What is the Turtle trading strategy?

Turtle trading is a complete mechanical system: it says what to trade, how much, when to enter, where the stop goes and when to take profits, leaving almost nothing to judgment. The entries are channel breakouts in the style of Richard Donchian, known as the father of trend following: buy when price exceeds the highest high of the past 20 or 55 days, sell short when it breaks the lowest low.

The drawing at the top of this page shows one System 1 trade: a break above the 20-day high, a 2N stop below the entry, and an exit when the trend ends and price falls through the 10-day low. It is a slower cousin of the breakout trading strategy. Any edge comes from holding the few big trends, not from being right often.

N
The Turtles’ name for volatility: a 20-day exponential average of the true range, essentially a 20-day ATR, in price units.
Unit
One slice of a position, sized so that a 1N move equals 1% of account equity. A full position is up to four units.
System 1
A 20-day breakout entry and a 10-day breakout exit, with a filter that skips a signal after a winning breakout.
System 2
A 55-day breakout entry and a 20-day breakout exit, with every signal taken.

The Turtle experiment: Richard Dennis vs. William Eckhardt

In mid-1983, Chicago commodities trader Richard Dennis and his long-time friend and partner William (Bill) Eckhardt argued over whether great traders are born or made. Dennis believed trading could be taught; Eckhardt thought aptitude decided it. To settle it, they advertised for trading apprentices in Barron’s, The Wall Street Journal and The New York Times.

According to the rules document later published by former trainees, more than 1,000 people applied, 80 were interviewed and 13 made the first class. They trained in Chicago for two weeks at the end of December 1983 and started trading small accounts in January; a second group was trained a year later. The name came from Dennis, back from Asia: he said they would grow traders the way turtle farms in Singapore grow turtles.

The trainees agreed to keep the rules secret for ten years. In 2003, Curtis Faith and other original Turtles published The Original Turtle Trading Rules free online, and Faith described the method in his 2007 book Way of the Turtle. Accounts differ in detail, and a famous group’s results are not typical: by the rules’ own account, Turtles who did not follow them consistently were dropped from the program.

The Turtle trading rules: System 1 and System 2

The published rules contain two related entry systems, and each trader chose how much of the account to put in each. A breakout means price trades past the high or low of the preceding 20 or 55 days.

The Turtles bought the moment price traded through the breakout, during the day, without waiting for the close; if a market gapped open through the level, they entered on the open. Hypothetical System 1 trade, as in the top drawing: the 20-day high is $80.00, price trades at $80.01 and you buy. Weeks later the trend fades and price drops through the 10-day low at $88.00, and every unit is sold there. Just as often, or more often, the same breakout reverses at once and the 2N stop takes you out.

  • System 1 entry: buy one unit when price exceeds the 20-day high by one tick, or sell short one tick below the 20-day low.
  • System 1 filter: skip the signal if the last breakout would have been a winner. A breakout counted as a loser if price moved 2N against it before a profitable 10-day exit.
  • Failsafe: when a System 1 signal was skipped, enter at the 55-day breakout so a major trend is not missed.
  • System 1 exit: sell all units on a 10-day low (cover shorts on a 10-day high).
  • System 2: enter on a 55-day breakout and take every signal; exit on a 20-day low (a 20-day high for shorts).
  • Stop: 2N from the entry on every trade. Whichever comes first, the stop or the exit breakout, ends the trade.

N: the Turtles’ volatility measure (a 20-day ATR)

N is the average daily range of a market, gaps included. The rules define it as the 20-day exponential moving average of the true range, known today as the ATR. The true range is the largest of three distances: high minus low, high minus the previous close, and the previous close minus low.

The formula: N = (19 × previous N + today’s true range) ÷ 20, starting from a simple 20-day average. Hypothetical: N is $2.00 and today’s true range is $3.00, so the new N is (19 × 2.00 + 3.00) ÷ 20 = $2.05. One wide day nudges N instead of jolting it.

On TradingView, the built-in Average True Range indicator with Length set to 20 and its default RMA smoothing computes the same average. The drawing below uses the more common 14-period ATR and a 1.5 × ATR stop; the Turtles used the 20-day version and a 2N stop. Either way, the stop widens when the market gets wilder, and the position shrinks to match. The ATR guide covers the indicator in detail.

Turn your phone sideways to read the drawing. Tap to close.

A hand-drawn chart with candles on top and an ATR (14) line panel below that rises as the candles get bigger; on the price chart, an entry line and a stop placed 1.5 × ATR below it, marked by a bracket labeled 1.5 × ATR.

Turtle position sizing: the 1% unit, the 2N stop and adds every ½N

Units were sized so that a 1N move equals 1% of the account: unit = 1% of equity ÷ (N × dollars per point). For a stock, one point is $1 per share, so the formula becomes 1% of equity ÷ N, in shares. The drawing below works it through: a $100,000 account, 1% = $1,000, N = $2.00, unit = 1,000 ÷ 2.00 = 500 shares.

Watch the trap: the 1% is the value of a 1N move, not the risk. The stop sits 2N away, so a full stop-out on one unit costs 2%. Hypothetical: you buy 500 shares at $80.00 on the breakout; the stop is $80.00 − $4.00 = $76.00, and hitting it loses $2,000.

