Turtle trading: the 20-day and 55-day breakout rules, tested on ETFs

Turtle trading is the set of trend-following rules that Chicago futures trader Richard Dennis taught to applicants with little or no trading experience in the 1980s. This article lays out the published original rules exactly, then shows our calculation of applying only the entry and exit lines to US ETFs, and the numbers behind why the results were weak. Figures are as of September 2026. Values recalculated every month are on the Turtle breakout backtest page in Questions and answers.

The 1983 bet

In mid-1983, Dennis and his longtime colleague William Eckhardt argued over whether great traders are born or made. Dennis thought trading could be taught, and Eckhardt thought innate talent decided it. They agreed to recruit people, train them and give them real accounts. Ads in Barron's, the Wall Street Journal and the New York Times drew more than 1,000 applicants. Dennis interviewed 80, picked 10 and added 3 people he knew, so the first class had 13. They trained in Chicago for two weeks at the end of December 1983, started with small accounts the next January, and from February most received accounts of $500,000 to $2 million. A second class trained in December 1984, bringing the total to 23 (Wikipedia). The name came from Dennis, back from visiting a turtle farm in Singapore, saying he would grow traders the way they grew turtles there (a 1989 Wall Street Journal article).

All the performance figures in circulation are claims by participants or the press. The original rules document says the Turtles earned an average annual compound return of 80% over the next four years, and there are reports that the Turtles' combined profit reached $175 million by the end of the experiment. This site has verified neither. After heavy losses in 1987 and 1988, Dennis stepped back from managing client money in the spring of 1988.

The rules stayed private for a long time. In 2003 several original Turtles released a document called The Original Turtle Trading Rules for free at OriginalTurtles.org, and Curtis Faith, the youngest Turtle, published Way of the Turtle in 2007. The rules below follow that document.

The original rules

• Markets: 21 liquid US futures markets: 30-year and 10-year Treasuries, 90-day Treasury bills, Eurodollars, six currencies (Swiss franc, Deutschmark, British pound, French franc, Japanese yen, Canadian dollar), gold, silver, copper, crude oil, heating oil, unleaded gasoline, coffee, cocoa, sugar and cotton. The only stock index was the S&P 500.
• N (volatility): the 20-day average range. The daily true range (TR) is the largest of high minus low, high minus the previous close, and the previous close minus low. N is updated daily as N = (19 × previous N + today's TR) ÷ 20, and the first N is the simple average of 20 days of TR.
• Unit: the number of contracts sized so that a 1N price move changes the account by 1%. Unit = 1% of the account ÷ (N × dollars per point).
• System 1: entered one unit long when price exceeded the prior 20-day high by a single tick, and one unit short when it broke the prior 20-day low. It skipped a breakout if the previous breakout would have been a winner, judged by whether a profitable exit came before a 2N stop, whether or not that trade was actually taken. After a skip, it took the 55-day breakout when one came. Longs exited at a 10-day low and shorts at a 10-day high.
• System 2: entered long on a break of the prior 55-day high and short on a break of the 55-day low, taking every breakout whatever the previous result. Longs exited at a 20-day low and shorts at a 20-day high.
• Entry timing: the Turtles did not wait for the close. They entered as soon as price touched the breakout level during the day, or at the open if the market opened beyond it.
• Adding units (pyramiding): one unit was added each time price moved ½N in the trade's favor from the previous fill, up to four units in one market.
• Stops: a position was closed when price moved 2N against the entry. Since 1N is 1% of the account, a unit's loss was capped at 2%. When units were added, the stops on earlier units moved up by ½N, so all stops gathered 2N from the latest unit. An alternative whipsaw stop cut at ½N and re-entered if price returned to the original entry.
• Limits: 4 units in one market, 6 in closely correlated markets, 10 in loosely correlated markets, and 12 in one direction (long or short) overall.
• Account reduction: for every 10% lost from the starting account, the account size used to compute units was cut by 20%, and restored once the year's starting equity was regained.

The original document stresses taking every signal, because most of a year's profit might come from only two or three large winners. It is a structure in which rare large trends pay for many small losses, and trading many markets at once raises the chance of catching those rare trends.

Breakouts come from the Donchian channel

The band between the N-day high and the N-day low is called a Donchian channel. Richard Donchian (1905 to 1993) ran Futures, Inc., known as the first public futures fund, from 1949, and he is known for the four-week rule: buy above the four-week high and sell below the four-week low. Four weeks is about 20 trading days, so the Turtle System 1 entry line is the channel's upper line and the System 2 exit line is its lower line.

The two trades in the figure (bought on a break above the 20-day channel and sold on a break below it) show the shape of a breakout rule. The October 2022 breakout came in a bear-market rally and ended at −0.8% after 36 trading days. The January 2023 breakout lasted 137 trading days and returned +29.5%. The second trade's exit signal came on a close 4.6% below the top. Confirming that the trend had ended took that much pullback.

