Two readers wrote back to the August letter with the only fair question — fine, when? — and the September letter answered it with dates: a window from Sept. 11 to Nov. 3, a center of gravity in the fortnight after the Fed, and 60/40 odds that the S&P 500 gives back at least 7% from its Aug. 13 record before the vote. It also wrote down the specific way it expected to be wrong. That was two weeks ago. Fed week was the first exam, and the honest thing to do on a Saturday is to sit the paper down and mark it.
So this is a grading exercise, in five parts. What the letter said would happen and what did. The pattern in the numbers that keeps showing up. When the drop comes, if it comes, and the scenario in which it does not. Why a 10-year Treasury at 5% is the whole story and not a side plot. And the Risk Atlas — the eight things that can actually go wrong, published in July with a defense assigned to each — re-graded square by square. The meter at the top is the summary. The rest is the working.
1. Proof of the reasoning: the letter’s calls, marked
A thesis without dates is an opinion. The September letter had nine dates and six reasons, and each of them can be checked against a newspaper. Here is the ledger.
| What the letter said | What happened | Grade |
|---|---|---|
| Sept. 11 inflation: hot, and hike odds go past 80% and the 2-year through 4.5%; cool, and “my 60 drops toward 45.” Named as the specific way the thesis would be wrong. | Core +0.3% against the 0.1% test. Hike odds went to 95% by decision day. The 2-year closed Friday at 4.741%, its highest since July 2024. | Confirmed |
| Sept. 16 Fed: a hike would be the first this year and “the trigger the bear has waited on since March.” | Raised 12–0 to 3.75–4%. Sixteen of 18 officials see at least one more. Futures put ~50% on October. | Confirmed |
| The Fed is not coming to help — the one reason the letter would keep if it had to throw away the other five. Prediction markets at 93% on zero cuts in 2026. | Friday’s Journal prints the Polymarket market itself: 95% on zero cuts this year, 2% on one, $53 million wagered. | Confirmed |
| The bond market is voting. 10-year 4.783%; prediction markets 42% on 5% before 2027; Norway cutting government bonds. | The 10-year settled above 5% on Sept. 16, the first time since 2007, and closed the week at 4.995%. The Bank of Japan raised to 1.25%, the highest since 1995, giving the largest foreign holder a reason to buy fewer. | Confirmed — six weeks early |
| The fuel and the consumer: diesel $5.85, wages behind prices, the average stock underneath the index — 90 new lows against 47 new highs with the index 1% from a record. | Consumer prices +3.4%; wages behind for five months; 30-year mortgage 6.95%. New lows against new highs: 151–99 Thursday, 182–29 Friday. | Confirmed — and widening |
| Sept. 18 expiry — the plumbing date; the two weeks after are on average the weakest fortnight of the year. | Passed quietly: the S&P moved 12 points. The fortnight starts Monday. | Pending |
| The drop: base case 6–10% from the Aug. 13 high, the low between late September and mid-October, bought into the vote. | The S&P 500 is 1.9% below 7,798.99. The VIX closed at 14.81. Nothing has moved. | Pending |
| The bear’s veto: while the curve stays in contango and vol-of-vol stays under 90 through expiry, buy no hedges. | Honored. No hedge was bought. The two inputs were not measurable this week and the veto was treated as standing, not lifted. | Honored |
Five confirmed, two pending, none wrong. That is a better grade than most letters get, and it is also the least useful kind of right: every reason showed up and the thing the reasons were supposed to cause has not. A thesis that is right about the causes and wrong about the effect is either early or wrong, and the market has not yet said which.
2. The pattern
Three weeks of the Journal’s trading diary, read in order, is the pattern. On Sept. 4, with the index one percent from a record, the NYSE logged 90 new 52-week lows against 47 new highs. On Sept. 17, the relief-rally day when two-thirds of issues advanced, it was 151 lows to 99 highs. On Friday, on a session the headlines called quiet, 182 lows to 29 highs. Six to one. The index is 1.9% from its high; the utilities average closed Friday at its 52-week low; the S&P real-estate, financial and transport sectors each lost more than 2% on the week. The Value Line index of the typical stock is up 4.2% this year against 11.8% for the S&P 500 and 68% for the semiconductor index.
Four other things rhyme with it. The 2-year yield made a two-year high on a day the VIX made a one-month low: the bond market is pricing the next hike and the options market is pricing nothing, which is exactly the “tail bid, body asleep” configuration the letter described. The Dow has fallen three straight weeks while the Nasdaq rose, the same divergence told through the two averages. Friday’s biggest percentage gainers were two-times-leveraged single-stock funds on bitcoin-treasury companies, up 27% to 33% in a session, while the broad tape fell: speculation at the edges, erosion in the middle. And the buyer base for Treasurys is thinning — Norway in September, Japan on Friday — at the moment the supply is about to grow.
None of these is a timing tool. Together they describe a market in which the strong cohort is getting stronger and narrower and everything else is already in a correction the index does not report. There are two ways that resolves. The cohort pulls the index down to meet the rest, which is the letter’s base case. Or the rest catches up, which requires a Fed that stops, and the Fed has said it has not.
