In September and again in October I put numbers on a pullback. Here is how each call scored, when the 2018 and 2022 drops really came, and who should turn conservative this week.
I was right about the ingredients and wrong about the clock. Here is every call, graded, and who should turn conservative this week.
The S&P 500 closed Wednesday at 7,743.98, by our own pull. That is a third straight gain, 0.96% under its Oct. 6 record. The Fed held rates at 3.75% to 4%, and two officials voted to raise them.
The letter was written Wednesday night. Thursday brought the week’s big earnings, and they cut both ways. By our own pull, after the close:
Alphabet missed. Earnings per share of $2.12 against $3.01 expected. Its free cash flow went to about minus $9 billion, the same turn Meta made the week before. The stock fell 12.7%, its worst day since it listed in 2004. Amazon missed too, by a little on revenue ($201.9 billion against $202.3 billion) and by 4% on earnings per share ($1.89 against $1.96), and fell about 10% after the close. Microsoft beat. Azure grew 48%.
The new fact is the spending, not the profits. Microsoft and Alphabet both spent less on AI gear in the quarter than analysts had penciled in, and Microsoft cut its full-year plan to $170 billion. The suppliers felt it at once: Micron fell 5%, SanDisk 5%, Intel 4%, CoreWeave 5%. Micron had risen 6% the day before. The two names in that list we hold at size, Micron (23 books, part of the North Star’s first third) and CoreWeave (one sleeve), are the ones the capex signal hits first.
What it does to the scorecard: nothing yet. The S&P 500 fund fell only 0.4% on the day. The equal-weight fund rose 1.2%. That is the Oct. 22 shape again: the money moved inside the market, it did not leave. The VIX closed at 16.74 and vol-of-vol at 94.5. By the fund’s close the index sits about 1.3% under its record, with three sessions to the vote. Two misses in one week are not an earnings recession, and FactSet still has the quarter’s profits up 29.5%. But “earnings are strong” now carries a footnote: the biggest spenders are earning less than hoped and spending less than planned, and the companies that sell to them are priced for the old plan.
Nothing is bought or trimmed on a Thursday. The reserve decision stays where the letter put it, in writing after Nov. 3. The one thing that moved is the question it has to answer.
Prices and results are the desk’s own pull after the close on Oct. 29 (Cboe delayed quotes, company releases and wire reports), not from the papers. Amazon’s after-hours price can change by the open.
Seven Calls, Graded
In September I wrote down odds, levels and a date. In October I cut the odds and wrote a test that would prove me wrong. It only counts if I come back and grade it.
The pattern is plain. I was right about the ingredients and wrong about the clock. Inflation came back, the Fed leaned toward raising, and the crowd was all in. Since the October Letter, none of it pushed the index more than 2.5% below its record. The plan did its job anyway. Nothing was bought at the top, two thirds of the reserve are intact, and the levels stand. The calls that leaned on a date were the ones that missed.
When the Drops Really Came
2022: the vote was Nov. 8. The bottom had already come on Oct. 12, 19 trading days before it. From that low to election day the fund rose 7.1%, and a year after the vote it was up 16.3% (our own data).
2018: the vote was Nov. 6. The drop came after it: the Fed raised on Dec. 19 and the low was Dec. 24.
So the midterm set the timing in neither year. The Fed did, both times. Since 1950 the S&P 500 has been higher twelve months after every midterm vote, by the published record. That is a pattern, not a promise, and it says nothing about the weeks right before the vote, which were the hard ones in 2022.
Both drops started from a record, with the Fed raising or about to. Today the index sits under 1% from its record, the Fed just held with two votes for a hike, and the crowd prices a December hike near 80%. The setup rhymes. The timing never announces itself. Jan. 3, 2022 felt like a great day. Going conservative in December was late both times. In 2022 the bottom had come in October. In 2018 the worst day to sell was Christmas Eve, and the record was back by April. So the switch is set by your own clock, before the drop, not by the headlines during it.
By Your Own Clock, Not the Headlines
The question on my desk is when to switch people to more conservative, and whether to do it now. My answer, by situation, as general rules.
