🇰🇷 한국어 원문 · Originally written in Korean on Oct 26, 2026. Slides have been re-labeled in English.
“Buy the breakout to a new high, and sell when the trend bends.”
Whenever trend-following investing comes up, Turtle Trading is always part of the conversation. It is famous for its simple rules—and for the story that complete novices were taught nothing but these rules and went on to make fortunes. What happens if I apply those exact rules to South Korean stock indices, specifically using index ETFs (Exchange-Traded Funds, index funds traded on an exchange like regular stocks) that any individual investor can buy? An ETF does not face single-stock delisting risk, making it a much safer vehicle.
To find out, I backtested the classic Turtle Trading rules on daily candles across 36 years of the KOSPI (South Korea’s main benchmark stock index, similar to the S&P 500) and 30 years of the KOSDAQ (South Korea’s tech-heavy secondary stock market, similar to the Nasdaq).
First, What Is Turtle Trading?
In 1983, American futures traders Richard Dennis and William Eckhardt made a bet: “Are great traders born, or can they be trained?” Dennis recruited novices through a newspaper ad, taught them a set of mechanical rules for two weeks, and funded them with his own money. He called them “Turtles,” and many of them famously generated massive returns.
The rules I tested here match “System 2” from Way of the Turtle (2007), written by original Turtle alumnus Curtis Faith:
- Buy (Entry): Buy when today’s price breaks above the highest high of the prior 55 trading days (roughly three calendar months).
- Sell ① (Trend Exit): Sell when the price drops below the lowest low of the prior 20 trading days (roughly one month).
- Sell ② (Hard Stop-Loss): Sell unconditionally if the price falls by 2N from the entry price. Here, N is the 20-day average of ATR (Average True Range, the average daily price range). For a broad stock index, N is typically around 1.5% to 2% of the price, making a 2N stop roughly 3% to 4%.
The original Turtles spread risk across dozens of liquid futures markets, going both long and short. For this test, I simplified the setup into the most practical form an individual investor can easily execute: trading a single South Korean index ETF—either KODEX 200 (tracking the top 200 blue-chip stocks on the KOSPI) or KODEX KOSDAQ150 (tracking the top 150 stocks on the KOSDAQ)—going 100% long on a buy signal and sitting 100% in cash (at 0% interest) on a sell signal. All trades are assumed to execute at the next day’s open after a signal triggers, deducting a 0.015% commission plus 1 tick of slippage.
For the years before those ETFs existed, I used the underlying indices as proxies. Because KODEX 200 was listed in October 2002, I used the KOSPI index prior to that date; because KODEX KOSDAQ150 was listed in October 2015, I used the KOSDAQ Composite index before that. The KOSDAQ index launched in July 1996, and because 1996 data lacks daily high and low prices, the KOSDAQ backtest starts in 1997—making 30 years the maximum available history.
Key Metrics Needed to Read the Results
- CAGR (Compound Annual Growth Rate): The annualized compounded rate of return—what constant yearly percentage gain would have produced the final result.
- MDD (Maximum Drawdown): The worst peak-to-trough percentage drop during the period. An MDD of -76% means KRW 100 million once shrank to KRW 24 million from its peak.
- Sharpe Ratio: Excess return earned per unit of volatility (price swings). Around 0.3 is average for a single equity index, while above 0.5 is considered solid.
- Win Rate, Payoff Ratio, and Expectancy: Win rate is the percentage of trades that ended in profit; payoff ratio is average win divided by average loss; expectancy is the average return expected per trade (
Win Rate × Avg Win − Loss Rate × Avg Loss). - Market Exposure: The percentage of trading days during the entire period that the strategy actually held stocks rather than cash.
Conclusion: It Works, But Not as Well as You Might Think
- KOSPI 36 Years (1990–2026): Turtle returned 5.9% annualized (CAGR) with a -34% maximum drawdown (MDD). Buy-and-hold returned 6.5% annualized with a -76% MDD. Turtle earned slightly less return, but cut the worst drawdown by more than half.
- KOSDAQ 30 Years (1997–2026): Turtle returned 3.9% annualized versus -1.0% for buy-and-hold (-91% MDD). On paper, Turtle won by a wide margin. However, almost all of that outperformance came from a single year—the 1999 dot-com bubble. Looking only at the period since KODEX KOSDAQ150 was listed in 2015, Turtle returned just 1.7% annualized, leaving virtually nothing on the table.