3. When — and the scenario in which it never comes
The center of the window was always Sept. 16 to Oct. 14, and we are four days into it with the plumbing date behind us. What is left on the calendar, in order: the two weeks after expiry, historically the weakest fortnight of the year, beginning Monday; Micron’s earnings on Sept. 30, the memory shortage going on the record with a number attached; the September jobs report on Oct. 2; September inflation on Oct. 14, the same week the banks open third-quarter earnings and companies that carried $100 oil through the summer tell you what it cost them; the Fed again on Oct. 27–28, the same week the largest technology companies report; and Nov. 3. Anthropic’s $100 billion offering, which the letter counted as an October drain on liquidity, has moved to November, which puts it at the far edge of the window rather than removing it.
Here are the updated odds, and the reasoning is the ledger above. Base case, 45%, unchanged: a 6–10% pullback from the Aug. 13 high with the low between the last week of September and mid-October, bought into the vote; the dividend book and the utilities that have already fallen outperform on the way down; long bonds do not help. Bear case, 20%, up from 15%: two of its ingredients — the hike and a 5% 10-year — have arrived ahead of schedule, the breadth divergence is wider than it was, and the largest stock sale in history now lands in the week of the vote. No drop, 35%, down from 40%: the specific exit the letter named for this path, a cool inflation print on Sept. 11, closed when the print came in hot. The sum is 65/35 on a 7%-plus giveback before Nov. 3, from 60/40. Not a forecast. A budget, with the arithmetic shown.
Is there a scenario in which this does not happen? Yes, and it deserves its full hearing, because it has been right since March. Earnings: FactSet has third-quarter S&P 500 profit growth at 28.9%, the third straight quarter above 25%, and roughly 32% for the year. That is not a market that needs a miracle; at 20 times forward it needs the growth to arrive, and so far it has. The economy: unemployment 4.1%, household net worth $186 trillion, retail sales beating in August. The 1994 template — a hiking cycle into a strong labor market, a bond-market massacre, and a stock market whose worst stretch was under 10%. The volatility regime: the VIX has closed between 14 and 18 for most of the summer, and regimes like that have a long history of outlasting the people who bet against them; 2017 never fell 3% from a high. The market is already hedged: the crash-protection index has sat near 150 for months, and hedged markets grind rather than gap. The election-year levers: an administration with a $5,000 check on the table will pull every one of them between now and Nov. 3, and every one is good for stocks in the short run. And the oil curve: December 2027 crude at $73.51 says the market expects the strait to reopen, which would take the fuel reason off the board in a single headline.
There is a third possibility that is neither, and it is the one we think about most. The drop already happened, underneath the index, and the index never reports it. Six new lows for every new high is a correction in the average stock. If the cohort carrying the average keeps carrying it through November, the letter will have been wrong about the index and right about the market, and a plan that only knows the index will have missed the whole thing. The Keel and Midterm Dividend books are graded on the average stock for exactly this reason.
What would change our mind, written down: a new record close above 7,798.99 before Oct. 2, which would say the cohort can carry the index through the weakest fortnight and the base case is wrong on timing; a September core print of 0.1% or less on Oct. 14, which reopens the exit the Sept. 11 print closed; and a Fed on Oct. 28 that says one was enough. Any one of those moves the no-drop case back above 40. All three would move the meter under 4.
4. Treasurys at 5%: why, and why it is the whole story
The 10-year settled above 5% on Wednesday for the first time since July 2007 and closed the week at 4.995%. The 2-year closed at 4.741%, the highest since July 2024. The Bloomberg long-Treasury index yields 5.36%. There are five reasons, and they are all in this weekend’s paper.
The Fed. The target is 3.75–4% with roughly a 50% chance of another increase in October and 95% odds on no cut this year. The front end of the curve is the Fed’s next move, priced. Inflation. Consumer prices are up 3.4% and crude is $100; a 5% nominal yield is a 1.6% real one, which is not generous by the standards of any decade before 2009. Supply. A $1.2 trillion “dividend” check proposal on top of an already large deficit, in a market that has to be paid to absorb it. Demand. Norway’s fund asked to cut government bonds from 70% of its bond book to 50%; on Friday the Bank of Japan raised to 1.25%, the highest since 1995, and Japan — the largest foreign holder, already down from $1.2 trillion to $1.1 trillion — now has a 3% domestic bond to buy instead. The term premium. The extra yield investors demand to lend for ten years rather than roll bills, which had been near zero for a decade, is being rebuilt by all four of the above at once.
That is the arithmetic the letter said gets harder every week the 2-year sits above 4.3%, and it has now run out of room. It does not mean stocks fall on Monday; markets have run at a zero equity premium before, on the argument that earnings grow and coupons do not, and the 28.9% third quarter is that argument in its strongest form. It does mean that every dollar of return from here has to come from earnings growth that arrives on time, because there is no longer a valuation cushion to absorb a disappointment. It also explains something about the Treasury market that the headline “5%” obscures: a lender to the government is now paid as much as an owner of its largest companies, with a contract instead of a hope. That is why Japan’s pension fund can think about coming home, why Norway can cut its bond book without missing the income, and why the reserve in this desk’s books sits in bills and floating-rate paper at about 3.8% rather than in either stocks or long bonds.
