"Oil v Gold trade unwinding."
Behind this observation came a list of pairs: Gold vs Oil, HK/China vs Japan/Korea/TW, Value vs Growth.
Put these words together and one picture comes to mind easily. The most popular assets start to get sold. The assets that the market shunned for a long time suddenly gain strength. Trades that used to work all become hard at the same time. For traders, this deserves attention. Behind the price moves, a group of investors may be leaving positions they have held for a long time.
I checked the recent data. I think we need to state the conclusion precisely first: there is clear evidence of shrinking tech positions and of large shifts in cross-market relative strength. The public evidence available now is not enough to confirm that all three pairs are being unwound together in the same direction. Value versus growth in particular: the latest data do not fit this tidy story.
This distinction is also the starting point for understanding the opportunity.
First, go back to October.
On November 3, Investing.com reported Goldman Sachs prime brokerage data: its clients' net selling of global equities in October was the largest in seven months. Net leverage fell 4.2 percentage points to 73.4%. Within that, tech selling over two weeks was about $20 billion. This is an institutional client sample with a defined scope and period. It supports the view that tech positions are shrinking. investing.com
Prices also show a clear split. Nasdaq's October review records that the Nasdaq-100 fell 6.6% and the S&P 500 fell 0.9%, while the S&P 500 Equal Weight Index rose 5.0%. In the same month, the Russell 1000 Value Index rose 4.4% and the Growth Index fell 6.1%. The modest headline decline hid a wide gap between winners and losers inside the market. Nasdaq
The pressure on Asia was more direct. Janus Henderson's monthly review shows that the Philadelphia Semiconductor Index fell 15.3% in October. In US dollar terms, the MSCI country indexes for Korea, Taiwan and China fell 15.4%, 8.3% and 7.0%. These numbers show that semiconductors and Asian tech were under pressure. They are also a reminder: we cannot lump different instruments and different start and end points together and say "everything has fallen more than 20% recently." janushenderson.com
In early November, a new catalyst appeared.
After the US midterm elections, the market reassessed the Middle East conflict and the policy outlook. A Reuters report on November 5 confirmed that Brent and WTI both fell more than 8% in the previous session. A Saxo analysis on November 4 linked the rebound in gold to a weaker dollar, lower Treasury yields and some short covering. Hopes that the conflict will ease are still part of market pricing. They are not the same as a peace agreement that is already in place. Reuters
So this round of change has at least two layers: the tech de-risking that already happened in October, and the repricing that new events triggered in early November. The election can be a new catalyst. It is not where all the earlier position changes began.
First, "A vs B" only says that A and B are being compared. It does not tell us which side was long and which side was short.
If a trade was originally "long oil, short gold," exiting it means selling oil and buying back gold. But oil falling and gold rising is not enough to confirm that both moves came from the same investors. It proves even less that this is the most crowded pair in the market.
To make the direction clear, we can start with three assumptions:
| Assumed original position | Trades needed to unwind it | Possible relative price pressure |
|---|---|---|
| Long oil, short gold | Sell oil, buy back gold | Gold strengthens against oil |
| Long Japan/Korea/Taiwan equities, short China equities | Sell Japan/Korea/Taiwan equities, buy back China equities | China equities strengthen in relative terms |
| Long value, short growth | Sell value, buy back growth | Growth strengthens against value |
This table describes a mechanism. It does not confirm current real positions.
We also need to separate net exposure from gross exposure. The US Office of Financial Research defines gross notional exposure as the sum of the absolute values of long and short exposure. Office of Financial Research
Take a purely arithmetic example. An account has net assets of 1 million yuan. It holds 1 million yuan of longs and is also short 1 million yuan of stock. Notional net exposure is zero, but gross exposure is 2 million yuan. If the longs fall 5% and the shorted stocks rise 5%, the account still loses 100,000 yuan before costs.
Now take a separate de-risking example and ignore price changes for the moment. If the account sells 300,000 yuan of longs from its starting position and buys back 300,000 yuan of shorts, gross exposure falls from 2 million yuan to 1.4 million yuan. Notional net exposure stays at zero. When trades like this happen in a cluster, the market can show a very sharp pattern of "old winners fall, old losers rise."
This also explains why some long-short funds suffer badly while the broad market falls only a little. Both of their legs may move against them at the same time.
