconstructed against verified priors and reported only where they tied. The US block was done last, at roughly 06:50–07:15 AEST, fifty to seventy-five minutes after the bell, and three of its four authoritative sources had not published: AP's tabulation of record was found in a search index but 404s on fetch at the only domain carrying it, and no syndication partner had indexed it; Treasury's official par curve still shows 25 September as its last row, as expected before ~18:00 ET; and the Fed Rate Monitor's post-17:00 ET settled refresh had not fired when checked at 17:07 and again at 17:10 ET — the table was byte-identical to its 15:25 pre-close snapshot. It then refreshed DURING THE BUILD, to a 16:45 ET post-close read, and the numbers moved materially: October hike pricing from 74.6% to 71.2% and December's two-hike modal from 60.3% to 58.9%. This note publishes the later read, and the episode is the clearest vindication of the framework's rule against quoting this page's mid-afternoon snapshot: the figure that would otherwise have led the edition was wrong by 3.4 points on the most-quoted number in it. Note also that the source recomputes its own comparison window — its “last week” column moved from 57.4% to 59.7% between the two fetches — so that column should be quoted with the fetch time attached, not carried. So US cash closes are corroborated or derived rather than verified, the US curve is a vendor read with the primary named as tomorrow's first target, and the most-quoted number in the note carries an explicit pre-close caveat. Two US figures are not published at all: the Nasdaq Composite and the Russell 2000 closes. That is the correct outcome under the framework rather than a failure of it — a shorter accurate edition beats a complete invented one — but it is worth stating plainly that on a Tuesday this note is at its weakest exactly where the reader's attention is highest.
Corrections — fifteen, six material, and one invalidates the central premise of a view opened yesterday
(1) MATERIAL, and it breaks a view: the S&P 500's 200-day moving average is approximately 7,205, not 7,679.54 — an error of 474 points, or 6.6%. That figure was carried into V028 yesterday as half of a claimed “two-mechanism level”. Two independent sources agree at 7,205.20 and 7,207.26; the series' own year-end history (6,286.45 at 31 December 2025) makes ~7,205 the arithmetically plausible value; and the outlier source prints its MA100 below its MA200 while price sits 1.8% off the high, which is not a coherent ordering. The plausibility check that would have caught this — testing a moving average against the arithmetic of its own window — is the same check this desk ran successfully on a 50-day last week and did not run on this 200-day. It was available at the time. (2) MATERIAL, same view: the dealer gamma flip is ~7,700, not 7,680, on three Monday reads clustering 7,699/7,702/7,709; the single source printing 7,680.77 places spot above its own flip while reporting negative net gamma, which cannot both be true. So the “confluence at 7,680” was an artefact on both legs. (3) MATERIAL: Micron reports Thursday 1 October at approximately 06:30 AEST, not Wednesday at 05:00 — verified from the company's own investor page (call 2:30pm Mountain on Wednesday 30 September). A full day and ninety minutes wrong on a named earnings catalyst for the most-crowded trade in the market. (4) MATERIAL: the BoJ “quarterly hikes to 2% by June 2027” signal came from Makoto Sakurai, a FORMER board member, not a serving official. Yesterday's edition and the views ledger both described him as “a BoJ official”, which materially overstates the source; his actual base case is December, with October live only on a large forecast revision. (5) MATERIAL: the “$1.27m/day record” VLCC rate cannot be tied to a named assessor and is withdrawn. The verifiable statement is “above $1.2m/day, Gibson Shipbrokers, 21 September”. (6) MATERIAL: the Trump–Xi summit DID take place, on 24–25 September, and a tariff deal was announced on Saturday 26 September — ~$30bn of goods each way, with China lowering on agriculture, seafood, wood, cosmetics and medical devices and committing to 10m metric tons of US coal in each of 2027 and 2028, the US lowering on small appliances, toys, holiday decorations and children's car seats, and rare earths explicitly unresolved; verified across six outlets. Yesterday's edition carried the summit as unconfirmed. ⚠ The reported extension of the truce to January 2027 could not be verified and is flagged separately. (7) The September Tankan consensus is settled at +26 for large manufacturers (from +22) and 36 for non-manufacturers — two independent outlets, same twelve-institution panel; the +25 variant is retired. (8) German October consumer climate: −27.4 was the CONSENSUS and the prior was −26.8 revised, so the fall was 3.8 points, not 3.2. (9) Taiwan was closed on Monday 28 September (Teachers' Day) and had also been closed Friday the 25th — it reopens today, not after a Thursday close. (10) The Hindenburg universe is 4,733 eligible US common stocks, not the 4,751 carried; it is recomputed daily and must never be carried. (11) Iron ore's sub-$100 streak is fifteen, not fourteen. (12) Tokyo CPI is Thursday in ET but Friday 09:30 in AEST — this desk had it on the wrong side of the dateline. (13) The UK Chancellor is John Healey under Prime Minister Andy Burnham. (14) BoE Mann speaks three times this week and next, not twice. (15) The carried ECB October pricing of ~48% could not be confirmed from any source; the best-sourced figure is 38.6%, dated 25 September — less hawkish than published, a direction the dovish tone-meter shift corroborates.