What 5% does not do is make long bonds attractive to us. The Bloomberg long index has lost 4% over the past year at a 5.36% yield; that is what duration does while the buyer base is thinning. Short paper earns nearly the same coupon with none of the price risk, and it resets upward if October delivers.
5. The Risk Atlas, re-graded
In July this desk published a map of the eight things that can actually go wrong, with what each did that week, which sleeve of the Capital Wealth Growth Portfolio is assigned to it, and an honest grade. Two squares were thin: the bond sleeve, deliberately at half the audit’s recommendation, and behavior. Here is the same map after Fed week.
| Risk | July reading | This week | Assigned to | July → now |
|---|---|---|---|---|
| Market | Fed held; Dow −1,153 in a day | Fed raised; the S&P is 1.9% from its record; the Dow fell three straight weeks; VIX 14.81 | Sizing; a reserve now at 15% in the Keel and Midterm Dividend books, from 0.5% in July; a written entry rule at −8% and −12% | Covered → Covered, stronger |
| Rate | 30-year 5.228%, highest since 2007 | 10-year 4.995% after a settle above 5%; 2-year 4.741%; long-Treasury index 5.36%; the buyer base thinning | No long duration anywhere in the books; bills and floating-rate paper reset with the Fed | Thin → Thin by design. The price risk is covered because we own no duration. The risk that moved is to the equity multiple: a zero premium over the bond. That square is now the live one. |
| Inflation | Stuck near 3%; three dissents pointing up | 3.4%; wages behind for five months; the Fed hiking; a year for it to work | Energy sleeve (XOM +36%, CVX +37%, COP +41% YTD) and dividends that grow | Covered → Covered |
| Geopolitical | WTI’s $10 round trip in five sessions | Day 200 of the war; a ship hit in the strait Friday; crude $100.30 and falling; Dec. 2027 at $73.51 | Chevron, Exxon, ConocoPhillips at weight; the defense sleeve; Valero’s written tripwire | Covered → Covered, and too expensive to add |
| Credit | Nvidia backstopping ~$250 billion of OpenAI financing | Turkey liquidating 131 funds; $349 billion queued to exit private equity; high yield at 6.85% | The avoid list: nothing priced by its own manager, nothing that borrows to buy what it holds | Covered → Covered, tested |
| Concentration | SK Hynix −10% on a record quarter | S&P +11.8% vs. the typical stock +4.2%; semis +68%; 182 lows to 29 highs; the index is the concentration | Sleeves; the equal-weight fund first on the reserve’s list; largest single position 3.5% | Partial → Partial. Honest grade: the offset is in place and it is small relative to the index’s own bet |
| Event | Humana, Boston Scientific, J&J’s $5.5 billion talc deal | Buffett’s exit; Xi’s health and a summit unconfirmed; 90 GLP-1 suits; a $100 billion IPO moved to November | Position caps; no single outcome large enough to change the year | Covered → Covered |
| Behavioral | Gold slept through a 1,153-point day | The published rule held through a hot print, a hike, a 5% settle and a 1.14% relief rally: nothing bought in three sessions, no hedge bought under the veto | Written rules, pre-committed and dated, graded in public | The hard one → Passing, so far. Three sessions is not a record; it is a start |
Six covered, one partial, one thin by design: the same tally as July. What changed is where the thin square’s risk lives. In July it was in bond prices, and we covered it by refusing to own duration. This week it is in stock multiples — a zero premium over the bond — and the cover for that is the reserve, the dividend book and the entry rule, all of which were rebuilt between the two readings. The behavioral square is the one to keep watching. Every rule is easy to follow in the week nothing happens. The fortnight that starts Monday is the one it was written for.
Where the meter sits, and what it says to do
6.1 of 10, up from 4.3 the day the letter was written. Every ingredient is on the board: the reason at nine, the calendar at seven, the breadth at eight. The price is at two and the hedging at two, because the index has not moved and the options market has not asked it to. In plain language: the fuel is loaded and there is no spark. Loaded is not the same as burning, and the difference is the whole of what to do on Monday.
The reserve stays a reserve; a third of it goes to work at an 8% drawdown from the Aug. 13 high, roughly 7,175 on the S&P, another third at 12%, about 6,863, and the last third the week of the vote regardless, because it was always November’s shopping money. The shopping list is the one written on a Sunday in September: the equal-weight index first, then the power builders, then the memory names after Micron reports. No hedges while the veto stands. No duration while the 10-year is above the earnings yield. The energy sleeve held, not chased. And a name that cuts its guidance in October gets sized down that day, on the rule, which is the only part of this letter that requires nobody to be right about anything.
For a household the four checks have not changed since the letter, and the week made each of them worth more: how much duration is in the bond fund you did not choose; twelve months of expenses in cash, at 3.8% rather than 0.44%; the growth fund that turned out to be a momentum fund; and the beneficiary pages. None of them needs the meter to be right. If the drop comes, they are the reason it is survivable. If it never comes, they are the reason nobody had to guess.
December can grade the rest.