The short answer
Turtle Soup fades failed 20-period breakouts. We tested the rules on gold:
- Daily Turtle Soup Plus One, since 2015: 64 trades, 35.9% winners, −$218 per ounce after costs, profit factor 0.71.
- Hourly 20-bar Turtle Soup, two years: 487 trades, 40.5% winners, −$890 per ounce, profit factor 0.75.
- The opposite trade (buy the 20-bar breakout, fixed 2R target): 1,141 trades, −$516 per ounce.
Fading false breakouts lost on both timeframes, and following the same breakouts with a fixed target lost too. The trigger alone did not carry an edge on gold; what happened after the entry (the exit) mattered more.
Turtle Soup is a published, rule-based strategy, so our test is closer to the original than our tests of discretionary ideas like CRT or the Judas swing. We still had to choose the exit, which the original rules leave partly to the trader. We state every choice below.
Where Turtle Soup comes from
The Turtle traders of the 1980s bought 20-day highs and sold 20-day lows. Laurence Connors and Linda Bradford Raschke, in their 1995 book Street Smarts, noticed that many of those breakouts failed, and named a strategy for trading the failure: Turtle Soup. The idea is the same one modern traders call a liquidity sweep: stops cluster beyond obvious extremes, get triggered, and the move runs out of fuel.
Exact rules
Turtle Soup (original, used for the hourly test)
| Rule | Long (short is the mirror) |
|---|---|
| Set-up | The current bar makes a new 20-bar low |
| Age filter | The previous 20-bar low was set at least 4 bars earlier |
| Entry | Buy stop at the previous 20-bar low, valid for the next 2 bars |
| Stop | The low of the bar that made the new 20-bar low |
| Target | 2R (variant: 1R) |
| Time exit | After 24 hourly bars |
Turtle Soup Plus One (used for the daily test)
| Rule | Long (short is the mirror) |
|---|---|
| Set-up | Today makes a new 20-day low and closes at or below the previous 20-day low |
| Age filter | The previous 20-day low was set at least 3 days earlier |
| Entry | Tomorrow, buy stop at the previous 20-day low (one day only) |
| Stop | The low of the set-up day |
| Target | 2R (variant: 1R) |
| Time exit | After 5 trading days |
How we tested it
Data. COMEX gold futures (GC=F) daily bars from January 2015 to October 2026 and 1-hour bars from 30 October 2024 to 30 October 2026. Execution. Buy and sell stops fill at the level, or at the open if price gaps through. Inside the fill bar we assume the usual path (up bar: open, low, high, close; down bar: open, high, low, close) and count only stops or targets reached after the fill, with the stop first if both are possible. Cost. $0.30 per ounce per round trip. Results in US dollars per ounce.
The results
Daily Turtle Soup Plus One (2015 – 2026)
| Version | Trades | Win rate | Avg per trade | Profit factor | Net | Max drawdown | 2015–2020 | 2021–2026 |
|---|---|---|---|---|---|---|---|---|
| 2R target, 5 days | 64 | 35.9% | −$3.40 | 0.71 | −$217.90 | $240.60 | −$51.80 | −$166.10 |
| 1R target | 59 | 44.1% | −$6.02 | 0.48 | −$354.90 | $392.10 | −$43.00 | −$311.90 |
The daily version traded about five times a year. Its worst year was 2025 (−$143 on eight trades).
Hourly 20-bar Turtle Soup (two years)
| Version | Trades | Win rate | Avg per trade | Profit factor | Net | Max drawdown | Year 1 | Year 2 |
|---|---|---|---|---|---|---|---|---|
| 2R target, 24 bars | 487 | 40.5% | −$1.83 | 0.75 | −$890.00 | $1,346.40 | −$336.60 | −$553.40 |
| 1R target | 409 | 47.7% | −$0.82 | 0.84 | −$337.40 | $461.40 | −$296.30 | −$41.10 |
Contrast: trading with the 20-bar breakout
| Version | Trades | Win rate | Profit factor | Net | Year 1 | Year 2 |
|---|---|---|---|---|---|---|
| Buy new 20-bar high / sell new low, 1 ATR stop, 2R | 1,141 | 36.0% | 0.96 | −$516.20 | +$453.10 | −$969.30 |
What the results say, in plain English
- Failed breakouts kept failing. When gold reclaimed an old 20-bar low, it often rolled over again and hit the stop under the new low. On H1 the stop was hit 285 times against 197 targets.
- Both sides of the trade lost with fixed targets. Fading the break lost, and following the break with a 2R target lost too (although it made money in the steadier year 1). This is what a market without a simple edge at this level looks like.
- The exit is where trend-followers made money. Our Donchian channel test traded the same kind of 20-bar breakout but exited on a 10-bar channel, letting winners run for days. That version made $2,588 per ounce over two years of hourly data. A fixed 2R target cuts off the large trends that pay for all the losers.
