🇰🇷 한국어 원문 · Originally written in Korean on Oct 21, 2026. Slides have been re-labeled in English.
At the bottom of the June-July correction, I cut my losses and sold.
And impressively, I sold almost exactly at the bottom.
I’ve been kicking myself over that.
I later found out that some people were buying at the very moment I was selling, because they judged it to be the bottom.
They had several rules of thumb, and one of them was the 10-year moving average.
“Buy the index below its 10-year moving average and you can’t lose.”
And they didn’t only talk about buying.
They said they sold once the market climbed out of the bottom, and that’s how they made money.
In other words, when to buy and when to sell come as a set.
Now another correction is here.
KOSPI is down 39% from its June high (9,114), and KOSDAQ is down 36% from its 1-year high.
I didn’t want to make the same mistake twice.
So I backtested both halves together.
- When to buy: Does buying below the 10-year average really never lose? Which of RSI, MACD, Bollinger bands, the length of a decline, the drawdown from the high, and the link to the S&P 500 and Nasdaq actually points to the bottom?
- Once bought, when to sell: At a profit target? After a set drop from the peak? In pieces?
The data starts in 1990 for KOSPI and 1996 for KOSDAQ: 36 years of KOSPI history.
First, What I Found
The backtest boiled down to four findings.
① “Buy below the 10-year average” mostly held for KOSPI only.
For KOSPI, buying below the 10-year moving average left you up a year later 90% of the time.
For KOSDAQ it was 61%. From 2007 to 2011, KOSDAQ’s 10-year average still contained the 2000 dot-com bubble prices, so even “below the average” was expensive. Since 2012, dips below the 10-year average have mostly been good entries for KOSDAQ too.
② The best buy signal was “cheap on a long view” plus “the US is oversold too.” In this post I call it signal A.
- KOSPI signal A: below the 10-year average or 30%+ below the 1-year high, while the Nasdaq’s RSI is below 35 (the US is oversold too). It was up 60 days later 90% of the time.
- KOSDAQ signal A: 20%+ below the 1-year high + below the 10-year average + a sharp short-term drop (down 8%+ in 5 days, or 6-day RSI below 15).
③ Among selloffs, “Korea-only” was the weakest.
This is when Korea falls 30% or more while the US (S&P 500) sits near its high. It was up 60 days later only 42% of the time for KOSPI and 32% for KOSDAQ.
Joint Korea-US selloffs, where the US fell too, were actually closer to the bottom.
Maybe that’s because macro shocks like Covid set off global panic selling, while a Korea-only crash tends to be the start of a long bear market?
④ How long to hold before selling? KOSPI did better held longer, KOSDAQ did better sold sooner.
KOSPI had big rebounds like 1998 and 2020, so a split take-profit or a trailing stop worked better. KOSDAQ rebounds faded, so selling everything at +15% was best.
For the “Korea-only” setup, holding on lost money; only taking a quick +10% worked.
- Profit target: sell once up +10% or +15% from the buy price.
- Trailing stop: sell when the price falls 12% from its highest close since buying. Rides big rebounds all the way.
- Split take-profit: sell half at +10%, sell the rest 10% below the subsequent peak.
Put together as rules for each index, it looks like the table below.
For KOSPI, the core rule is to buy on signal A and use a split take-profit; joint selloffs are held longer, and Korea-only crashes are a quick +10% trade.
For KOSDAQ, buy on signal A and sell at +15%.
Now let’s apply these rules to today’s market.
The Call: Wait on KOSPI, Buy KOSDAQ in Stages
KOSPI: wait.
The drawdown (-39%) and the drop over the last 10 days (-21%) are among the largest on record.
But the index is still 96% above its 10-year average, and the US is near its high (Nasdaq RSI 43).
It’s the Korea-only crash setup from above, and signal A is off.
In that setup, holding lost money; only quick +10% exits worked.
So if I go in, it will be small, and I’ll take +10% and get out.
If the Nasdaq’s RSI drops below 35 and the US corrects too, signal A turns on, and I’ll look at it properly then.
KOSDAQ: buy in stages, take profit at +10-15%.
It has fallen below its 10-year average (804) and is 36% below its 1-year high.
On October 15-16 a sharp short-term drop joined in, and KOSDAQ signal A fired.
KOSDAQ rebounds don’t last, so locking in at +10-15% beat holding for longer (in the backtest, +15% was slightly better).
That said, there was almost no trading below 780 over the past year, so support is thin all the way down to the July low (645).