Two standing items CLOSED, and one gap closed after several editions
⭐⭐ GOVERNOR COOK: CLOSED, VERIFIED FROM THE PRIMARY. The Federal Reserve's own Board-members page lists “Lisa D. Cook — Governor” among the seven sitting Governors, and its FOMC page lists her among the twelve current voting members. A standing daily check carried since edition 001 is retired. A search result suggesting a 28 September Cook speech was spurious — the URL 404s and the 2026 speeches listing carries no September Cook entry — but the membership pages settle the substantive question outright. Incidentally confirmed from the same pages: Warsh is Chairman and Powell is listed as a Governor. ⭐ THE US SEPTEMBER CPI DATE IS PUBLISHED: Wednesday 14 October, 08:30 ET / 22:30 AEST, from the BLS October schedule — an unfilled gap for several editions. It lands two weeks before the FOMC and three days before the 17 October blackout, which makes it the last major inflation print of the cycle. From the same schedule: PPI 15 October, ECI Q3 on 30 October. ⭐ THE RBNZ DATE IS SETTLED FROM THE ISSUER after two editions: Wednesday 28 October, 14:00 NZ = 12:00 AEDT — and it is a Monetary Policy REVIEW, not a full MPS, with a 15:00 NZ press conference. ⭐ The SPI is quotable for the first time in eleven editions — December contract APZ2026 at 8,727.5 — though the quote is stamped 26/09 and the +62.5 point basis to cash is December carry rather than a signal.
Conflicts and how they were resolved
The US 10-year cluster is RESOLVED at 5.24%. Two clusters were in conflict (5.223 against 5.242) and an independent third source settled it: it prints 5.24% for 28 September with a +0.07 change implying a 5.17% prior, which matches the official par 10-year for 25 September exactly. Positive evidence that the lower cluster is a morning read rather than a settle: a quote page showed 5.22 at 07:50 CDT, so 5.22 was an early level the market then rose above. The 30-year is NOT resolved and is published as a range of 5.53–5.58% — three sources give 5.525, 5.55 and 5.576, and the middle matches neither of the others. The Swiss market index gap was resolved by the reconciliation test itself, which is a first: two sources gave two Mondays, and the CFD page's −6.41 change implies a prior 4.00 points wrong while the cash page's −2.80 ties exactly and states its own prior. The Hang Seng was resolved against its own page's quote box: the box fails the chain by 16.16 points while the narrative ties to within 0.9, so 24,643 is used. The Brent contract-month conflict was resolved by reading the live page's contract label — three clusters existed because the vendor's Brent page tracks December-26 while its WTI page tracks the front month, and the highest cluster ($107.71, November-26) was correct. The Tankan consensus and the French cabinet date were both resolved on recency and source count (+26 on two outlets; 1 October on two sources including one published Monday morning against an undated page giving 30 September). Left open with both sides published: the 30-year Treasury; the Nasdaq Composite close; the US 2-year, whose lower read implies a prior 5.4bp from the official series — too wide for pure par-versus-on-the-run methodology; the gilt 10-year's stated prior (5.35% against a carried 5.3593%); Karoon Energy's Monday move (−12.9% against −12.3%); net gamma magnitude (−$2.22bn against −$6.0bn); the call wall (7,800 against 8,000) and max pain (7,725 against 7,450); the payrolls consensus (+90K against +98K); and the 50-day moving average, where two sources agree at 7,636.09 against a vendor's 7,699.99 and the desk's own plausibility rule cannot adjudicate between two Friday reads from different providers.