Yes, this week, for the short clock. That is anyone drawing income with under two years of spending outside stocks. Or anyone with a known bill in two to three years, a house, tuition, a wedding. Move that money now, in steps, while the index sits within 1% of its record. Selling near a high is the cheapest an umbrella ever gets. A 13-week Treasury bill auctioned at 4.050% this week.
Yes, a rebalance, for the five-years-out. Put the mix back to the target you set at the start. After two years of tech gains, the stock slice has likely drifted up. That is a conservative move with a rule behind it, not a forecast.
No, for savers still adding with ten or more years to go. Keep buying. In 2022 the fund took 23 months to get back. A saver who kept buying bought the dip without trying. One who switched had to guess the way back in.
Not a blanket switch for everyone. Moving everyone to conservative is a market call. It has to be right twice, out now and back in later. I have already been early once this fall.
So the model books' reserve can go to work next week while short-clock money steps out. The reserve is long-clock money that was waiting for a price or a date. The umbrella is money you need soon.
Crowd Prices, With a Timestamp
Our own Polymarket pull at 7:31 p.m. PT Wednesday, crowd prices and not forecasts. A December hike 78.5%. Another hike in 2026 at all 79.5%. A U.S. recession by the end of 2026 6.5%, from 7% on Oct. 16. The S&P 500 touching 7,200 by Dec. 31 31.5%, and touching 7,400 53%. News reports put CME futures near 82.6% on a December hike. Volatility, by our own Cboe pull at the close. The VIX is 16.91, from 19.99 on Oct. 16. The curve slopes up, from 15.60 at nine days to 21.24 at six months, a calm base. Vol-of-vol 96.61. SKEW, the price of crash insurance, 148.74, from 168.59. We carry crowd odds as a price with a timestamp, not a forecast.
Count the years of spending you hold outside stocks. If it is under two, fix that this week, in steps, from the stock slice.
Check the mix against the target you set. If it drifted up after two strong years, rebalance back. That is a rule, not a forecast.
Check what your cash earns. The average bank money-market account pays 0.45%. A rate that starts with a zero loses to 3.7% inflation. Bills pay about 4%.
That is where the predictions stand. The ingredients were real. The clock was mine, and it was wrong. The next dates are the vote on Nov. 3 and the Fed's December meeting. I'll put the reserve decision in writing after the vote and grade it the same way.
The sky is still gray, and I still don't know the day it rains. The umbrella isn't a forecast. It is years of spending outside stocks, a mix back at its target, and cash that earns a four. If you'd like to check all three with me while the sky is still mostly dry, fifteen minutes with your statement is usually enough.
Two companies supply 34% of the S&P 500's profit growth. Count the names your index fund leans on. — Nvidia and Micron supply 34% of the S&P 500's expected 30% profit jump, and ten companies about two-thirds.
Microsoft offered buyouts to 60,000 people, a quarter of its staff. Here’s how to weigh one. — Microsoft offered voluntary buyouts to about 60,000 employees, a quarter of its staff and the largest in its history.
A money reporter knows $7 a week is $364 a year. She bought her 5-year-old the stuffie anyway. — A reporter who knows $7 a week is $364 a year buys her 5-year-old the plush toy anyway.
A compliment on a client’s Rolex helped land him. A vintage Datejust costs about a third of new. — U.S. women’s watch sales on Chrono24 more than doubled in three years, and a 1984 Datejust cost about $6,000 against $18,000 new.
Half of private clubs have wait lists again. The joining fee at the family-friendly ones is $60,000. — Family-focused clubs charge a median joining fee of $60,000, double golf-only clubs, and half of private clubs have wait lists.
No. 188 · Part 1 · the weekend paper · No. 189 · Part 2 · the Monday to Wednesday papers · The October Letter
Which clock is yours?
The sky is still gray, and I still don't know the day it rains. The umbrella isn't a forecast. It is years of spending outside stocks, a mix back at its target, and cash that earns a four. If you'd like to check all three with me while the sky is still mostly dry, fifteen minutes with your statement is usually enough.