But the recent Goldman sample has one detail we cannot ignore. The report says the ratio of selling existing longs to short selling was about 4:1. It does not say "selling longs versus covering shorts." So this data more directly shows de-risking driven mainly by cutting longs. The drop in net leverage alone also cannot tell us how much gross exposure fell. investing.com
The first step in reading fund flows is to stop calling every sale the same kind of unwind.
The weak oil and strong gold of early November fit the price impact of exiting a "long oil, short gold" trade. But the overall positioning in gold does not support a casual claim that the market was broadly and heavily short gold.
The CFTC disaggregated report visible as of November 5 reflects COMEX gold futures positions on October 27. Managed money held 122,508 long contracts and 16,946 short contracts, a net long of 105,562 contracts. Other (Combined)
This does not rule out that some funds held gold shorts. It also does not rule out short covering in early November. What it shows is that "some short covering happened" and "the market was broadly heavily short gold before" cannot substitute for each other.
Also, the CFTC's aggregate categories cannot tell us whether a given gold short matches an oil long held by the same entity. CFTC
The oil-gold relationship also has no fixed direction. In an article for the LBMA, James Steel of HSBC stresses that the effect of oil prices on gold usually passes through the dollar, interest rates and the stock market. You cannot simply use the oil price to forecast the gold price. LBMA
For trading, this means you still need to ask more questions. After oil falls, do real interest rates fall? Does the dollar keep weakening? Can buying in gold continue? One large down candle in crude oil cannot answer these questions.
Narrow the window to November 4 and a clearer regional split appears. The table below uses US-listed ETFs from the same issuer and the change in fund NAV on the same day:
| Market proxy | ETF | NAV change on November 4 |
|---|---|---|
| China equities | MCHI | +4.02% |
| Japan equities | EWJ | +0.83% |
| Korea equities | EWY | -0.31% |
| Taiwan equities | EWT | -0.47% |
The source is the official iShares product pages. These are ETF NAV changes, not the closing moves of the four local market indexes. MCHI also does not represent all Hong Kong stocks or all A-shares. MCHI
The observation that Chinese equities clearly outperformed Japan, Korea and Taiwan that day holds. But on the same day, China also had a separate policy backdrop. From November 4, the central bank cut the rate on its relending facility for stock buybacks and shareholding increases by 0.25 percentage points to 1%. www3.xinhuanet.com
So the price move may reflect several forces at once: overseas de-risking, short covering, better policy expectations and new buying. From "China up, Korea and Taiwan down" alone, it is hard to separate the contribution of each.
The more useful thing to watch next is this: once Asian tech stocks stop falling sharply, can Chinese equities keep their relative strength? If they can, new allocation demand is more worth studying. If the relative advantage disappears quickly, the earlier move may have depended more on short-term position adjustment. This is a judgment that needs to be verified. It is not a direct measurement of where the money came from.
"Value has outperformed for a long time" has data behind it. According to the Nasdaq review, as of the end of October, the Russell 1000 Value Index was up 24.5% year to date, and the Growth Index was down 0.2%. Nasdaq
But "outperformed for a long time" does not imply "has now started to reverse."
On November 4, the NAV of the iShares Russell 1000 Value ETF (IWD) rose 2.56%, and the Growth ETF (IWF) rose 0.35%. On this pair of proxies, measured on the same day and the same basis, value still came out ahead that day. IWD
So a recovery of growth against value can go on the list of scenarios to watch. The evidence is not enough to describe it as a confirmed collective unwind.
It is also easy to confuse equal weight with growth. RSP tracks the S&P 500 Equal Weight Index and cannot stand in directly for the growth style. RSP/SPY is better for observing weight concentration and breadth inside the index. For growth and value, use style measures from the same index family where possible. Invesco US
Several price ratios moving at once does not mean they tell the same story.
In markets like this, the most tempting questions are usually these. How many more days will it fall? When will the rebound start? How long will the window last?
History gives good reference points. It also often overturns answers that are too neat.
In the week of August 6, 2007, several quantitative long-short equity funds suffered unusual losses. The related strategies then rebounded sharply on August 10. Research by Khandani and Andrew Lo proposed an explanation with data behind it: large market-neutral portfolios exited quickly and triggered de-risking in other strategies. The authors did not treat the reasons for any specific fund's forced liquidation as established fact. mit.edu
The follow-up research is more instructive. Using simulated factor portfolios and transaction data, the authors found that signs of a broader adjustment had already begun in July and continued to the end of the year. dspace.mit.edu
If you look only at the most violent week, it is easy to conclude that "the forced selling is over, and mean reversion comes next." Extend the time frame and you find that a short-term price recovery can happen while long-term deleveraging continues.