Rejected outright
The carried 200-day of 7,679.54 and the whole technical block it came from, which on the same fetch also served a corrupt pivot table. An “intraday pivot block presented as daily”, caught by the one-line check: for classic pivots R1 − S1 = High − Low, and the published block's R1 − S1 is 10.50 points, which would assert a 10.5-point daily range on the S&P 500 — and it is centred on Friday's close. A Nasdaq Composite read of 26,820 / −0.92%, identified as a 12:16 ET midday snapshot by a quote page carrying the identical figure stamped “Market Open”. Every one of five 24-hour crypto signs from one aggregator, which printed all five POSITIVE on a session all five assets fell — the sign-inversion defect has now recurred and the levels-from-one-source, signs-from-another protocol is validated twice. A Sunday-stamped AUD/USD row and a Sunday-stamped USD/KRW row. A GBP/USD row showing current equal to previous close with a −0.0002 change. A USD/CHF page printing two different levels for the same pair. A Stoxx 600 quote of 638.68 with a +0.00% change field — an uncomputed field, not a real change, and contradicted by the intraday wrap it would have to agree with. An index level of 51,498 for the Dow from a page whose own prose gives 51,484 and reconciles. Vendor change columns for a seventh consecutive edition. And five sell-side attributions on dateline grounds — see below.
Not published
The Nasdaq Composite and Russell 2000 Monday closes. The Stoxx 600 Monday close (fourth failure). Any OAT–Bund mark for 28 September. Any twelve-pair FX close set — see the §03 caption; only the DXY is published as a close. Monday's VIX close, VIX3M, IVTS and the contango day count. Monday's SKEW, put/call, percentage above the 200-day, McClellan and the highs/lows series — the entire sentiment stack failed to update, so where figures appear they carry their own 24 or 25 September date. ASX 200 volume, breadth and the A-VIX (four routes; the A-VIX is the top gap on decision day). Any daily pivot levels from any source. Monday's Korean single-stock prices (the available figures are explicitly “earlier in the week” and contradict the session). Japanese single-stock prices, a seventeenth consecutive edition, hence no Japanese relative value. Any credit observation after 24 September. SoftBank secondary levels or a new-issue concession, a sixth consecutive failure across five sources — this should go to a dealer run or be dropped. Lipper weekly flows, a fifth consecutive failure on both routes — recommend formally dropping it as was done with the percentage above the 50-day. Any fresh CTA or vol-control estimate. The Australian monthly CPI consensus from a forecaster survey (a 4.1% y/y figure on one calendar is single-sourced and a very large step from 3.5%). The Cotality HVI release date. The national euro-area flash CPI dates, times and consensus across three routes. NZD/USD, any RBNZ commentary, KOSDAQ, KRW, India VIX, the Jakarta index level, EU gas storage (~70.9% appears under a September 2024 dateline elsewhere, so one source may be recycling the other's vintage), BNB's 24-hour sign, OPEC press releases (HTTP 402), EIA weekly inventories, the Riksbank's 4 November date, and the next-meeting dates for Banxico, Brazil, South Africa, Indonesia, Hungary, Turkey, Taiwan and the PBoC. The named FOMC voting roster remains structurally unavailable from the statement and is re-tasked to the 16 September minutes, due ~6–7 October.
Traps caught
A moving average wrong by 474 points, caught by range arithmetic rather than by a second source — and it is this desk's own published figure, which is the point. The intraday-pivot trap, for a second consecutive edition, caught by the R1 − S1 identity. A contract-month trap on Brent: the vendor's history header reads “October 2026 (LCOZ6)”, which is internally contradictory since Z denotes December, while its live page reads “Dec 26” — the live label is the accurate one, and reading it is what produced the correct settle. Phantom Sunday rows present on both the Brent and WTI dated tables (27 September), which corrupt every printed change on those pages. A quote page serving Friday's data under a live “Real-time” header — twice, on a major-indices page and on the Stoxx 600. A tracker serving Friday's 90.0% RBA probability under a Monday timestamp, caught because the site-wide stamp reads 25 September and the number is identical. A central-bank tracker attaching a probability to the WRONG BoJ meeting date (28 October, neither the 29th start nor the 30th decision). A vendor page whose quote box and narrative disagree, on four separate markets (Hong Kong by a chain-breaking 16 points, Italy, Spain, Indonesia). An index card internally corrupt enough to report the Hang Seng as “3,145.39”. A weekly sector table that would read as daily — deliberately avoided by using the evening wrap rather than the live blog. An ETF tracker still serving Friday's close under a current header (VTWO), which is why the Russell is not published. The dateline trap, five times, including one item dated 28 September 2023 whose matching day-and-month is exactly how it works. And Maysan (Iraq) and Mayun (Yemen) held apart for a ninth consecutive edition — the two are ~2,000km apart and they are the two ends of the same Hormuz bypass, which is the commercial reason the distinction matters: the East–West pipeline moves Saudi crude around Hormuz to the Red Sea and Bab al-Mandeb is the exit from that route. The pipeline has restarted and is running at 100% of its 7.0 mbd capacity with zero redundancy; Bab al-Mandeb has not reopened. The holiday check held and caught that Taiwan was shut.
