Exit Liquidity
The worst week of the year ended with three documents that do not agree. The tape says the S&P 500 has fallen 9.95% in three months and closed Friday at 6,910.39. The seasonal literature says a holiday rally is due. And the flow data says the retail investor, the crowd the rally story depends on, spent the week before the break buying protection at a record pace and almost nothing else. Monday’s session will grade which of the three was the record.
Friday’s board
| Index | Close | Change |
|---|---|---|
| S&P 500 | 6,910.39 | −3.70% |
| Nasdaq Composite | 22,617.73 | −5.28% |
| Dow Jones Industrial | 48,147.33 | −1.93% |
| FTSE 100 | — | −1.02% |
| DAX | — | −2.60% |
| Hang Seng | — | −3.79% |
The S&P fell 147 points, 2%, to 7,175 — a fourth consecutive losing session, with more than 90% of index stocks finishing red. Nothing in the week’s record shows a bid arriving at any point during the fall.
The week, as printed
Thursday, November 19, the S&P 500 fell 147 points, 2%, to close at 7,175 — its fourth consecutive losing session, with more than 90% of index stocks finishing red. The VIX rose 15.8% that day to 28.44, having sat at 22.82 when Goldman’s seasonal note arrived. Friday was worse. The S&P 500 closed at 6,910.39, down 3.70%. The Nasdaq Composite closed at 22,617.73, down 5.28%. The Dow closed at 48,147.33, down 1.93%. The board was global and uniform: the FTSE 100 down 1.02%, the DAX down 2.60%, the Hang Seng down 3.79%. The VIX touched 40 intraday and closed at 37.66.
The three-month chart is the shape of the event. The index is down 9.95% over that span and sits roughly 12% below its 52-week high of 7,844.52. Months of sideways trade at the top, then a near-vertical descent. Nothing in the week’s record shows a bid arriving at any point during the fall. What the record shows instead is two kinds of paper being prepared for Monday: a story about why prices should rise, and a set of positions taken by people who expected them to fall.
The comforting promise
Every November, the seasonal case is republished. Goldman Sachs’ FICC and equities desk, in a note reported by TheStreet, lays out the historical pattern: from mid-November through December 31, the S&P 500 rises about 2.4% on average, in four phases — a mid-November dip, a Thanksgiving ramp of roughly 1.1% into early December, an early-December stall that surrenders about two-thirds of the gain, and a late-December rally. The final stretch is the reliable part: since 1990, the index has risen between December 15 and year-end in 27 of 36 years.
Goldman’s own data then discounts the promise under current conditions. In the 17 years since 1990 when the VIX stood at 18 or higher in mid-November, the average year-end gain shrank by nearly half, to 1.34%. The index rose in only 59% of those years, and the median path included a drawdown of 3%. The VIX was 22.82 when the note arrived and 28.44 by Thursday’s close — above the threshold, and rising. The one figure in the bulls’ favor is real: in 14 of those 17 high-volatility years, the VIX itself fell by year-end. Fear, historically, mean-reverts. What the seasonal study is, graded plainly, is a population statistic about Decembers in general. It is not a measurement of this one, and the desk that published it halved its own number before the week was out.
The forward targets sit in the same category. Goldman Sachs, Morgan Stanley, and Barclays have each set a 2027 S&P 500 target of 8,700; UBS holds 8,900. These targets were published into a tape that had just broken their own near-term work: Morgan Stanley’s chief strategist models 14% earnings growth to $420 and had identified 7,100–7,300 as the support zone being tested. The index closed Friday at 6,910.39, below the bottom of that zone. Goldman’s earnings path has U.S. growth slowing from 36% in 2026 to 11% in 2027, and both JPMorgan and Goldman expect the Federal Reserve to be raising rates into the new year — JPMorgan projects three consecutive quarter-point hikes across December, January, and March. The 8,700 targets require the earnings to do all the work while the multiple compresses. That is a forecast, and it is recorded here as one.
The seasonal case, at its own discount
| Measure · mid-Nov → Dec 31 | All years since 1990 | VIX 18 or higher · 17 yrs |
|---|---|---|
| Average gain | +2.4% | +1.34% |
| Index rose in the period | — | 59% of years |
| Rose in the final stretch · Dec 15 → year-end | 27 of 36 years | — |
| Median drawdown on the path | — | 3% |
| VIX fell by year-end | — | 14 of 17 years |
The last row is the one figure in the bulls’ favor, and it is real: in 14 of those 17 high-volatility years, fear itself mean-reverted by year-end. The empty cells are the promise’s own: the study does not print them.
