A false breakout is one of the most frustrating things in trading. Price breaks the 20-day low, you sell following the breakout, and a couple of hours later the market turns and knocks you out with your stop. Linda Raschke and Laurence Connors proposed making money from exactly that: trading against those breakouts. They called their strategy “Turtle Soup” because it is cooked with the stops of those who trade breakouts with the “Turtles” system. It sounds logical. I wrote a robot with the rules from the book and ran it on seven markets for 72 months. I’ll show you what came out and which robot is better to use instead.
- Loss in all 7 markets. Classic Turtle Soup on D1 lost between 10.1% (EURUSD) and 32.6% (USDCHF) over 72 months. Only 9–18% of trades were winners.
- The cause: strong trends. In these years gold doubled in price and USDJPY went from 105 to 160. In a market like that, a breakout of an extreme tends to continue more than reverse, and the strategy keeps positioning itself against the move.
- Optimization didn’t save it: on EURUSD H4, out of 34 profitable parameter sets in training, only 6 stayed positive in the forward test. The best survivor made 17% in 72 months: less than 3% a year.
- What works better: the “Holy Grail” robot, Raschke’s other strategy, but trading with the trend: +69% in 72 months on USDJPY H4 with a drawdown of 8.7%, forward +21%.
📊 Download the “Holy Grail” robot that does work, for free
The strategy rules
Turtle Soup is described in Raschke and Connors’ book “Street Smarts” (1995). The name is a joke on Richard Dennis’s famous “Turtles” system, which bought the breakout of the 20-day high and sold the breakout of the 20-day low. Raschke noticed that many of those breakouts are false and proposed entering in the opposite direction. The rules for buying (selling is the mirror image):
- New low: today price fell below the low of the last 20 days.
- The previous low is not recent: the previous 20-day low was set at least 4 days earlier.
- Entry: when price comes back above the previous low, the robot buys. Entry is only possible on the day of the false breakout.
- Stop: a little below that day’s low. After that, Raschke advises managing the trade with a trailing stop; in the robot, the stop is moved up to the low of the last two candles.
- Position size: based on a risk of 1% of the balance per trade.
What it looks like on the chart
Below are two real trades from the robot, from the best parameter set the optimizer found on EURUSD H4, and both are from the forward period. The blue stepped line is the high of the last 30 candles; the yellow one, the low. The purple dot marks the previous extreme that price broke.
The trade this strategy is loved for. On March 10, 2026, EURUSD broke above the high of the last 30 candles, but it didn’t hold and went back below the previous high. The robot sold at 1.16602, with a stop at 1.17201, above the breakout high. Take profit of one stop: 1.16003. The next day price reached the target: +$108. The breakout was false, just like in the book.
The trade that repeats far more often. On May 13, 2026, EURUSD fell below the 30-candle low and went back above it for a while. The robot bought at 1.17234, with a stop at 1.16991. But the “false” breakout turned out to be real: two hours later price kept falling and hit the stop: −$109. The problem is that these two pictures can’t be told apart in advance: at the moment of entry they look the same.
By hand it looks like this: mark the low and the high of the last 20 days. If price broke one of them and then came back to the other side of the level, enter against the breakout, with the stop beyond the day’s extreme. It looks very pretty in the history, when you already know which breakout was false.
How I ran the test
MetaTrader 5 Strategy Tester, FxPro demo account with its quotes, spreads, commissions and swaps. Period: October 2020 to October 2026, 72 months. Deposit of $10,000, leverage 1:500, “1 minute OHLC” mode: the robot sees how price moved within the day and enters at the moment of the return to the level, as in the book. Why exactly 72 months, I explain at the end of the article.
Turtle Soup over 72 months: a loss in all seven markets
| Instrument (D1) | Result | Profit factor | Max drawdown | Winning trades |
|---|---|---|---|---|
| EURUSD | −10.1% | 0.81 | 21.2% | 16% |
| WTI (oil) | −20.1% | 0.55 | 23.3% | 13% |
| USDJPY | −20.5% | 0.60 | 25.5% | 12% |
| BITCOIN | −22.5% | 0.69 | 25.9% | 18% |
| GBPUSD | −24.6% | 0.50 | 36.0% | 18% |
| GOLD | −31.6% | 0.28 | 37.0% | 9% |
| USDCHF | −32.6% | 0.45 | 40.0% | 13% |
Not a single market in positive territory, and in almost all of them the drawdown exceeds 20%. There are three reasons:
- The 2020–2026 market was a trending market. Gold rose from about $1,900 to $4,000, USDJPY from 105 to 160, and bitcoin went through several huge waves. In a market like that, a breakout of a 20-day extreme tends to continue more than reverse. Raschke’s strategy was born in the 90s and is designed for a market where breakouts often fizzle out.
- The stop is too close. A stop beyond the day’s extreme is usually a fraction of the ATR. Any new retest of the level hits it, and on gold only 9% of trades were winners.
- The two-candle trailing stop doesn’t let profits grow. Even when the reversal did happen, the stop was moved up so close that the trade was closed on the first pullback.
The same classic rules on EURUSD H4 work even worse: −82.9% in 72 months. On a lower timeframe there are more false and real breakouts, and the spread eats stops that are already small to begin with.
What if optimization saves it?
