What Is Turtle Soup in Trading? Two Rule Sets

Turtle soup is a setup that fades a failed breakout: price pushes through a level plenty of traders were watching, fails to hold, and turns back inside the range. The original version from Larry Connors and Linda Raschke is a precise rule set built on a new 20-day high or low, with a stop order for entry and a stop loss one or two ticks beyond the day's extreme.

The ICT version keeps the idea of fading a sweep and throws the 20-day number away, which leaves three decisions to you.

What is turtle soup in trading?

A reversal setup traded against a breakout. FXOpen describes the trigger in one line: "A failed breakout occurs when price returns inside the prior range after sweeping a swing level." The trade goes the opposite way to the break, which is why Alchemy Markets calls it "a contrarian alternative to breakout trading".

What are the original turtle soup rules?

A new 20-day extreme, an entry stop just inside the old one, and a stop loss just beyond today's extreme. TradingMarkets, Larry Connors' own publication, puts the premise this way: "Larry Connors and Linda Raschke in their book Street Smarts revealed a strategy that looks to fade 20-day breakouts, levels that the famous Turtles trading group were alleged to key off in initiating trades."

NetPicks prints the rules in full and credits Linda Raschke, "where I got this strategy from". For a buy:

• "The market makes a new 20 day low and the lower that price shoots below the former low, adds to the trade setup."
• "The last 20 day low must be at least four days in the past (using daily charts)", so "you can't compare the low of today to the low of yesterday if that was the prior 20 day low".
• The entry is a buy stop "5-10 ticks above the prior 20 day low", and "Forex traders may want to use the same in pips". It is good for one day only.
• "Your stop loss, if entry occurs, goes 1-2 ticks below the low of today".

The sell side is the mirror: a new 20-day high, the previous one at least four days back, an entry stop 5 to 10 ticks below it, and a stop 1 to 2 ticks above today's high.

Two further details. There is a re-entry rule, that if you are stopped out the same day or the next "you can place a stop order, buy or sell, at the same price as your previous entry". And the exit is left open, with "Trailing stops are a good approach as is scaling partial out at 1R".

The "Plus One" in the name is a second timing. TradingMarkets describes the delayed version: "For a Turtle Soup Plus One Buy, the strategy goes long after a market makes a new 20-day low but has not made a low in the prior four sessions. The entry signal is at the prior 20-day low and the stop goes at the previous day's low." The original is faster: "Turtle Soups occur in the same day rather than on the following day". NetPicks says its own stop order lasts only that day, under either of its timings, and TradingMarkets does not say how long the Plus One entry stands.

Those are the published rules as their sources state them. Whether they make money is not something we have measured.

Why is it called turtle soup?

The name is a joke at the expense of the original Turtle traders. Alchemy Markets tells the story: the system was "created by Richard Dennis and Bill Eckhardt in the 1980s", using "an indicator (Donchian Channels) to track the highest high or lowest low over a 20-day or 50-day period", and they "would enter a trade when a new high or low was detected". When those breakouts fail, the trapped traders are the soup.

Who gets the credit for the inversion changes from page to page. Alchemy Markets names Linda Bradford Raschke alone, FXOpen names her and Connors together, and NordFX credits Dennis and Eckhardt with the breakout system without naming anyone for the fade.

What is the ICT turtle soup, and how is it different?

It keeps the idea and drops the number. FXOpen is explicit: "In the ICT strategy, forex traders apply it on intraday charts using liquidity and order flow, not fixed 20-day breakout rules." Alchemy Markets says where the weight went instead: "Instead of focusing rigidly on 20-day or 50-day highs and lows, you'd place more emphasis on liquidity sweeps, market structure shifts (MSS), and lower timeframe confirmation."

	Original	ICT version
What defines the level	A new 20-day high or low, with the previous one at least four days back	A key swing high or low on a higher timeframe, with no stated lookback
How you enter	A stop order 5 to 10 ticks inside the old extreme, good for one session	Discretionary, after a close back inside or a market structure shift
Where the stop goes	1 to 2 ticks beyond today's extreme	Beyond the sweep, with no published distance
Where it is traded	Daily charts, with an intraday variant on the same daily level	Intraday charts, "using liquidity and order flow"

The original carries numbers for its entry and its stop, so most of it is testable as written, though you still have to set the exit and what counts as a marginal break. The ICT version needs three decisions from you first.

How do you identify a turtle soup setup?

Write down three things, because the published versions answer them differently or not at all.

The lookback comes first. The original fixes it at 20 days with the four-day separation rule. The ICT pages say to mark key swing highs and lows and give no period, so that number is yours.

Then the size of the break. NetPicks leans on it loosely: a deeper poke through the old level "adds to the trade setup", and "Marginal breakouts may be ignored". Neither phrase puts a figure on marginal, so set one and keep it for the whole test.