In March 2020, gold also went through a sell-off. The World Gold Council's analysis at the time held that cross-asset liquidation, leveraged positions and margin needs may have amplified the decline. Because gold is easy to turn into cash, investors may also have sold it to cover losses in other assets. World Gold Council
On March 23, the Federal Reserve announced that it would buy US Treasuries and agency MBS in the amounts needed to support smooth market functioning. To understand how market conditions changed afterward, this kind of liquidity support needs to be considered too. Federal Reserve issues FOMC statement
This case is a reminder that the long-term role of a safe-haven asset and the holder's immediate need for cash are two separate things. Both can affect the price at the same time.
If the selling comes from an urgent need for cash, "it is already cheap" is often not enough to stop it. The more useful questions are whether funding pressure has eased and whether the market can absorb sell orders again.
The BIS review of the August 2024 turmoil records that TOPIX fell about 12% in one day on August 5. By the close on Friday, August 9, the S&P 500 had recovered its losses since Monday, and TOPIX had recovered most of them. This refers to the decline over those few days, not the whole adjustment from the earlier high. The BIS linked the amplified move to the exit from leverage, higher margins and yen-funded carry trades. bis.org
This shows that waiting for confirmation has value, and also a cost. If you wait until every explanation is complete and every indicator looks good, the main part of the recovery may already be over. But the reverse is also true: entering early does not give you an automatic advantage.
Put the three cases together for a more reliable understanding. The most violent trading shock may last only a few days. Position adjustment may continue longer. The final trend still depends on fundamentals and funding conditions. You cannot compress them into a uniform rule of "forced selling for 5–15 days, then a 1–4 week high-win-rate window."
"Catching the unwind opportunity" sounds like one thing. In practice it contains at least three different trades.
Suppose there is already evidence that the "long oil, short gold" trade is being unwound. Opening a new "long gold, short oil" position is a bet that the exit pressure will continue.
The more meaningful things to watch here are these. After the gold-oil relative strength breaks out, does it hold on pullbacks? When bad news appears, does the new strong side still attract buyers? Does the leg that was crowded keep meeting selling pressure?
This kind of trade does not require buying at the extreme low, but it depends heavily on the entry point. If the price has already covered most of the move, the remaining room may not cover the loss if the thesis fails. Even if the story still holds, the trade may not be worth taking.
Here you might reconsider the old winners that were sold off, or even put the original pair back on. But this assumes the old relationship still has a sound economic basis.
I would watch for these changes. New bad news no longer pushes the price to new lows. After the first rebound, pullbacks become smaller and slower. The relative price starts to form higher lows. These signs can only improve the quality of a judgment. They cannot on their own prove that the forced selling is over.
Take oil as an example. "Oil prices start to stabilize" is closer to a clue for reassessing an oil price recovery. It cannot directly justify adding new oil shorts. Following the unwind and trading the recovery after the unwind have different entry reasons and different exit conditions.
Some position adjustments are only a short shock. Others lead the market to price assets by new standards.
If earnings expectations, the policy environment or funding costs change in a lasting way, prices may not return to the old mean. The opportunity may then last longer, but it needs new evidence. Do earnings expectations keep adjusting? Does demand for the assets go beyond brief short covering? Can the new leading sectors hold their advantage through repeated swings?
"The old winners have fallen a lot" is not enough to decide that they should return to where they were. The historical mean has no obligation to become the future price target.
Instead of assuming a fixed number of days, I prefer to set a different observation horizon for each trade:
| Change you want to capture | Time scale for planning observation | Key judgment |
|---|---|---|
| One round of urgent position exits and the short-term recovery | Intraday to several trading days | Whether extreme selling, spreads and volume start to recover |
| Several groups of funds continuing to cut or rebuild positions | Several days to several weeks | Whether the relative trend survives multiple pullbacks and new news |
| A lasting revaluation driven by earnings, policy or interest rates | Several weeks to several months | Whether the new fundamentals keep coming through |
These are scales for planning research and holding expectations. They are not win-rate ranges from historical statistics, and they are not a countdown for the current market.
In practice, I would write down three things in advance.
First, what change would invalidate the reason for the trade. If the trade is China strengthening against Korea and Taiwan, watch whether the relative relationship breaks. If you only watch whether the China long leg rises, you miss losses caused by the other side rising more. If the trade is gold benefiting from easing rate pressure, then a renewed rise in the dollar and real interest rates is a reason to reassess.