I’ll buy in pieces, not all at once.
What if I had followed these rules in 2026?
- KOSDAQ (KODEX KOSDAQ 150): bought Jul 9 → took +15.9% on Oct 7 → re-bought Oct 16 at 12,905 won, now +1.6%.
- KOSPI (KODEX KOSPI): bought on signal A Jul 31 → closed by Oct 14 with the split take-profit, +11.2%. The Korea-only quick trade: bought Jul 29 → +13.7% on Aug 14, re-bought Oct 14 → now -8.3%.
While I was cutting losses at the bottom, the rules were buying, selling into the rebound, and buying again into this decline.
Without a sell rule, just holding, KOSDAQ would be at -3.8% right now.
The numbers confirmed it: selling matters as much as buying.
Read These Numbers With Care
Right after the conclusion, I need to add something.
Crashes like this don’t happen often.
Over 36 years, each rule produced only 6 to 30 trades.
- Few cases. One or two different outcomes turn a 100% win rate into the 80s. The error bars are roughly ±15 percentage points. And a handful of big events (the 1997 IMF crisis, the 2000 dot-com bust, the 2008 financial crisis, 2020 Covid) show up in several rules at once.
- The win-rate illusion. Profit-target exits won 80% of the time even when buying on a random day. So this post judges a rule by whether it earns more per year than random entries, not by its win rate.
- Cherry-picking. The rules were picked from 135 buy-sell combinations. I kept only those that worked both before and after 2008, but luck may still be in there.
- 2026 is unprecedented. KOSPI more than doubling in a year and then falling 39% has never happened before. Past patterns may not hold.
So read these numbers not as “do this and you can’t lose” but as “in the past, these conditions tilted the odds a little in your favor.”
Below is everything I reviewed to get there.
1. What Was Compared, and How
I ran every combination of 15 buy signals and 9 sell rules.
Buy when the signal fires, sell when the exit rule triggers, then wait for the next signal.
What I cared about most was the benchmark.
Indexes rise over long periods, so a signal only means something if it beats “buy on any day and sell with the same rule.”
Selloff signals also stay on for days at a time, so signals within 20 trading days of each other count as one case.
And since the Korean market opens before the US trades that day, each Korean day T only uses the US close from the night before.
The indicators and sell rules are defined in the table above.
As shown earlier, the most important ones are the 10-year moving average, the drawdown from the 1-year high, the Nasdaq’s RSI, and the three sell rules.
2. When to Buy: Is “Below the 10-Year Average” Really a Sure Thing?
For KOSPI, it mostly held.
Buying below the 10-year average left you up a year later 90% of the time, with a median gain of +31%.
Right after the 1998 IMF crisis, the March 2020 Covid crash, and September-October 2022 all fall in this zone.
But it wasn’t a sure thing. Buying when KOSPI first slipped below its 10-year average in August 2000 meant -22% a year later.
For KOSDAQ, it didn’t.
Buying below the 10-year average was up a year later only 61% of the time, with a median gain of just +3%.
Buying in November 2007 meant -60% a year later.
KOSDAQ’s 10-year average still contained the 2000 dot-com bubble prices (2,834), so from 2007 to 2011 the average itself sat far too high.
“Below the 10-year average” only means cheap if the average isn’t inflated by a bubble.
Since 2012, the times KOSDAQ dipped below its 10-year average (2019, 2020, 2022, 2024-25) were mostly good buying opportunities.
3. When to Buy: RSI, Bollinger Bands, Duration and Depth
Honestly, this was the most surprising result.
Buying when RSI was below 30, below the lower Bollinger band, or on the 5th straight down day was up 60 days later 47-58% of the time.
No different from buying on any day (56% for KOSPI, 51% for KOSDAQ).
After a 10%+ drop in 5 days, KOSPI bounced within 5 days 71% of the time, but over 3 months the edge mostly disappeared.
How long it had been falling was no guide.
From the 7th straight down day on, results got worse (60-day win rate 36% for KOSPI, 27% for KOSDAQ).
A long losing streak didn’t mean a bounce was near; it meant the selling pressure was strong.
How far it had fallen only mattered on a 1-year view.
Buying 30-40% below the 1-year high gave a 1-year median of +20% for KOSPI and +13% for KOSDAQ.
But of the bear markets that hit -20%, 62% went on to -30%, and 75% of those reached -40%.
“It’s fallen this much, it must be the bottom” is the most dangerous thought.
4. When to Buy: Selloffs Shared With the US Were Closer to the Bottom
Korean indexes follow the previous night’s US session (correlation 0.3).
