Week 43 — Oct 19 – Oct 23, 2026

The index is the least informative number of the week. The S&P fell 0.43%; underneath it, long Treasuries rose 1.61%, energy 1.60%, health care 1.45%, the equal-weighted index 1.14%, commodities 1.07%, utilities 0.83%, staples 0.73% and financials 0.68% — while megacaps fell 1.70%, emerging markets 1.34%, the Nasdaq 0.76% and gold 4.03%. The distance between the best and the worst of those is nearly six points in a week the headline called flat. This is the fourth week in five in which the average stock has beaten the index, and the rotation now has a clear shape: out of the complex that drove the year, into the parts of the market that pay you to wait.

It is worth saying plainly that the book is on the wrong side of this. The US sleeve is held a third each in the S&P, momentum and megacap growth, and it holds nothing at all in staples, utilities, health care or financials. That is not an oversight to be corrected this week — the engine allocates across sleeves and has no sector dimension, by construction — but a reader is entitled to know that the thing working best in October is a thing this model cannot express. Whether it should be able to is registered as a research question and is not decided.

Thursday was a position being reduced, not a forecast being changed. Its trigger was earnings, not data: Meta and Tesla reported on Wednesday evening, both beat on the headline, and both were sold on what they intend to spend — Meta raising the floor of its 2026 capital-expenditure guidance to $135–145bn, Tesla cutting its own — after which the AI complex fell from Thursday's open. In one session megacaps fell 3.35%, momentum 3.67%, gold 2.63%, copper miners 6.27% and emerging markets 2.91%, while the equal-weighted index rose 0.73%. Four things argue that this was de-risking rather than a call on growth. Jobless claims printed 194,000 against the 200,000 expected. The October composite business survey came in at 57.6, with manufacturing beating at 56.0. The odds on a quarter-point rise at the December meeting eased only from 78.5% to 72.5% — worth about a basis and a half on a two-year that fell sixteen on the day. And high-yield spreads widened while Treasuries rallied hard, which is not how credit behaves when the news is good.

Friday's data then argued against the rally rather than for it. The final October Michigan survey revised both of its inflation-expectation measures up: one-year to 4.7% from 4.6%, and the five-year to 3.5% from 3.4%. Long-run expectations are the series the committee says it cannot let drift, and they drifted in the last scheduled reading before it meets. Consumer sentiment itself came in at 48.0 against 46.3 expected, so this was not a weak survey — it was a survey in which people feel slightly better and expect slightly higher prices.

The curve rallied across the board and the move was entirely in real yields. Two-year 4.74% to 4.69%, five-year 4.89% to 4.82%, ten-year 5.09% to 5.00%, thirty-year 5.43% to 5.33% — a bull flattening of four basis points in 2s10s. Over the same week ten-year breakevens went 2.35% to 2.34% while the ten-year real yield fell from 2.74% to 2.64%. The market revised what it expects to be paid for lending, not what it expects prices to do. That is the second week running with the same decomposition.

Credit widened again, and again only at the weak end. High yield went 3.31% to 3.40%, nine basis points, while the investment-grade Baa spread was unchanged at 1.46%. This is the fourth consecutive week of that split, and the cumulative move is now 67 basis points on high yield since 23 September against essentially nothing on Baa. A credit market repricing with no funding stress behind it is usually about the borrowers rather than the system — slower, and easier to mistake for noise.

The plumbing is where the week's most durable news is. Bank reserves fell $59.5bn to $2.963 trillion, below three trillion for the first time, and the overnight reverse-repo facility that absorbed the balance-sheet run-off for two years holds $0.20bn. It is empty in all but name, so from here the run-off comes out of bank reserves directly. On Thursday the secured overnight rate printed level with the rate paid on reserves, after weeks two to three basis points below. One print is one print, and the twenty-one-day average that arms the funding leg is still at −1.6bp against the +5bp that trips it. But it is the one trigger that soft growth and decelerating inflation cannot arm on their own, and it is the only one of the three that moved toward us this week.

The duration gate shut, and the rally is what shut it. The ten-year at 5.00% is now 0.25pp cheap to fair value, exactly at the threshold where long bonds begin to scale in, against 0.33pp on Tuesday. Long bonds go to zero and the defensive dollar sits in short-dated bills instead. The irony is worth naming rather than hiding: long Treasuries were the best asset in the book this week, and it is precisely that rally which has priced them out of it.

The outside view, and one of it disagrees with us. Torsten Slok's Daily Spark spent four of six sessions this week on the same argument: that the AI trade's earnings foundation is narrower than its price. His 24 October note reports revenue per worker and profit margins rising for the Mag 7 and flat for the S&P 493, with no sign yet of AI lifting productivity, and concludes that mainstream corporate America has begun treating token spending as a cost centre rather than a source of growth. We do not need to accept the conclusion to notice that the market spent the week acting on it: megacaps fell 1.70% while the average stock rose 1.14%. Read that way, the rotation this note has called positioning may have a fundamental reason underneath it, and the two readings are not exclusive.

His 23 October note is a direct challenge to this model's monetary reading, and it deserves to be said plainly. Under the title Monetary Policy Is Not Restrictive, he cites the Richmond Fed's third-quarter CFO survey: 19% of firms report that financing costs or access to credit constrain their investment and spending plans, and 81% report no such constraint. This gauge's monetary cycle reads −0.85, which is to say tight. Both can be true — the monetary composite measures the policy rate, the curve and real yields against their own history, not whether firms feel constrained — but a reader is entitled to know that the people actually borrowing do not report the condition our cycle score describes. If corporate America is right, the restrictive reading is a statement about the price of money and not about its availability, and the second is what usually breaks things.

Michael Howell of CrossBorder Capital, interviewed on 23 October, is bullish on gold and cautious about the path to it — his framing is that gold may face another correction before the next major liquidity upswing, with China's gold demand and the coming debt-refinancing wall as the longer arc. Gold fell 4.03% this week, its worst since early June, and the book holds 12%. We take the near-term caution as corroboration of what the tape did rather than as a forecast, and note that his liquidity argument and this model's liquidity cycle are measuring related things on different clocks.

Mark Meldrum's 23 October review puts the midterm elections on Wednesday week as the risk event for bonds and the dollar, and flags the divergence between crude and natural gas. On the oil half we can check our own series rather than repeat his, and the check goes his way: the authoritative Cushing spot price last printed $107.89 on 20 October, which is the 96th percentile of its past year of observations and within six per cent of April's $114.58 high, against a December low of $55.44. Crude is not merely elevated; it is trading near the top of its range while the inflation leg reads it year on year. His captions are machine-translated and we have not quoted him, only his subject.