Then what confirms the failure, which is where the same chart gives different answers:

• A market structure shift, with no close test stated. Alchemy Markets mentions the shift throughout and only describes a close inside its chart examples.
• Both, but not in the same order. FXOpen wants the close on the sweep candle itself, "Price should sweep the level with a wick that extends beyond it, then close back inside the prior range on the same candle", then a market structure shift on the lower timeframe. innercircletrader.net, a tutorial site signed ICT Trading that credits the setup to Michael Huddleston, runs that order too: "The diagnostic is the body close. If price closes past the level, the breakout is real. If the wick pierces but the body closes back inside, the Turtle Soup is in play". NordFX runs it the other way round: its numbered steps put "Confirm Market Structure Shift" before the entry, and the entry itself comes "once the price closes back above the support level", with bullish candlestick patterns or other indicators as confirmation on top of both.
• Signs rather than a test. Opofinance lists "Quick rejection of the breakout level", reversal candlestick patterns and "Lack of follow-through in the breakout direction", and defines none of the three, so there is nothing in it you could apply the same way twice.

"Same candle" matters in FXOpen's version. A close back inside on the sweep candle and one three candles later are both a close back inside in plain English, and they are different setups on the same chart. The original rule set sidesteps the question by using a resting stop order instead of a close.

Is turtle soup the same as a liquidity sweep or a false breakout?

No. The sweep and the false breakout are the ingredients, and turtle soup is one named recipe built from them. A liquidity sweep is price running past a marked level and taking out the orders resting there. A false breakout is the general case of a break that does not hold, whatever defines the level. Turtle soup adds the claim about where the level comes from, either the 20-day extreme or a higher-timeframe swing, plus an entry mechanism.

The Donchian channel that marks those 20-day extremes has its own rules, including what window it measures and whether the current bar counts. Settle that before you trust the lines on your chart.

What timeframe is it traded on?

The original is a daily-chart setup. The ICT versions use a higher timeframe for the level and a lower one for the trigger, with no agreement on which pair. NordFX marks levels on "the daily or four-hour chart" and watches for the sweep on "15-minute or 1-hour charts". Alchemy Markets names "the weekly, daily, and 4-hour charts" for the zones, then drops lower "to confirm whether the sweep is actually failing". innercircletrader.net says "The 15-minute is the most reliable parent chart for marking the levels" and calls the 1-minute the sharpest entry chart while warning that "it amplifies noise".

NetPicks says the original can be run intraday as well: monitor the daily 20-day extreme and manage lower down, or apply the 20-period lookback to whatever timeframe you trade.

How do you backtest turtle soup, step by step?

Write the rule first, then step forward one bar at a time so each level is marked before its sweep appears. Marking it afterwards guarantees a hit.

1. Fix the stretch of history before you look at it. ChartLabs session setup lets you pick random dates by length and period, or hand-pick a range.
2. Write the rule: lookback and separation, what counts as more than a marginal break, and which confirmation you are testing.
3. Mark the levels on the higher timeframe at the left edge of your test, before stepping forward.
4. Step bar by bar. When price runs past a marked level, log the candidate at that moment, before the next candle resolves it.
5. Apply the confirmation test exactly as written, including the candidates that fail it, and record which of those breaks turned out to be real. Without that column a sample says nothing about how often the filter said no.
6. Review each one afterwards on the chart with the markup you had at the time.

In ChartLabs you step real historical candles one at a time with the next bar hidden and place the trades by hand, and the order bar carries market, limit and stop orders with stop-loss and take-profit, which is what the original rule set needs. Notes and tags sit on the trade, so one lookback can be filtered apart from another afterwards.

For the ICT version, two built-ins in the Indicators menu, under Standard and the Smart Money & ICT heading, can mark parts of steps 4 and 5. Liquidity (Equal Highs/Lows, Sweeps) labels a Sweep when the first candle to trade past a swing high or low closes back inside it, which is FXOpen's same-candle reading. Its swings are pivots that stand clear of 5 bars on both sides by default, not a 20-day extreme, and if that first candle closes beyond the level it is not marked at all, so a close back inside three candles later never shows as a sweep. Market Structure (BOS / CHoCH) marks a structure shift only on a close beyond the last confirmed swing, never a wick, and calls a break against the trend a CHoCH. For the original 20-day levels, Donchian Channel, also under Standard, defaults to Length 20. Each one draws by its own rule, so check it against the rule you wrote before you count with it.

innercircletrader.net measures readiness in calendar time: "Beginners should expect six to twelve months of demo work before risking real capital. A hundred logged setups on the 5-minute is a reasonable benchmark before going live." The count is a fair benchmark. The calendar is not the only way to reach it, because those setups already exist in past data.

What goes wrong when people test this by hand?

Marking the swept level after the reversal is already on screen. The rest is a smaller version of the same problem: choosing the lookback per candidate, which turns every near miss into a hit; dropping the four-day separation rule when it is inconvenient, which widens the setup; switching between the close test and the structure shift depending on which one worked; and logging only the setups that reversed, so the real breakouts never reach the sample.

One more: treating someone else's success figure as a measurement. A figure quoted for this setup tends to arrive without a sample, a period, a method or an instrument, so there is nothing in it to reproduce.

Our beginner's guide to backtesting covers the loop, manual backtesting builds better traders explains what stepping bar by bar teaches that a report cannot, and the forex backtesting guide covers sessions and pair behaviour.

Test it before you trade it

The original rules carry enough numbers to check on history once you have settled the exit and the marginal break, which makes them a good place to start even if you end up trading the ICT version. Our data is sourced from Dukascopy, tick-level where it is available, and cleaned to remove anomalies, across forex, indices, commodities and crypto, on one plan at $20 a month, and the 7-day free trial needs no credit card: start the free trial.