“A population statistic about Decembers in general. It is not a measurement of this one — and the desk that published it halved its own number before the week was out.”
The positioning
Against the targets, the positioning. Bank of America’s November Fund Manager Survey, reported by Bloomberg, shows institutional investors cutting their overweight in technology by the most since 2006. Barclays, in the same week it held an 8,700 target, raised healthcare and high-quality stocks. The published word is bullish. The surveyed allocation is leaving the sector the rally depends on.
This desk does not grade the difference as hypocrisy. It grades it as the mechanism. A target is a statement made to the market. An allocation is a position taken in it. When the two diverge, the position is the one with money behind it, and the statement is the one that keeps a bid in the market while the position is adjusted. Whoever supplies that bid is the counterparty to the adjustment. The question the week’s remaining document answers is whether that counterparty has shown up.
Statement vs. allocation
“A target is a statement made to the market. An allocation is a position taken in it. When the two diverge, the position is the one with money behind it.”
What retail actually did
Citadel Securities’ strategist Jory Weiss, in a client note covering sessions through Tuesday, November 17, records retail doing almost the opposite of what the dip-buying story requires. Cash equity activity was subdued on both Monday and Tuesday: gross retail volumes ran at roughly the 35th percentile of the past year on Monday and the 37th on Tuesday, and Monday’s net buying sat at the 34th percentile, about one-third of the one-year daily average. Exchange-traded funds accounted for all of Monday’s net equity buying. In every other sector, retail investors were net sellers.
What retail bought, in size, was protection. On Tuesday, retail traders spent approximately $104 million net on put option premiums — the highest single-day total since March 31 and the third highest of the past twelve months. Index hedging led: roughly $57.2 million of net put premium on the S&P 500, 8.30 times the one-year average, at the 100th percentile, on pace for the highest daily SPX hedging level since March 2025. Single-stock protection concentrated in the chip complex: Taiwan Semiconductor at $32.0 million, 12.70 times its average and also a 100th-percentile print; Nvidia at $7.85 million, at the 98.4th percentile; ASML, AMD, and Micron all at or above the 88.8th percentile.
The date is the fact that matters. This was Tuesday, November 17 — before Thursday’s 2% fall and before Friday’s 3.70%. The largest retail hedging wave of the year was placed ahead of the two worst sessions of the week, by the cohort the seasonal story casts as the rally’s fuel. Who stood on the other side of those put sales is not in the record, and this desk does not invent it. What is in the record is sequence: protection first, break second, and no dip bid in the cash market at all. As of the last flow data published, the counterparty has not arrived.
Protection first
The highest single-day total since March 31, and the third highest of the past twelve months. What retail bought, in size, was protection.
| Name | Net put premium | Vs 1-yr avg | Percentile |
|---|---|---|---|
| S&P 500 Index · SPX | $57.2M | 8.30× | ■ 100th |
| Taiwan Semiconductor · TSM | $32.0M | 12.70× | ■ 100th |
| Nvidia · NVDA | $7.85M | — | 98.4th |
| ASML · AMD · Micron — single-stock protection | — | — | ≥88.8th |
SPX hedging ran on pace for its highest daily level since March 2025. Who stood on the other side of those put sales is not in the record, and this desk does not invent it.
Sequence, as recorded: protection first, break second, and no dip bid in the cash market at all. As of the last flow data published, the counterparty has not arrived.
The mechanism
The thesis this piece tests was stated by the desk’s principal before the weekend, and it is recorded here in his terms. Retail buys the dip out of fifteen years of conditioning — a decade and a half in which every decline of consequence was repaid, so the reflex is trained, not foolish. Institutions, smelling turmoil, are risk-averse unless they can short the mechanism itself; when they decide to reduce, they require a bid to sell into, and the conditioned dip-buyer is that bid. If the current break is a wobble, the bid is rewarded as it always has been. If it is a repricing — of AI capital expenditure, of the credit that funded it, of the multiple — then the bid absorbs the transfer, and the dip-buyers hold the loss. It is the structure this publication’s No Counterparty papers describe in another market’s language: the public is the gambler, the casino’s creditor, and the landlord’s mortgage holder — so when the waves land, there is no one to transfer the loss to except the party who arrived last, buying.