- Low Win Rate, High Payoff Ratio: Out of every 10 trades, roughly 6 end in a loss. However, because winning trades average around +13% while losing trades average around -4%, the expectancy per trade is roughly +3%. That is the textbook profile of trend following.
- On an Exposure-Matched Basis, It Did Not Beat the Index: Turtle sits in cash for roughly two-thirds of the entire timeline. So I compared it against a simple benchmark that holds 38% in the index ETF and 62% in cash from day one. Since KODEX 200’s listing in 2002, that static 38% allocation achieved both a shallower MDD (-23.9% vs. -27.4%) and a higher Sharpe ratio (0.45 vs. 0.37).
In short, Turtle Trading on a single Korean index was not a strategy that boosted returns, but one that gave up a large chunk of bull-market gains in exchange for sidestepping major crashes. When applied to a single equity index, that trade-off is steep.
Evidence 1. What 36 Years Looked Like
KOSPI: Safely in Cash During Every Major Crisis
During the 1997–98 Asian Financial Crisis, the 2000 dot-com bust, and the 2008 Global Financial Crisis, a buy-and-hold KOSPI investor had to stomach peak-to-trough drawdowns of -50% to -76%. Turtle spent most of those crashes in cash, having exited early via its 20-day low rule or 2N stop-loss as soon as the index rolled over.
There is a recent example as well. After peaking on June 22, 2026, KODEX 200 fell 41% through October 23. Turtle had entered on April 22 and exited on July 9 when the 20-day low broke, closing that trade with a +26% gain.
KOSDAQ: A Win Built on a Single Year
In 1999 alone, Turtle made +142% on the KOSDAQ (while buy-and-hold surged +239%). Then, when the bubble burst in 2000 and buy-and-hold crashed -79%, Turtle stopped out at -34%. Virtually the entire 30-year gap between Turtle and buy-and-hold was decided in those two years. From 2000 through 2026, Turtle’s cumulative return on the KOSDAQ was just +7% total—or 0.3% per year.
Scorecard
Evidence 2. Why It Makes Money Despite Losing 6 Times Out of 10
- Over 36 years on the KOSPI, Turtle took 83 trades (2 to 3 trades per year). Only 41% of them were profitable.
- The 32 trades exited via the 2N stop-loss kept losses small, averaging -4.4%.
- The 51 trades exited via the 20-day low had a 67% win rate and averaged +8.1%. A handful of extended trend rides (up to +53.5%) paid for dozens of small stop-outs.
- Still, a stop-loss does not always guarantee a small loss. The worst KOSDAQ trade (-21.5% in March 2000) happened when the market gapped down below the stop price overnight and executed at the opening price. Most recently, in October 2026, a KOSDAQ 150 trade gapped straight through its stop and closed at -13%.
An expectancy of around +3% per trade means that if I follow the rules mechanically, each trade nets roughly +3% on average over the long run. However, losing streaks reached 5 to 6 consecutive trades—meaning an investor must be able to stomach half a dozen straight stop-outs without abandoning the system to capture that positive expectancy.
Evidence 3. Strong in Crises, Weak in Grinding Bull Markets and Sideways Ranges
Splitting the 36-year history into three distinct eras makes the strategy’s personality unmistakable:
- Through 2002 (Asian Financial Crisis & Dot-Com Boom/Bust): Turtle crushed the indices—returning 5.7% vs. -3.6% annualized on the KOSPI, and 11.4% vs. -15.3% annualized on the KOSDAQ.
- 2002–2015 (China Boom, 2008 Financial Crisis & Range-Bound Market): Buy-and-hold won. Holding the KOSPI returned 11.5% annualized, whereas Turtle made just 2.8%. In a market that repeatedly rallied and paused, Turtle kept buying 55-day breakouts near local tops only to get stopped out on pullbacks.
- 2011–2024 (14 Years of Box-Range Markets): KOSPI Turtle earned a cumulative total return of just +9% over 14 years, posting flat or negative returns in 11 of those 14 years. Simply holding the KOSPI over the same stretch gained +48%.
What pulled KOSPI Turtle’s full-period return up to its current level was the explosive bull run of 2025–2026. Excluding those last two years drops its 1990–2024 annualized return to 3.7%.