I gave the optimizer free rein: channel length of 10–40 candles, age of the previous extreme, minimum stop of 0.3 to 1.2 ATR, take profit of 1 to 4 stops, trailing stop, time limit on the trade and a trend filter using EMA 50–200. H4 timeframe, EURUSD and gold markets. The optimization was done only on 10/2020–09/2024, and afterward I checked what it found on 10/2024–09/2026, months the optimizer didn’t see.
- EURUSD H4: out of 247 parameter sets, 34 were profitable in training, and in the forward test only 6 of them stayed positive. The other 28 “worked” only on the data they were fitted to.
- GOLD H4: out of 245 parameter sets, 17 were profitable in training. The ones that stayed positive in the forward test had drawdowns of 19–59% along the way.
- The best survivor on EURUSD H4: 30-candle channel, stop no smaller than 1.2 ATR, take profit of one stop and trading only in the direction of the EMA 150: in training +8.7%, in the forward test +7.8% with a profit factor of 1.56. Over 72 months, +17.1% with a drawdown of 4.7%.
Formally, this parameter set passed the forward test. But I don’t recommend it, and I’ll explain why. First, 17% in six years is less than 3% a year: that doesn’t pay for the risk or even for a VPS rental. Second, this is no longer Raschke’s Turtle Soup: the EMA 150 filter only lets you enter with the trend, and a take profit of one stop turns the system into short trades on pullbacks; almost nothing is left of the idea of catching reversals. And third, when 6 out of 34 parameter sets that were profitable in training survive, that looks more like luck than an edge.
What to use instead of Turtle Soup: robots that passed the test
If Turtle Soup ends up positioned against the trend, the logical thing is to look at robots that trade with it. These robots, in the same 72-month test, held up when checked on data the optimizer didn’t see:
| Robot | Over 72 months | Max drawdown | Forward 10/2024–09/2026 |
|---|---|---|---|
| “Holy Grail”, USDJPY H4 | +69.3% | 8.7% | +20.7%, profit factor 2.39 |
| TTM Squeeze, GOLD H4 | +55.0% | 4.7% | +36.1%, profit factor 2.73 |
| Heikin-Ashi, GOLD H4 | +96.4% | 13.8% | +54.4%, profit factor 1.71 |
| For comparison: classic Turtle Soup (7 markets, D1) | from −10.1% to −32.6% | up to 40.0% | — |
📊 Download the “Holy Grail” robot for USDJPY
📊 Download the TTM Squeeze robot for gold
Where to run them and how to download them
The robots work in MetaTrader 5 with any broker. My test was done at FxPro. The other brokers in my ranking also work: RoboForex and Tickmill. The robots are free for clients of these brokers:
- Open a MetaTrader 5 account with one of the brokers above (or use the one you already have).
- Enter the Telegram bot with the button and send your account number.
- After verification, the bot sends you the file: the robot, the ready-made parameters and the instructions. Run it in your broker’s tester, it takes an afternoon, and then put it live.
If you want to check the false breakouts yourself, the bot also has the Turtle Soup robot, with the classic and the optimized parameters, to experiment with in the tester. It’s not advisable to put it on a real account.
My conclusion
Turtle Soup is a beautiful idea: profiting from those who got caught in a false breakout. But in 2020–2026 the market didn’t punish those who traded breakouts, but those who positioned themselves against them. The classic version loses in all seven markets, and optimization only found one parameter set that earns less than 3% a year and already bears little resemblance to the original. False breakouts didn’t disappear; they simply can’t be told apart from real ones in advance, and a close stop beyond the extreme gets hit before the reversal has time to arrive. Protect your deposit: if you’re going to look for reversals, don’t do it against a strong trend. And to trade, use a robot that passed the test, for example the “Holy Grail” from Linda Raschke herself: it enters with the trend on the pullback, instead of standing in its way.
Today I also published the analysis of the “Holy Grail” robot, which passed the test on USDJPY H4. The system Turtle Soup was invented against I analyzed separately: the Turtles strategy. Earlier, the grid with martingale, double top and double bottom and pivot trading also failed the test.
Why 72 months?
The market changes fundamentally every 3–5 years: participants and volatility change, and trends give way to ranges. That’s why I don’t run 10–20 year tests: they mix different markets into a single average figure, and after a market change the strategy has to be changed anyway. Turtle Soup is a clear example. It was created in the 90s as a bet that breakouts often fizzle out. The last six years were years of strong trends, and the same logic became a loser. Maybe, when markets spend a lot of time in ranges again, false breakouts will once again feed systems like this one, but you have to trade the market that exists today.
To look for strategies that fit the current market I built my app “Find Your Strategy”, in the browser or in Telegram. There you can test any idea on up to 72 months of history, and every 2 weeks the app runs tens of thousands of strategies on its own and shows you the ones that work over the last 6, 12 and 36 months. It doesn’t promise profits: it helps you notice in time that the market has changed and that an old idea has stopped working.
Backtest results do not guarantee future results.
Sources
- Linda Bradford Raschke and Laurence A. Connors: “Street Smarts: High Probability Short-Term Trading Strategies” (M. Gordon Publishing, 1995), chapter “Turtle Soup” (false breakout of the 20-day extreme).
- Curtis Faith: “Way of the Turtle” (2007): rules of the Turtles system with the 20-day channel breakout.
- My own tests in MetaTrader 5 on the FxPro demo server, October 2026.
Where to trade
These are the brokers and exchanges I use that work with clients from Latin America. Always check that the service is available in your country and start small.
Risk warning: CFDs, Forex and crypto derivatives carry a high risk of losing money due to leverage. This article is not investment advice.