Second, how much the whole portfolio would lose if the trade fails. Two legs of the same notional amount can have different volatility, market sensitivity, and funding and borrow costs. Low notional net exposure does not mean low portfolio risk. A ratio chart helps observation, but actual profit and loss also depends on the weights and costs of the two legs.
Third, how long you are prepared to wait if the price does not move as expected. A trade built for a position recovery within a few days should not quietly become a long-term investment after it fails. Time is also one dimension for testing a trade hypothesis.
As for RSI, z-scores or multiples of standard deviation, they can describe "how far the deviation is." They cannot prove on their own "why it must come back." If the relationship itself has changed, the past mean and volatility may lose their value as references.
COT data can also serve only as cross-validation. The report usually takes its sample on Tuesday and comes out on Friday. As of November 5, the gold data above still stopped at October 27, and the next scheduled release is November 6. Even when the next report comes out, it cannot cover all the changes after November 4. It helps to judge the structure of positions. It cannot announce in real time that "the forced selling ended today." CFTC
At this point, it is better to keep three scenarios open than to rush to pick one fixed outcome.
If this is mainly position adjustment, the recent sharp rises and falls may start to retrace once exit pressure eases. Then the question is which relative trends can last. Do not automatically assume that all the old winners will recover.
If the fundamentals are also changing, the new pattern of relative strength may continue. For example, the risk premium in crude oil may keep falling, together with changes in real interest rates and the dollar. Or Chinese policy support may gradually turn into lasting demand for the assets. This kind of judgment needs later facts to carry it forward.
If the pressure spreads further into funding and liquidity, traders may face more forced selling. In that case, cutting gross exposure and keeping the capacity to wait is a valuable choice in itself.
The two nearest clear checkpoints are the US October jobs report at 8:30 a.m. ET on November 6 and the US October CPI at 8:30 a.m. ET on November 10. The dates come from the official BLS calendar. bls.gov
Beyond the data itself, I would watch how the market absorbs it. Do the assets that led the decline still make new lows on bad news? Can the assets that recently gained strength hold up once the supportive news fades? These reactions can help test the trade hypotheses above.
Pairs that look diversified may share the same risk. Commodities, Asian regions and US styles sound diverse by name. Behind them, they may all depend on the dollar, funding costs or the risk budgets of the same group of funds. A factor that one pair is long may also happen to be the factor another pair is short. What you need to check is the portfolio's real exposure, not just the names of the holdings.
Relative strength is sometimes a measurement problem first. Cross-market comparisons should use the same currency, the same return basis and comparable timestamps. US-listed Asian ETFs and local indexes trade in different hours. Continuous commodity futures also involve contract rolls. Before you trust a clean breakout in a price ratio, first check how much of it comes from the assets and how much from the way you measure.
Options can lose money even when the direction is right. Suppose you expect a slow recovery after the panic fades and buy short-dated out-of-the-money options while implied volatility is high. Time decay and falling volatility may both offset the directional gain. CME's explanation of Vega clearly shows how changes in implied volatility affect option prices. Spread structures can help limit some risks, but they do not automatically raise the win rate of your judgment. The Greeks
Looking back, the most valuable part of "crowded trades are unwinding" is that it pushes us to check again: who holds what, who needs to exit, and what reasons remain to support the price after they exit.
I will leave two questions as a starting point for further research on any pair:
Does the money this trade expects to make come from other people having to exit, or from the asset getting better in the future?
If the people who had to trade stop tomorrow, would I still want to hold this?
Start with the pair closest to your own holdings and try to answer these two questions. That usually leads to a clear judgment more easily than chasing every asset that is reversing at once.
The weights of MSCI Korea and MSCI Taiwan are concentrated in Samsung Electronics, SK hynix and TSMC. Tech and chip stocks also dominate the Russell 1000 Growth Index. In October, the Philadelphia Semiconductor Index fell 15.3%, Korea fell 15.4% and the Growth Index fell 6.1%. The regional pair and the style pair very likely share one semiconductor factor. They look diversified, but in fact they are one concentrated bet. (Inference)
Impact: news on chip earnings or AI capital spending will move both pairs at the same time.
The web pages below were checked on November 5, 2026. Fund product pages and the CFTC current report page update on a rolling basis. The article body keeps the data dates used for this piece.
This article is research analysis and a framework for thinking. It is not a recommendation to buy or sell any security, commodity or derivative. The scenarios in it are not price forecasts. Investing involves risk, and you may lose your principal.