Results on QQQ and SPY (as of September 2026)

The Questions and answers calculation takes only the entry and exit lines from the original rules. When the close is above the prior 55-day high, the whole account buys at the next trading day's open, and when the close falls below the prior 20-day low, everything is sold at the next open (System 2). System 1 uses 20 and 10 days. Volatility sizing, adding units, the 2N stop, the skip after a winner and short selling are not included. $10,000 went in on January 3, 2000, with dividend-reinvested prices, a 0.15% fee on each buy or sell, and no interest on cash. The two right-hand columns were calculated separately by the operator from the same trades: the ETF's own annual gain on the days the rule held cash and on the days it was invested.

QQQ System 2 made 63 round trips, and 54% ended in a gain. The average winner gained +7.3% and the average loser lost −4.8%. 18 of the 63 ended within a month (21 trading days). By decade, it beat holding only in 2000 to 2009 (−2.7% a year against −6.9%), and returned 3.3% a year in 2010 to 2019 (holding 17.9%) and 10.0% from 2020 onward (holding 21.0%). SPY System 2 trailed even in 2000 to 2009, at −1.7% a year against −1.1% for holding. System 1 traded more often, and 57 of its 117 QQQ round trips ended within a month.

Why it was weak on a single ETF

What we confirmed with numbers and what remains an explanation we could not put in numbers are listed separately.

• The index rose more while the rule sat out (calculated): on the days QQQ System 2 held cash (52% of all days), QQQ rose 10.3% a year, and on the days it was invested, 7.0% a year. For SPY System 2 the figures were 11.8% a year on cash days and 4.1% on invested days. Of QQQ's 20 biggest up days since 2000, 19 fell on days System 2 held cash, and so did 18 of its 20 biggest down days. Large moves clustered in stretches that mixed declines and rebounds, and a rule waiting for a 55-day high missed the rebounds in those stretches.
• The result was the gain on invested days minus costs (calculated): on the days QQQ System 2 was invested, QQQ rose 140% in total. With 63 round trips of fees (about 0.3% each) taken out, $10,000 comes to about $19,900. The actual result was $19,916. What the rule missed was the 286% gain on the days it held cash.
• Gains were cut short (calculated): the 20-day-low exit was often hit by pullbacks within an uptrend. The largest gain in QQQ System 2 was +21.0%, and winners were held for 76 trading days on average. On the same QQQ, the 50/200 cross had a largest gain of +58.0%, and its winners were held for 371 trading days on average.
• The original sizing becomes leverage on a single ETF (calculated): since 2000, QQQ's median N was 1.56% of price and SPY's was 1.17%. A unit that risks 1% of the account per 1N comes to about 64% of the account for QQQ and about 85% for SPY, and the four-unit limit for one market comes to about 257% and 341%. The original sizing cannot be followed without borrowed money, so this calculation simplifies to fully invested or fully in cash.
• Diversification and short selling are missing (explanation): the Turtles traded both long and short across 21 markets that move differently. When one market went sideways, trends in other markets could cover the losses, and in a market in a long decline a short position could make money. A long-only calculation on one ETF has none of this structure. This site does not run short-selling or multi-market calculations, so it cannot show this difference in numbers.

Related concepts and search terms

The Turtle name is famous, but each part of the rules is a widely used idea.

• Trend following, managed futures: trading long and short with the trend across many futures markets. Firms that do this are called CTAs (commodity trading advisors). The Turtles are a well-known example.
• Donchian channel: a band made from the N-day high and low. A break above the upper line and a break below the lower line are the signals.
• ATR (average true range): the average range introduced by J. Welles Wilder in his 1978 book New Concepts in Technical Trading Systems. The Turtles' N is a 20-day ATR.
• Volatility position sizing: smaller positions in more volatile markets, so that each market carries the same risk.
• Pyramiding: adding to a winning position in steps. The Turtles added at ½N intervals up to four units.
• Moving-average trading: rules in the same trend-following family that use moving averages are covered in Moving-average trading.
• New-high breakouts: the approach of an earlier trader who used breakouts to new highs as his trading trigger is in Livermore's trend breakout.

Search terms: turtle trading rules, Richard Dennis, Donchian channel breakout, 55-day breakout, ATR position sizing, Way of the Turtle, The Complete TurtleTrader, four-week rule.

Drawbacks and risks

• Only about half the trades win and large gains are rare, so losing stretches are long. QQQ System 2's max drawdown of −60.2% came between March 2000 and April 2003.
• Entering on an intraday breakout can fill at a worse price than the breakout level. This calculation simplifies to fills at the next day's open.
• Frequent trading lets costs and taxes cut deeply into results. QQQ System 1 had 234 fills after 2000.
• Futures trade on margin, so they move amounts larger than the account. Using the original unit size on an ETF would mean buying 2.6 to 3.4 times the account at the limit, as shown above.