Quantitative hedge funds took their worst single day since August on Thursday, in a decline that Goldman Sachs data described as a simultaneous drop across strategies that has been rare since the pandemic.
Systematic long-short funds fell 3.7 standard deviations on the day, according to Goldman’s prime brokerage data, a move large enough that it sits outside the range these portfolios normally produce. Momentum factors dropped 3.4 standard deviations. Crowded long positions sold off together while shorts rallied against them, the pattern that appears when positions financed with borrowed money are being closed.
It was the third such episode this year. The recurrence is the interesting part. A single violent day can be attributed to a news event or one large seller, but the same combination of falling crowded longs, rising shorts and elevated borrowing has now appeared three times in ten months, which points to something structural in how these portfolios are built.
The structure is the problem. Quantitative managers construct positions by ranking thousands of securities against factors, and those rankings push many funds toward the same names. When one fund needs to reduce risk, it sells its most liquid holdings first, which are also the names its competitors hold. The selling then shows up in factor returns rather than in any single stock.
Asia absorbed the damage. The monthly drawdown for these funds reached 6.8 percent, with positions tied to Korean artificial intelligence names largely cleared out and Taiwan holdings still near their highs. That distribution matters. When one crowded theme is emptied and another is not, the remaining exposure becomes the obvious candidate for the next round of selling.
Total borrowing across the prime brokerage book stood at 310 percent, a level that leaves room for further reductions before positions return to the size they held before the trade became crowded. Options expirations in recent weeks had been providing a cushion that dampens daily moves, and that cushion has largely expired.
The mechanics of unwinding explain why the pressure tends to spread rather than stay contained. Funds that reduce exposure raise cash against margin calls, which shrinks the collateral available to finance other positions. Prime brokers respond by raising requirements on the positions that look riskiest, which forces more sales.
The macro explanation is that the trade had become crowded because the story was good. Holding semiconductor and data center exposure has been one of the better-performing positions of the past 18 months, backed by earnings that kept beating expectations. Positions that work attract capital until the marginal buyer is another fund holding the same view, at which point the exit becomes narrow.
The behavior of the shares and the behavior of the businesses have diverged before. Semiconductor companies have reported record results this earnings season while their shares fell sharply, and the funds that own them use mechanical rules rather than views about the companies. They sell because prices fell and volatility rose, not because they have concluded anything about demand.
What usually follows is a period of reduced gross exposure. Funds cut the size of their books, clients ask about the drawdown, and position sizes stay smaller until volatility settles. That process can take weeks, and it typically removes the marginal buyer from the assets that had been rising.
This year’s episodes have also been shorter than those of 2020 and 2022, when the unwinding of crowded positions lasted months. Faster settlement cycles, better intraday risk reporting and limits on how much concentration a prime broker will fund have shortened the cycle. They have not eliminated it.
Analysts who track positioning said the risk now sits in the positions that were not cleared. Taiwan exposure remaining near its highs while Korean positions were sold is the kind of asymmetry that precedes a second leg, because the funds still holding the crowded trade are the ones with the most to sell if prices keep falling.
Nothing in the data says the underlying businesses are impaired. The names being sold are the same ones reporting record profits, and the selling comes from portfolios that rebalance on rules rather than on analysis. That disconnect is what makes the episode uncomfortable for investors who own the shares outright and have no mechanical reason to sell.
For the funds themselves, the question is whether the rules that produced the concentration are being adjusted. A systematic manager that survived two earlier episodes without changing its factor weights will have the same exposures the next time volatility rises, and its clients will have the same conversation about drawdowns that the previous two episodes generated.
Prime brokers have their own reasons to moderate these episodes. The desks that finance these funds earn fees on the borrowing and on the trades the funds place, so a client that blows up is a client that stops generating revenue. Several of the largest brokers have tightened the concentration limits they allow in a single theme and ask for more frequent reporting on intraday exposure than they did five years ago.
Clients provide a slower-moving check on the same behavior. Allocators who have paid performance fees through two prior drawdowns ask their managers what changed in the model, and a fund that answers with the same factor weights it held before is asking to be judged on returns that have not yet been produced.