- The 1R version was closest to break-even on hourly data, especially in year 2 (−$41). That hints that the reversal often happens but does not travel far.
- Small daily sample. 64 daily trades in almost twelve years is too few to be confident either way.
When Turtle Soup fails
- Fast, two-way markets. A sharp dip can make a new 20-bar low, reclaim it, and then fall again before any reversal takes hold.
- Wide set-up bars. If the bar that makes the new low is huge, the stop is far and the 2R target is very far.
- Second attempts. The original rules allow repeat signals; after one failure, the next attempt at the same level often fails too.
- News spikes. A data-driven spike through the 20-bar low can reverse and re-break within minutes.
How traders adapt it
Mean-reversion traders who like the Turtle Soup idea often add:
- A trend filter, such as only fading lows when the daily trend is up. See mean reversion on gold.
- Volatility context, skipping set-ups after very wide days.
- Partial profits at 1R with a trailing stop on the rest, combining the higher hit rate of a 1R exit with some exposure to larger moves.
We did not test these, so treat them as ideas for your own testing, not as fixes.
Risk sizing
The hourly version’s average stop was about $11.89 per ounce; the daily version’s about $17. Risking $100 on the hourly version:
- $100 / ($11.89 + $0.30) is about 8 ounces (0.08 lots).
- Its worst losing run was 18 trades, about $1,800 at that size.
Use the position size calculator for each trade, and decide in advance how many consecutive losses you will accept before pausing.
Checklist before you trade a failed breakout
- Confirm the level is a true 20-period extreme and that the old high or low is at least a few bars old, as the original rules require.
- Decide the trigger: same-bar reversal (original) or next-bar entry (Plus One). They behave differently, and our daily sample was small.
- Check the trend. Fading a new low in a strong downtrend is a very different bet from fading one in a range.
- Measure the stop. If the bar that made the new extreme is wide, the 2R target may be unrealistic; consider the 1R version or skip.
- Count repeat signals. After one failed reversal at a level, record whether the second attempt is any better before you take it.
- Track the opposite trade. Our contrast test shows how easily a market can punish both fading and following the same break.
Test it yourself
Turtle Soup is published as precise rules, so you can code it exactly. Test the original and Plus One versions on your platform’s XAUUSD data, then test one change at a time: a trend filter, a different exit, or a different lookback. Keep the contrast test as well, because it shows whether you are measuring the level or just the market’s trend. The strategies tested hub compares this result with 13 other strategies.
Practise it with other traders
Fading breakouts requires patience and a thick skin. In the XAUUSD Global community, traders who trade sweeps and failed breakouts post their set-ups, compare exits and share their journals. The first month is free, then $19.99 a month.
Educational content only, not financial advice. Trading gold with leverage carries a high risk of loss.
Frequently asked questions
What is the Turtle Soup trading strategy?
It fades failed breakouts of a 20-day high or low. If price makes a new 20-day low but the previous 20-day low was set at least a few days earlier, a buy stop is placed at the old low; if price trades back above it, you go long with a stop just under the new low. Shorts are the mirror image. It was described by Laurence Connors and Linda Bradford Raschke in their 1995 book Street Smarts.
Does Turtle Soup work on gold?
Not in our tests. The daily Turtle Soup Plus One version since 2015 made 64 trades and lost $218 per ounce after costs. A 20-bar hourly version over two years made 487 trades and lost $890 per ounce. A 1R-target version lost less but still lost.
What is Turtle Soup Plus One?
A variation where the new 20-day low must close at or below the old 20-day low. The next day, a buy stop is placed at the old low. It gives the breakout a full day to fail rather than reversing within the same session.
Is Turtle Soup the same as a liquidity sweep?
They describe the same market behaviour: stops beyond an obvious high or low are triggered and the move fails. Turtle Soup defines the level precisely (the 20-period extreme) and gives exact entry and stop rules; liquidity-sweep trading is usually more discretionary.
If fading breakouts lost, does trading the breakout work instead?
Not with fixed targets in our test. Trading with the 20-bar break on hourly gold (1 ATR stop, 2R target) made 1,141 trades and lost $516 per ounce. Our Donchian channel test, which lets winners run with a trailing exit, was positive on hourly data. The exit made the difference.
How many Turtle Soup signals does gold give?
Not many on daily charts: 64 Plus One trades in almost twelve years in our test, about five a year. On hourly charts the 20-bar version triggered 487 times in two years.
Sources and further reading
- CME Group — Gold futures overview — Contract specifications and trading hours for COMEX gold
- CFTC — Commitments of Traders — Weekly futures positioning, including gold
- Investor.gov — Stop order — How stop orders work, and their limits in fast markets
External sites are independent of XAUUSD Global. Check dates and data yourself.