One clarification belongs in the record, because the thesis is routinely misread as bearishness. It is not a claim about the models. The underlying technology stands as graded: real, and its largest supplier’s financials unfabricated. Nothing in the week’s tape grades the machines. What is being repriced is the add-ons — the leverage, the circular financing, the obligations kept off the balance sheet, the offering priced for the pop, the multiple itself: the financial apparatus layered over a working core. The add-ons are where the loss is manufactured, and they are, not coincidentally, the parts that require a counterparty.
The week supplied a miniature. Anthropic priced its offering at $83 a share, on an upsized 980 million shares. It opened at $111.50 and closed its debut at $97.17, up 17% — the pop the public was invited to buy. On day two it fell 8.5% to close at $88.91, printing an intraday low of $83.05, five cents from the price at which the issuer sold. No verdict attaches to two sessions. But the geometry is the thesis in one ticker: supply placed at $83, demand recruited at $97 and above, and the price returning to the print to ask who is holding what. Monday’s close, measured against $83, is this desk’s standing test of the deal — scheduled, not prejudged.
The Monday test
The thesis is falsifiable, and Monday, November 23, grades the first sitting. The baselines are Friday’s closes: S&P 500 at 6,910.39, VIX at 37.66, Anthropic at $88.91 against its $83 print.
The bail is confirmed if: morning strength, if any arrives, is sold into by midday, repeating Friday’s pattern of rallies used as exits; the VIX holds at or rises from 37.66 rather than bleeding off; Anthropic closes below $83, putting every buyer since the print underwater at the issuer’s own price; and when the week’s retail flow data is published — it arrives with a lag, and the verdict on this limb may land a day late — cash-equity buying finally appears while prices fail to hold their gains, the signature of a bid being used rather than rewarded.
The bail is refuted if: Anthropic holds above $83 through the close; the VIX falls materially from 37.66; the S&P 500 closes above Friday’s level on breadth rather than on a narrowing set of names; and retail buying, when it shows in the flow, meets a floor — gains that survive into the close instead of being harvested. A Thanksgiving-week bounce is the strategists’ base case, and it may well print. A bounce, on its own, grades nothing. What grades the thesis is who is on the other side of it, and whether the people who bought it still own it at the bell.
First sitting: Monday, November 23
Lag note. Retail flow data arrives with a lag — the verdict on Test 04 may land a day late. A Thanksgiving-week bounce is the strategists’ base case, and it may well print. A bounce, on its own, grades nothing.
“What grades the thesis is who is on the other side of it, and whether the people who bought it still own it at the bell.”
Sources and grading. Index closes and intraday levels: Yahoo Finance watchlist record, Friday, November 20, cross-checked against Barron’s (“Worried About Black Monday? What Wall Street Says Could Happen”), which agrees on the S&P 500 (−3.7%), Nasdaq (−5.3%), Nvidia (−14% Friday, −23% on the week, to $125.53), Anthropic (−8.5% to $88.91 after a $97.17 debut close against the $83 offer), and the VIX (touched 40, closed 37.66). Strategist readings in Barron’s — capitulation, a Thanksgiving bounce, a retest around the December 8–9 FOMC — are color, recorded as such, not as record. Seasonality, 2027 targets, rate paths, earnings paths, and the Bank of America survey: TheStreet (Farooque and Chatterjee, “Goldman Sachs issues sobering verdict on S&P 500 year-end rally,” ~November 20), reporting Goldman Sachs, Morgan Stanley, Barclays, UBS, and JPMorgan material and a Bloomberg-carried survey; the seasonal figures are population statistics and are graded as such. Retail flow and put-premium figures: Citadel Securities, strategist Jory Weiss, client note as circulated; the data runs through Tuesday, November 17, and the note’s exact publication date is [UNVERIFIED] in the record available to this desk. Anthropic offering figures: price $83, 980 million shares upsized, open $111.50, debut close $97.17, second-day close $88.91, intraday low $83.05. Gross proceeds of $81.3 billion are arithmetic from the share count and price; a larger figure circulating in this desk’s notes is an implied valuation, not the raise, and is not used. The exit-liquidity thesis is the principal’s, stated Saturday, November 21, and is presented as a thesis under test, not as a finding. This piece makes no recommendation. Monday’s closes will be graded against the $83 line at 6:43 p.m. ET, as scheduled.