Two new source warnings worth recording
A German financial portal's “stoxx_600” path serves the STOXX Europe 600 HEALTH CARE sub-index, not the headline index — wrong instrument as well as stale, and a live trap for any future run. And two mastheads publishing byte-identical syndicated market reviews remain one source, not two — that feed was previously caught printing a wrong Russell close and a wrong gold price, and it is the origin of one side of today's unresolved 30-year conflict, so its 10-year and 30-year figures are treated as single-sourced despite appearing in two places. On the other side of the ledger, a general macro-data vendor whose change columns this desk routinely discards proved genuinely useful as an independent LEVEL source and was what broke the 10-year tie — the “take levels, recompute changes” rule is about its change fields, not its levels.
Tomorrow's first verification targets
The RBA statement and press conference at 14:30 and 15:30 today, which settle V004 and V006 and bear on V023 — and, because the Bank's releases archive is four months stale, the statement text may have to come from wire copy. Then, in order: Treasury's official 28 September par row (due ~08:00 AEST today), which settles the 2-year basis, the 30-year range and both curve measures; AP's tabulation for 28 September, which should be retrievable at the usual syndication URLs once indexation catches up, and with it the Nasdaq Composite and Russell closes; any IG/HY/CCC observation after 24 September, with V017 38bp from confirmation and two dark sessions to fill; a settled Fed Rate Monitor read, to establish whether Monday's 71.2% held through the 17:00 ET futures settle; the OAT–Bund dedicated series' 28 September row; LME copper's 28 September official settle, which posts today; the Nikkei's 28 September row from the exchange archive, to convert a corroborated close into a verified one; the A-VIX; the Australian monthly CPI consensus before 11:30 tomorrow; the national euro-area flash CPI schedule before Wednesday's European open; the payrolls consensus, to fix the surprise threshold; and the French PLF date against a government communiqué, since a Thursday Conseil des ministres is unusual and the collision with the OAT auction is the tradeable part.
United States, the Fed and positioning
- US Treasury — official daily par yield curve, September 2026 (confirmed still at 25 Sep)
- US Treasury — tentative auction schedule · tentative buyback schedule Q3 2026 · expanded buybacks from 9 September
- Investing.com — Fed Rate Monitor (stamped “Sep 28, 2026 03:25PM EDT”; settled refresh had not fired at 17:10 ET)
- Federal Reserve — Board members · FOMC members — both list Cook; this closes the standing item
- Federal Reserve — 2026 speeches · FOMC calendar
- BLS — October 2026 release schedule — September CPI on 14 October
- Cleveland Fed — inflation nowcasting (September core PCE 3.49% y/y)
- StockAnalysis — SPY (settled 28 Sep close, $765.61, −0.74%) · QQQ · DIA · IWM (3:49 PM capture) · VTWO (serving Friday)
- Trading Economics — S&P 500 · Dow · US 10-year yield (the independent read that broke the tie)
- Investing.com — US government bonds · YCharts — 10-year Treasury
- CFTC — Traders in Financial Futures · CME legacy · COMEX · NYMEX · ICE Futures U.S. · release schedule
- WallStreetNumbers — S&P 500 moving averages and Financhill — SPX technicals — the two sources that corrected the 200-day
- CBOE SKEW · put/call · VIX term structure · highs/lows and McClellan · % above the 200-day
- ZeroGEX · GEXmetrix · QuantWheel · InsiderFinance — the four gamma reads
- ICI — combined flows · ICI — money market funds · AAII sentiment survey
- FactSet — Earnings Insight, 25 September 2026
- Micron — Q4 results date (the correction) · Motley Fool — 28 Sep midday
- Polymarket — October Fed decision (the versioned URL; the bare slug is a trap)
Rates, FX and central banks
- RBA — Coming Up (14:30 and 15:30 issuer-verified) · speeches list (no commentary since 22 Sep) · Iain Ross — “A Wage-price Spiral: What are the Chances?” · MR 2026-12, 5 May (the last issuer-confirmable decision)
- centralbank.watch — RBA (90.0%, but see the timestamp warning) · ECB · calendar