The session before that has almost no link. US influence carries over for exactly one day.
What mattered more was the nature of the selloff.
- Korea crashes while the US sits near its high: the 60-day win rate was the worst, 42% for KOSPI and 32% for KOSDAQ. 1990-92, the 1997 IMF crisis and 2000 fall in this group, and the declines dragged on.
- The US is oversold too: when KOSPI was 30%+ below its 1-year high or below its 10-year average, with the Nasdaq’s RSI below 35, it was up 60 days later 90% of the time.
The odds of a bottom rise when the US is shaken too. That was the most useful finding on the buying side.
For someone like me who sold at the bottom, “wait for a confirmed bounce” feels natural.
So I compared waiting for the first up day, a move back inside the lower Bollinger band, a MACD turn, a recovery of the 5-day average, or RSI crossing back above 30.
The win rate didn’t improve; the entry price just went up 1-3%.
5. When to Sell: The Profit-Target Win Rate Is an Illusion
I turned “sell once it climbs out of the bottom” into rules and attached them.
At first, the profit target looked perfect. Add a +10% target to KOSPI signal A and the win rate was 100%.
But buying on any day and selling at +10% won 82% of the time too.
Indexes usually recover eventually.
So instead of win rates, I judged each rule by whether it earned more per year than random entries under the same sell rule.
I also tried “out of the bottom” exits such as a recovery of the 200-day average, RSI 70, or the upper Bollinger band. They sold too early or waited until the price fell back, and did worse than profit targets.
6. When to Sell: The Right Sell Rule Revived Some Signals
Looking at buying and selling together, some signals that seemed useless on the buy side came back to life.
- KOSPI’s short-term selloff signals (MACD histogram below -1%, -8% in 5 days) beat the benchmark both before and after 2008 when paired with a 12% trailing stop. But the win rate is only 35-44%: a string of small losses, then one big rebound makes up for them.
- KOSPI’s Korea-only crash won all 6 trades only with a quick +10% exit. Raising the target to +15% or switching to a trailing stop lost money.
- KOSPI’s RSI below 30, lower Bollinger band and down-day streaks stayed at the random-entry level whatever sell rule was attached.
- For KOSDAQ, only the signals built on the 10-year average held up in both periods. Short-term signals worked before 2007 and fell apart after 2008.
I compared split take-profits too.
For KOSPI, selling half at +10% and the rest 10% below the next peak was the best split, and a plain 12% trailing stop on everything earned even more.
Riding big rebounds like 1998 and 2020 all the way was what counted.
KOSDAQ was the opposite: selling everything at +15% worked best. Its rebounds didn’t last.
7. Final Rules and the ETF Check
All of this is summarized in the rule table shown earlier. To recap:
- KOSPI core: buy when cheap on a long view (below the 10-year average or 30%+ below the 1-year high) and the Nasdaq is oversold; sell half at +10% and the rest 10% below the peak. 17 wins in 17 trades, 6.8% a year, above buy-and-hold (5.9%), with about half the maximum drawdown.
- KOSPI aggressive: the same entry, sold on a 12% trailing stop. The highest return at 9.2% a year, but only a 50% win rate.
- KOSPI secondary and quick trade: hold joint Korea-US selloffs longer with a 12% trailing stop; trade Korea-only crashes quickly for +10%.
- KOSDAQ core: buy in a bear market + below the 10-year average + after a sharp drop, sell at +15%. 4.2% a year while buy-and-hold returned -1.9%.
You can’t buy the index itself, so I also checked with the KODEX KOSPI and KODEX KOSDAQ 150 ETFs.
Buy at the next day’s open after a signal, with commissions and a one-tick worse fill.
- KODEX KOSDAQ 150 + KOSDAQ core rule: 7.3% a year, maximum drawdown -26%. Buy-and-hold: 2.5% a year, -57%.
- KODEX KOSPI + KOSPI aggressive rule: 14.5% a year, above buy-and-hold (13.0%), with a smaller maximum drawdown (-34% vs -41%).
One caution.
On a day like July 31, when the market swung 18% in a single session, the same rule turned into +14% or -16% depending on whether you bought at the open or the close.
In a market like this one, it seems right to split purchases over several days even when the signal fires.
In June and July, when I sold at the bottom, I had no rule for buying and no rule for selling.
This time, at least, I’ll set both before I move.
Note: This post is a personal record of a backtest and is not a recommendation to buy or sell any product. Past returns do not guarantee future results.
