Korea 2026: A Leverage Unwind That Is Not Over

This case is different from the others. Every other piece in this section covers an event that is finished, where the ending is known. This one is not — it is still happening.

So this piece covers only what is already complete and checkable: what happened, how the mechanism worked, and which numbers are now settled facts. It contains no judgement about where the index goes next and no view on whether current levels are high or low. Data runs through the close of 5 August 2026, and all index figures are closing prices.

In early 2025 the KOSPI was near 2,400. On 22 June 2026 it closed at 9,114 — eighteen months, +280%.

On 30 July of the same year it closed at 5,593. That is 26 trading days and -38.6% from the high.

And on the very next session, 31 July, it rose 17.91% in a single day — the largest one-day gain in the KOSPI’s history, surpassing the record set during the financial crisis in October 2008.

Those three numbers together are the most useful thing in this case.

What is already settled

Start with what can be checked.

• The all-time intraday high of 9,385.59 came on 19 June; the closing high was 9,114.55 on 22 June.
• 23 June fell -9.99% in one session — the start of the unwind.
• 28 July fell -10.84%, 29 July another -5.98%. Reports describe this as the first back-to-back circuit-breaker days on record for the index, and the ninth circuit breaker of the year.
• July closed -22.2%, the worst month since the 2008 financial crisis.
• The trigger was external: news that China had begun mass production of domestic deep-ultraviolet lithography equipment, combined with disappointing results from SK Hynix, repriced the entire AI/semiconductor chain. Korean indices carry an unusually heavy weight in that chain.
• Media tallies put forced liquidations of margin positions at roughly ₩2.3 trillion over two and a half months, with foreign investors pulling about $13 billion out of Korean equities in July alone.
• The 31 July rebound came with roughly ₩7 trillion of foreign net buying and was triggered by strong US technology earnings. The same day, new cash-deposit requirements for leveraged ETF investors took effect.
• The Korea Exchange has publicly said it is evaluating a temporary short-selling ban and tighter daily price limits. As of the data cutoff, no decision had been made.

All of that has happened and will not change. What is not settled is everything that comes after — and this piece leaves it alone.

How leverage turns a decline into forced selling

A large share of the retail money that drove the rally came in through margin borrowing and leveraged ETFs.

Both share one property: they generate sell orders that have nothing to do with what you want.

For margin positions the mechanism runs like this:

Price falls → equity ratio breaches the maintenance level → broker issues a margin call → if it is not met, the position is force-closed → closing it means selling → that selling pushes prices lower → more accounts breach their maintenance level → …

Note the critical link: a forced liquidation is not a sale because someone decided to sell. It is a sale because the rules require it. It does not look at valuation, at fundamentals, or at what you think.

This is another positive-feedback spiral — structurally identical to LUNA’s death spiral, the margin spiral in UK pensions, and the short squeeze in GameStop: a price move forces trades in the same direction, which amplifies the move.

The difference here is who was forced. This time it was retail investors themselves.

The ₩2.3 trillion figure shows how fast the spiral ran. It also shows how much of the selling pressure was not a decision at all, but an execution.

The largest up day in history happened inside the worst month in history

This is the most counterintuitive part of the case, and the most useful.

The +17.91% on 31 July did not occur in a bull market. It occurred inside a month that fell 22% — the session immediately after the low.

That is not a coincidence. Across the 31 trading days from 23 June to 5 August:

	Days
Fell more than 3%	12
Rose more than 3%	8
Moved more than 3% either way	20 of 31

Crashes and melt-ups are two faces of the same thing. Volatility clusters, and when it rises it rises in both directions. Markets do not become violent on the way down and gentle on the way up.

Two concrete consequences follow:

One: “wait for a bounce to get out” and “buy the bounce” are the same bet in this environment. Both wager that a market swinging 5–18% a day will move the way you need it to.

Two: a huge up day inside a downtrend does not tell you the downtrend is over. In this case the index rose 8.18% on 9 June and 6.24% on 15 July — and made new lows after each.

This is also why the training data on this site uses real market history rather than randomly generated series. A random walk has no volatility clustering, so it cannot teach you the feel of a tape like this one.