- Bank of Japan — MPM schedule (29–30 October, decision the 30th; Summary of Opinions 1 October) · release calendar
- Bank of England — upcoming events · Ramsden — “Quantitative tightening: the next chapter”, 28 September
- RBNZ — October Monetary Policy Review (28 October, a Review not an MPS) · Bank of Canada — 2 September
- Norges Bank — policy-rate decisions · Bank of Korea — 2026 meeting dates
- Econostream — ECB speak and the weekly tone meter (Schnabel drove the dovish drop)
- Trading Economics — Bund · OAT · BTP · gilts · JGBs · ACGBs
- ideal-investisseur — OAT–Bund dedicated series (last row 25 Sep at 105.4bp)
- Trading Economics — dollar index (the only FX close published) · investingLive — Monday indicative opens (the continuity test that passed)
- Tradingpedia — PBoC fix, 28 September (numbers used, prose rejected)
- Reuters via Investing.com — Sakurai, a FORMER board member (the correction)
Australia, New Zealand and Asia
- Market Index — ASX evening wrap, 28 September 16:36 AEST (in-body dateline verified; the sector table and movers)
- Investing.com AU — ASX 200 history (chain verified) · ASX 200 futures, APZ2026 (quotable again, stamped 26/09)
- ABC — markets live, 28 September (narrative and the 10-year; its index print was rejected)
- ABS — future releases (monthly CPI Wed 30 Sep 11:30 confirmed) · consumer sentiment · business confidence
- Australian Broker — auction clearance, Cotality data · MacroBusiness · fixed-rate repricing, Canstar data
- Cotality — dwelling values, three months to August
- Nikkei — official index archive (last row 25 Sep; 28 Sep not yet posted) · News On Japan — 28 September session
- Nippon/Jiji — Tankan consensus +26 (twelve-institution panel)
- Trading Economics — China · Hong Kong (narrative used, quote box rejected) · Hang Seng Tech history (the gap closed)
- Korea Times — economy · Korea JoongAng Daily (the open that also ties to the prior)
- HDFC Sky — Indian market close, 28 September · Stockpil
- Trading Economics — NZX 50 · interest.co.nz (the 3pm read that overstated the session)
- Asian market holidays (Taiwan shut 25 and 28 Sep; Golden Week 1–7 Oct) · JPX calendar
Europe and geopolitics
- Trading Economics — DAX · CAC · IBEX · FTSE MIB · FTSE 100 — country pages reconcile; the euro-area aggregate does not
- finanzen.ch — SMI close, 28 September (states its own prior; the read that won the reconciliation test)
- Investing.com — Stoxx 600 history (no 28 Sep row) · MarketScreener (+0.00% uncomputed field)
- ifo — business climate 89.9 · NIM — consumer climate −30.6 (prior −26.8 revised)
- MoneyVox, 28 September — PLF 2027 “jeudi” · Banque des Territoires — “le 1er octobre” · francebudget.fr (gives no date) · AFT — next long-term OAT auction, Thursday 1 October
- BDI — 2027 federal budget (the two bases, kept apart)
- Al Jazeera — Trump rejects the roadmap, 26 September · Axios, 27 September · Newsquawk headlines
- Al Jazeera — the pipeline drones launched from IRAQ · Houthis seize MAYUN island, YEMEN — two theatres, ~2,000km apart, held apart for a ninth edition
- Axios — the US–China tariff deal, 26 September · Fox News · Epoch Times
- ForexFactory calendar — event list, times, consensus and priors; its impact column is permanently broken and every H/M/L in §10 is this desk's own
Commodities, credit and digital assets
- OilPrice — Brent futures curve by contract (Nov-26 $107.71; the page that resolved the contract-month conflict) · Investing.com — WTI history · WTI live, contract label CL1!
- Kitco — gold, 16:12 ET · Trading Economics — gold · USAGOLD — three of the five routes on the break
- Westmetall — LME copper cash and 3M (last row 25 Sep) · copper · iron ore · lithium · Henry Hub · TTF
- FRED — IG OAS · HY OAS · CCC OAS — all three still last-observed 24 September · ConvexTrade — HY corroboration
- TankerMap — Hormuz transits (~1/day seven-day average; its “~21/day” is static boilerplate) · gCaptain · The Deep Draft — Gibson VLCC rates above $1.2m/day
- CoinGecko (levels only — every 24h sign was inverted) · CoinDesk (signs and percentages) · Farside — BTC ETF flows · ETH · Coinalyze — derivatives (per-asset OI) · Polymarket — CLARITY Act
- IEA news — no emergency stock release announced 21–28 September