Evidence 4. A Fair Exposure-Matched Comparison—and Turtle vs. Volatility Harvesting
Much of Turtle’s shallow drawdown comes from the simple fact that it sits in cash most of the time. Since KODEX 200 was listed in 2002, Turtle held stocks on only 38% of trading days. So I compared it against a static portfolio holding 38% in KODEX 200 and 62% in cash throughout. Turtle edged out the static portfolio in annualized return by 0.5 percentage points, but the simple 38% buy-and-hold had both a smaller maximum drawdown and a higher Sharpe ratio. That means there is no compelling reason to trade actively just to take on the same level of market risk.
I also compared Turtle on the exact same daily dataset against the Volatility Harvesting strategy I covered in earlier posts (a ladder system that buys in tranches each time the price drops by a set step and sells each tranche on a rebound):
- KODEX 200: Turtle (6.0% annualized, Sharpe 0.37) outperformed Volatility Harvesting (2.8% annualized, Sharpe 0.17).
- KODEX KOSDAQ150: Volatility Harvesting (6.4% annualized, Sharpe 0.34) beat Turtle (1.7% annualized, Sharpe 0.04). In a market like the KOSDAQ—characterized by wide swings and frequent reversals—harvesting mean-reverting volatility fit the market structure far better than chasing breakouts.
- 50/50 Blend of Both Strategies: Because the daily returns of the two strategies have low correlation (0.15 to 0.45), blending them 50/50 on KODEX 200 reduced MDD to -21%. On KODEX KOSDAQ150, however, blending dragged the Sharpe ratio down from 0.34 to 0.25. Moreover, because the drawdown reduction largely comes from Turtle sitting in cash two-thirds of the time, blending is not much different from simply running Volatility Harvesting with smaller capital.
Limitations and Interpretation
- Heavy Reliance on Outliers (Luck): Turtle’s 30-year victory on the KOSDAQ rests almost entirely on a single year (1999), and its recent KOSPI numbers lean heavily on the 2025–2026 rally. From a tail-risk perspective, a track record dominated by one or two extreme episodes should be treated with deep skepticism (Nassim Taleb, Fooled by Randomness).
- Multiple-Testing Adjustment: I also tested 27 parameter combinations around the baseline rules (entry lookbacks of 40, 55, and 70 days; exits of 10, 20, and 30 days; stops of 1.5N, 2N, and 3N). The classic 55/20/2N specification landed right in the middle of the pack. Under the Deflated Sharpe Ratio (DSR, Bailey & López de Prado, 2014)—a statistical test that penalizes the Sharpe ratio for the number of parameter trials—the result fell short of statistical significance, meaning random chance cannot be ruled out.
- Index Proxies vs. Actual ETFs: The pre-listing history uses price indices that exclude dividends. On the KOSDAQ side, the broad KOSDAQ Composite and the large-cap KOSDAQ 150 also behaved quite differently (returning 0.8% vs. 1.9% annualized between 2015 and 2026).
- Execution Assumptions: If I assume trades execute intraday at the exact breakout price rather than at the next day’s open, the 30-year KOSDAQ annualized return jumps from 3.9% to 6.0%. In a gap-prone market like the KOSDAQ, execution timing makes a major difference—though intraday stop orders in live trading can also suffer worse slippage during fast markets.
- Difference from the Original Turtle System: The original Turtles diversified across dozens of uncorrelated futures markets, sized each position so that 1N equaled 1% of account equity, and pyramided (added units) into winning trends. What I tested here answers a narrower question: “What happens if an individual investor applies the breakout entry and exit rules to a single Korean stock index?”
So, Does It Work?
Applied standalone to a single South Korean index, the Turtle rules “get you out of major market crashes, but charge a heavy toll in missed bull-market gains.” Over 36 years, Turtle failed to beat buy-and-hold on the KOSPI, and while it beat the KOSDAQ over 30 years, that entire edge came from a single year in 1999.
If I were to use these rules in my own portfolio, I would not use them as a standalone replacement for index investing, but rather as a secondary risk-management signal telling me when to trim equity exposure—reducing weight when the 20-day low breaks and restoring it when the 55-day high is reclaimed. In a future post, I will test whether applying Turtle the way it was originally intended—spread across multiple uncorrelated assets such as US equities, bonds, and gold—leads to a different conclusion.
Disclaimer: This post is for information and analysis purposes only and does not constitute a recommendation to buy or sell any financial product. Past backtest performance does not guarantee future returns.
