KOSPI closes at 6,852.58 on August 20, 2026 — 381 points back in one session

August 19, 9:06 a.m.: a sell-side sidecar. August 20, 9:57 a.m.: a buy-side sidecar.

The same circuit fired on consecutive sessions. Only the direction changed.

1. The numbers recorded on August 20

Of the 398.66 points lost the day before, 381.41 came back. The gap left over is 17.25 points.

Put differently: after two sessions the KOSPI sits 17.25 points below its August 18 close of 6,869.83. The index travelled roughly 780 points across those two days, and ended 2% of that distance from where it started.

The KOSDAQ stayed quiet again, rising 1.99% while the main board rose 5.89% — the same structure as the previous day's fall. What moved was not the market but one sector.

2. What went up

The line worth pausing on is SK hynix across two days.

The company announced the largest share cancellation in Korean corporate history and fell 9.75%. It announced nothing the next day and rose 12.73%. This page covered the first half of that on August 20 in "SK hynix to cancel 40 trillion won of shares — the largest ever, and the stock fell 9.75%."

The most defensible reading is contextual rather than causal. On August 19 two separate downward forces landed at once: the US 30-year yield at its highest since 2007, and an overnight selloff in US chip stocks. The cancellation news was buried under both. When those forces lifted, it showed up in the price. No report decomposed this numerically.

A second factor arrived on the morning of August 20: reports that Samsung Electronics will approve a shareholder return in the 100-trillion-won range at a board meeting this month. Expectations that began with SK hynix spread to Samsung, which rose 9.49%. That story is covered separately in "Samsung's 100-trillion-won shareholder return."

3. Who bought and who sold

Set side by side, the pattern is unmistakable. Foreign and retail investors swapped places.

Retail bought 5.64 trillion won on August 19 and sold 2.65 trillion won on August 20 — roughly half of it back, into a market 5.89% higher. Foreign investors did the reverse.

Neither column tells you who was right. Retail investors did not necessarily sell the same shares they bought, and there is no public data on the average price on either side. What is confirmed is that the buying and selling sides traded roles across two sessions.

4. The won strengthened both days

On August 19 this page recorded a mismatch: foreign investors sold 4 trillion won of Korean equities and the won got stronger, which is the opposite of the textbook direction.

On August 20 there was no mismatch. Foreigners bought 2.2991 trillion won and the won strengthened a further 5.1 won to 1,392.6.

So the currency moved the same way on both days, and only equities changed direction. This page still has not confirmed what drove the August 19 mismatch. But with two days side by side, the simplest reading is that something other than equity flows was setting the currency's direction that week.

5. What is left and what could not be confirmed

• Net foreign buying — three figures appeared for the same session: 2.2991tn, 2.2703tn and 1.7092tn won. This is the cut-off-timing problem covered on August 15 in "What net foreign buying means," but the 28.8bn won gap was not traced to a specific cause.
• The one-day lag — confirmed as a fact; unexplained as a mechanism.
• Intraday range — not cross-confirmed, so not stated.
• Next checkpoint — Samsung's board is due to fix the size of its shareholder return this month. Expectations are already in the price, so a figure below the reported range could move the market the other way. Details are in "Samsung's 100-trillion-won shareholder return" and "What a special dividend is."