Leveraged ETFs: being right about direction is not enough

Leveraged ETFs rebalance daily. What they promise is N times today’s move — not N times the move over a stretch of time.

The size of that gap can be calculated directly. Taking the 22 June high as the start and 5 August as the end:

	Cumulative return
KOSPI index	-27.4%
2× long, daily rebalanced	-52.8%
2× short, daily rebalanced	+34.7%

Look at the 2× short row first. The directional call was exactly right, the leverage was pointed the right way, and the naive expectation would be +54.8% — twice the index’s 27.4% decline. The actual result was +34.7%, a shortfall of 20 percentage points.

Where did those 20 points go? Almost entirely into one session.

At the close on 30 July, that 2× short position was up 114.3%. One day later, at the close on 31 July, it was up 37.5%.

Down 35.8% in a single day — two thirds of the entire profit handed back.

Now look at the 2× long row: -52.8%, while “twice the index decline” would be -54.8%. It actually lost slightly less than the naive multiple. That is not an illusion — in a sustained decline, daily rebalancing automatically shrinks the position, which cushions the loss.

So the real conclusion is not “leveraged ETFs always lose more.” It is:

The outcome of a leveraged ETF depends on the path, not on the start and end points. For the same “index down 27%,” a different route produces results tens of percentage points apart. You can have a view on direction when you buy. You cannot have a view on the path.

For completeness: at the 30 July low, the 2× long was down 65.3% while the index drawdown was -38.6%. A 2× long that had not yet been liquidated had already lost two thirds of its capital.

Circuit breakers, rule changes, and whether you can trade at all

Nine circuit breakers so far this year, two of them on consecutive sessions. While a circuit breaker is active, the market stops matching orders.

At the same time, the new cash-deposit rule for leveraged ETFs took effect on 31 July, and the exchange is publicly discussing a temporary short-selling ban and narrower price limits.

For an individual trader all of this means one thing:

When you most need to act, whether you can act is not entirely up to you.

This is the same risk that appeared as market makers not answering the phone in 1987 and as brokers restricting GME purchases in 2021. It never shows up in calm markets, and it is most likely to show up in extreme ones — exactly when you need to move.

Any plan that depends on “I’ll cut the loss when it gets there” carries a hidden assumption: that the market will be open and your order will fill. In a market that fell more than 3% on 12 days out of 20, that assumption does not always hold.

No ending here

At the data cutoff the KOSPI closed at 6,623, down 27.4% from the high, with three of the last four sessions higher.

That says neither that the bottom is in nor that it isn’t. This piece takes no view and recommends no action.

To be explicit about why it is worth reading now: not because it predicts anything, but because the mechanical part has already run a full cycle and can be checked. The ₩2.3 trillion of forced liquidation has happened. The record daily moves have happened. The path decay in leveraged products is already sitting in their NAVs. None of those numbers change based on what the index does next.

There will be an ending. When it arrives, this piece will get one.

What transfers

One: separate your selling from the system’s selling

Selling pressure in a decline comes partly from judgement and partly from forced liquidation. The second kind has no reasoning, cannot be talked out of it, and does not stop because prices have “already fallen enough.” A market with heavy margin participation falls differently from a low-leverage one — faster, deeper, and with less regard for value.

Before judging the character of a move, ask whether its fuel is voluntary or forced. That is the same framework as GameStop, pointed the other way.

Two: volatility runs in both directions, and it clusters

“The biggest up day landed inside the worst month” is not a Korean curiosity — it is the norm. Reading a large rally as a trend reversal is the most common misuse of this fact.

Three: leverage changes more than the size of the move — it changes whether you survive long enough to be proved right

An unleveraged holder at -38.6% still owns the asset and can choose to wait. A 2× holder at the same moment is down 65.3% and may well not last long enough for their view to play out — even if the view is ultimately correct.

Leverage converts “how much do I lose if I’m wrong” into “can I survive until I’m right.” Those are different questions, and the second one usually decides the outcome.

Understanding that is worth far more than remembering how high the KOSPI went.

This is a mechanical review and risk-education piece covering the completed portions of an ongoing market event. It is not investment advice, does not recommend any security, and makes no inference about current or future market conditions. Index figures are taken from the closing-price data used in the chart (through 2026-08-05); other statistics are cited from public media reports.