Should you sell before a long holiday?
I counted 77 of them
Korea closes its market for several days twice a year — Lunar New Year and Chuseok, the autumn harvest festival — plus a scattering of one-day holidays that make three-day weekends. Before each one the same advice circulates: "You cannot sell while the market is shut, so go in flat." I counted what actually happened around 77 market closures from 2014 to 2026, every stock.
01The market, and how I split the holidays
KOSPI is Korea's main board (Samsung Electronics, SK Hynix); KOSDAQ is the secondary board, smaller and more volatile, closer to the Nasdaq. The session runs 09:00–15:30 Korea Standard Time (UTC+9). Korea has two multi-day holidays a year: Seollal (Lunar New Year, January or February) and Chuseok (the harvest festival, September or October). Each closes the market for three to five weekdays.
I took the trading calendar and flagged every gap of four or more calendar days between consecutive sessions. A normal weekend is three days (Friday to Monday), so four or more means at least one weekday was closed.
- Lunar New Year — January–February, 4+ days off · 12 holidays
- Chuseok — September–October, 4+ days off · 12
- Year-end — last session of December to first session of January · 12
- 3-day weekends — one public holiday attached to a weekend (Labour Day, Memorial Day, Liberation Day, Hangul Day and so on) · 41
D0 is the last trading day before the break, D1 the first trading day after it.
- The three days before: open two sessions before D0 → close on D0
- The last day: D0 open → D0 close
- Holding through: D0 close → D1 close (the part people worry about)
- The first day back: D1 open → D1 close
- The first week back: D1 open → open five sessions later
The baseline is the same stocks on any random day, same trade shape. The median is the middle result when every trade is lined up from worst to best: half did better, half did worse.
02All 77 together: nothing
| Win rate | Baseline | Median | Baseline | |
|---|---|---|---|---|
| Three days before | 42.8% | 43.1% | −0.41% | −0.41% |
| The last day | 43.7% | 40.3% | −0.22% | −0.32% |
| Holding through | 40.1% | 40.9% | −0.36% | −0.22% |
| First day back | 37.4% | 40.3% | −0.53% | −0.32% |
| First week back | 43.4% | 44.4% | −0.57% | −0.41% |
Combined, it is the baseline. Holding through a closure won 40.1% of the time against 40.9% for any random night. That does not mean nothing bad ever happened during a holiday in twelve years — it means bad things happened at the same rate as on any other night.
But this flat result is two opposite effects cancelling out.
03By holiday: they split
| Lunar New Year | Chuseok | Year-end | 3-day weekend | Baseline | |
|---|---|---|---|---|---|
| Three days before | 43.5% (−0.34) | 48.2% (−0.22) | 53.5% (+0.26) | 39.5% (−0.66) | 43.1% (−0.41) |
| The last day | 48.5% (−0.06) | 52.0% (+0.09) | 50.1% (+0.01) | 38.9% (−0.40) | 40.3% (−0.32) |
| Holding through | 52.0% (+0.11) | 44.3% (−0.22) | 43.7% (−0.22) | 35.3% (−0.64) | 40.9% (−0.22) |
| First day back | 48.1% (−0.10) | 42.3% (−0.22) | 39.4% (−0.44) | 32.6% (−0.82) | 40.3% (−0.32) |
| First week back | 55.6% (+0.65) | 47.1% (−0.22) | 51.7% (+0.20) | 37.4% (−1.43) | 44.4% (−0.41) |
Lunar New Year is the good one. Holding through it won 52% against a 41% baseline, and the first week back won 55.6% with a median of +0.65% after costs.
Chuseok is good for exactly one day — the last session before it (52%). After the break, nothing. "Stocks rally after Chuseok" is not in the data.
Year-end is good in the three days before (53.5%). Those are the last sessions of December, the same window as the post-cutoff rebound in my year-end tax-selling article.
Three-day weekends are bad across the board. The first day back won only 32.6%, the first week back 37.4% with a median of −1.43%. The gap between Lunar New Year and a three-day weekend on the first week back is 18 points. That is why the 77 together showed nothing.
04Twelve Lunar New Years, one by one
| Last session | Days closed | Holding through | Share up after | First week back |
|---|---|---|---|---|
| 2015-02-17 | 5 | +0.19% | 54.7% | +1.80% |
| 2016-02-05 | 5 | −4.00% | 9.0% | −1.51% |
| 2017-01-26 | 4 | −0.22% | 41.7% | −0.64% |
| 2018-02-14 | 4 | +1.13% | 75.0% | +1.71% |
| 2019-02-01 | 5 | +0.25% | 55.6% | +1.24% |
| 2020-01-23 | 4 | −3.41% | 11.4% | −3.78% |
| 2021-02-10 | 4 | +0.69% | 65.3% | +0.14% |
| 2022-01-28 | 5 | +2.17% | 81.9% | +3.87% |
| 2023-01-20 | 4 | +0.44% | 61.9% | +1.27% |
| 2024-02-08 | 4 | +0.59% | 64.4% | +1.49% |
| 2025-01-24 | 6 | −0.83% | 29.6% | −0.55% |
| 2026-02-13 | 5 | +0.82% | 63.1% | +1.56% |
The first week back was up in 8 of 12. The two bad ones have names: February 2016 coincided with the global selloff over oil and European banks, and January 2020 was the week COVID first spread. Only 9% and 11% of stocks were up after those breaks.
Take those two out and the remaining ten saw 30–82% of stocks up after the break, mostly above the baseline. "Something will happen while the market is closed" was right 2 times in 12.
05Why are three-day weekends bad?
This is the most striking number in the article, and the one to read most carefully.
Look at when the 41 three-day weekends fall: early May (Labour Day, Children's Day), early June (Memorial Day), mid-August (Liberation Day), early October (Foundation Day, Hangul Day). And that stretch of the calendar contained an unusual number of sharp selloffs: August 2015 (China's devaluation), May 2019, August 2020, August 2021, June and August 2022, February and May 2026. Eight of the 41 had a first-week median below −5%, and every one of them is on that list.
So the number says less "three days off hurts stocks" and more "May to August had a lot of bad weeks". Whether the holiday or the season is the cause, this data cannot tell. Either way, buying on the first day back from a three-day weekend was a poor choice for twelve years.
06Caveats
77 holidays, twelve per type. Multiplied by stocks that is tens of thousands of trades, but one holiday's market pushes all of that holiday's stocks the same way. "8 of 12" is more honest than a win rate.
The −0.22% median that keeps appearing for holding through is the cost alone: many stocks open exactly where they closed. Read it as zero.
Costs are the 0.22% buy-side deduction only, with no sell-side cost or spread. Delisted stocks are not in the data. Prices are from the Korea Exchange only from March 2025, when a second venue opened.
07What I take from it
- Combined, there is no holiday effect: holding through won 40.1% vs 40.9% on any night
- Lunar New Year was fine to hold through (52%) and the week after was good (55.6%), up in 8 of 12 years
- Chuseok was good for the last day only, then nothing
- The week after a three-day weekend was bad (37.4%), but that number is tangled up with May–August selloffs
- "Something happened while the market was closed" — 2 Lunar New Years out of 12 (oil in 2016, COVID in 2020)
"Go flat before the holiday" found little support in twelve years of data. Going flat before Lunar New Year usually cost money; going flat before Chuseok gained and lost nothing. The only case where being flat helped was the three-day weekend, and even there it is unclear whether the holiday or the season did it.
Holidays feel dangerous because you cannot sell,
but the data says those few days were no different from any other few days.
댓글 쓰기