Does the Turtle trading strategy work on EUR/USD?

EUR/USD, Daily chart, Oct 2, 2016 to Oct 1, 2026. Backtest run on October 2, 2026.

Key results

Total return
-2.7%
Buy & hold: +0.1%
Annual return (CAGR)
-0.3%
Buy & hold: 0%
Max drawdown
-14.6%
Buy & hold: -23.3%
Win rate
40%
Profit factor
0.94
Trades
80
Longest losing streak
8 trades
Period
Oct 2, 2016 to Oct 1, 2026
10 years
Market
EUR/USD
Daily

Equity curve vs buy & hold

StrategyBuy & hold
Cumulative return of the strategy and of buy and hold, EUR/USD, Oct 2, 2016 to Oct 1, 2026At the end of the period the strategy is at -2.7%, against +0.1% for buying and holding EUR/USD over the same days.
Cumulative return in percent of the starting capital. The dashed line holds the asset for the whole period.

Returns year by year

YearReturnTrades
2016+0.9%3
2017-0.7%8
2018+0.7%8
2019-6.2%9
2020-0.9%7
2021+4%8
2022-3.8%7
2023+1.8%8
2024-2.8%9
2025+6.6%7
2026-1.6%6

Calendar years, compounded from monthly results. The first and last years can be partial.

The Turtle experiment is trading folklore. In 1983 Richard Dennis and William Eckhardt taught a group of novices a mechanical trend-following system, and the core of it fits in one line: buy a breakout to new highs, sell a breakout to new lows, and exit on a shorter breakout in the opposite direction. System 1 used 20 days to enter and 10 days to exit. We tested a version built from calendar levels, inspired by this description: the previous month's extremes to enter and the previous week's to exit. It keeps the slow-in, fast-out logic, but it is an adaptation, not the original system.

The rules we tested

  • Market and period: EUR/USD, daily candles, October 2016 to October 2026.
  • Long entry: the close crosses above the previous month's high.
  • Long exit: the close crosses below the previous week's low.
  • Short entry: the close crosses below the previous month's low.
  • Short exit: the close crosses above the previous week's high.
  • Stop: a 350-pip catastrophe stop, never hit: all 80 trades closed on the weekly exit rule. No take-profit, so winners run as long as the weekly structure holds.
  • Sizing and costs: 100% of equity per trade, 0.02% commission per side, $10,000 starting capital. No volatility-based sizing and no pyramiding, unlike the original Turtles.

What the backtest shows

The shape is right for trend following: the average win (1.25%) is larger than the average loss (0.88%). With that payoff, the break-even win rate is about 41%. The system won 40.0% of the time, so it missed by little more than a point and finished at -2.7%, with a profit factor of 0.94. The pair itself ended almost exactly flat (+0.1%), so there was no long-term drift to ride either way.

Losing streaks are part of the deal: the longest run was 8 losing trades in a row. The worst drawdown was 14.6%, against 23.3% for holding the pair.

The yearly table shows how dependent the system is on the market's mood. 2019 was the worst year (-6.2%, 1 winner out of 9 trades), a year of failed breakouts. 2025 was the best (+6.6%, 4 winners out of 7), and 2021 also did well (+4.0%). Between those, most years were small gains or small losses. With about eight trades a year, commissions are not the problem: 80 round trips at roughly 0.04% each cost about 3% of the position over the decade.

Why it doesn't work here

Trend following makes its money from a few outsized winners that pay for many small losers. The Turtles found those winners by trading dozens of markets at once: when one market trended, it paid for the others. A single major currency pair spends much of its time in ranges, and over ten years it offered only a handful of trends long enough to carry the system. The weekly exit is also a compromise: tight enough to give back little in a reversal, but it closes trades on ordinary pullbacks, which keeps the average win at 1.25%.

What you could test next

  • Use a looser exit, such as the previous month's opposite extreme, to let trends run longer.
  • Use a faster entry, such as the previous week's high or low, to catch trends earlier at the cost of more false breakouts.
  • Add a trend filter, for example trading only in the direction of the 200-day moving average.
  • Run the same rules on gold or a stock index, where long trends have been more frequent.

Methodology and assumptions

Starting capital
10,000
Position size
100% of equity
Commission
0.02% per side
Data window
Oct 2, 2016 to Oct 1, 2026
Run date
October 2, 2026
  • No repainting: every signal is computed on closed candles only, so the backtest never acts on a price a trader could not have seen yet.
  • Past performance does not predict future results. A backtest is a historical simulation, not a forecast.

Frequently asked questions

Is Turtle trading still profitable?

On EUR/USD alone over the last ten years, this Turtle-style breakout was not: -2.7% over 80 trades with a profit factor of 0.94. The original Turtles traded a broad portfolio of futures markets, and a single currency pair gives the system far fewer large trends to catch.

What are the Turtle trading rules?

The original System 1 entered on a 20-day breakout and exited on a 10-day breakout in the other direction. Our version uses calendar levels instead: entry on a close beyond the previous month's high or low, exit on a close beyond the previous week's opposite extreme.

What win rate does a Turtle system need?

It depends on the payoff. Here the average win was 1.25% and the average loss 0.88%, so the system needed about 41% winners to break even. It got 40.0%.

Does trend following work on forex?

It needs sustained trends. In this test, EUR/USD produced profitable years when it trended, such as 2025 (+6.6%), and painful ones when breakouts kept failing, such as 2019 (-6.2%, 1 winner out of 9 trades).

Reproduce it, then change it

Every number on this page comes from the Backtrex engine. Rebuild the strategy in the app, then change the market, the timeframe or a parameter and see whether the result holds.

Ready to backtest your strategies?

Import your strategy as text, Pine Script or MQL5 and run your first test for free, no credit card. From import to a 10-year result, count about 2 minutes.

Create your account in 30 seconds and run your first backtest. No credit card.

Other backtests