Adds came every ½N from the previous fill, up to four units in one market. With N at $2.00 that means $81.00, $82.00 and $83.00, the ladder in the drawing. Each time a unit was added, the stops on the earlier units rose by ½N, so the whole position sat 2N below the latest add: $77.00, then $78.00, then $79.00. Adding only as price moves your way is pyramiding, an idea also found in Jesse Livermore’s trading rules.

The failure math: if price reverses right after the fourth add and hits $79.00, the four units lose $1, $2, $3 and $4 a share, $5,000 in total, or 5% of the account, more if it gaps. The rules also capped risk across markets: 4 units in one market, 6 in closely correlated markets, 10 in loosely correlated ones and 12 in one direction. A position size calculator helps with the share count.

Turn your phone sideways to read the drawing. Tap to close.

A hand-drawn notebook calculation on the left: account 100,000, so 1% = 1,000; N = 2.00; unit = 1,000 ÷ 2.00 = 500 shares. On the right, a stepped ladder with the entry, then +1.00, +2.00 and +3.00 (every ½N), unit boxes 1 to 4, and the stop raised with each add.

Why trend following has long losing streaks and deep drawdowns

The rules say it plainly: most breakouts do not turn into trends, so most Turtle trades lose. The system survives only if the few long trends earn far more than the many small losses, and the rules warned that most of a year’s profit might come from two or three trades. Skip one of those, and the losses may go uncovered.

Losses also cluster. In a sideways market, breakout after breakout fails. Hypothetical: five false breakouts in a row at $2,000 each cost $10,000, 10% of a $100,000 account. The Turtle rule for that moment was to cut the account size used for sizing by 20% for every 10% drawdown: you trade as if the account were $80,000, so at the same N the next unit drops from 500 to 400 shares.

Winners hurt too. A 10-day or 20-day exit waits for a new low, so the system gives back a large part of an open profit before it sells; that is the price of staying in for the big move. And the drawdown math in the drawing is unforgiving: a 10% loss needs 11% to recover, 25% needs 33%, and 50% needs 100%. See risk management for limits that keep a streak survivable.

Turn your phone sideways to read the drawing. Tap to close.

A hand-drawn equity curve that peaks, falls with the drawdown shaded, and recovers, beside three bars: −10% needs +11%, −25% needs +33%, and −50% needs +100% to get back to the peak.

How to adapt Turtle trading to stocks and ETFs today

The Turtles traded liquid US futures: bonds, currencies, metals, energy, softs such as coffee and sugar, and a stock index, but not grains or meats. Futures are leveraged, and those markets often move independently. A stock account works differently, so an honest adaptation changes a few things, and every change needs testing with small size first.

If you day trade on 5-minute to 1-hour charts, a Turtle-style system is a different animal: breakouts are noisier and costs weigh more. UpTickFlow, our invite-only TradingView indicator, is not a Turtle system and does not compute N or channel breakouts. It waits for a closed candle at a level, draws the entry, stop and target before the trade, scores the setup from 0 to 100 and stays quiet in chop: a second check, not a scanner.

  • Charts: use daily candles on liquid stocks and ETFs. On TradingView, add Donchian Channels (Length 20 is the default) for the 20-day high, a second copy at Length 10 for the exit, and Average True Range at Length 20 for N. System 2 uses 55 and 20.
  • Confirmation: the original bought at the first tick past the high. You can require a daily close above it instead, which skips breakouts that only wick through, at the cost of a worse price. That is your change, not the original rule.
  • Location: the 20-day high is the level, but check what sits above it. A breakout just under a multi-month resistance on a higher timeframe has little room. The original rules ignored this, so treat it as your own filter.
  • Size within your buying power: with the numbers above, one 500-share unit at $80.00 is $40,000, and four units come to about $163,000, more than the account. At 0.5% per N, a unit is 250 shares, a 2N stop risks 1%, and four units total about $81,500.
  • Diversify: ETFs that track different asset classes (stocks, bonds, gold, commodities) trend at different times. Ten tech stocks are one bet, so apply the correlated-unit limits to them as a group.
  • Gaps: stocks gap on earnings and overnight news, and a stop order cannot prevent a gap. Check the earnings date before every entry.
  • Shorts: the rules sell 20-day lows too. Shorting stocks adds borrow costs and squeeze risk; see short selling.

When the Turtle strategy fails

Turtle trading fails often, and the failure looks exactly like the setup. The worst environment is a range like the one in the drawing below: every push to the top prints a fresh 20-day high, the trend follower buys, and price falls back. The range traders in the drawing sell exactly where the breakout trader buys.

Dennis told Jack Schwager in Market Wizards that the rules were not the hard part; sticking to them when things go badly was. Most of the failures below are tests of that.

  • Chop: repeated false breakouts, each a small loss, can last for months.
  • Gaps: a 2N stop is a plan, not a fill. Overnight news can open a stock far below it.
  • Giving back: the slow exits return a big part of open profits, and a trader who starts taking profits early removes the large wins the system depends on.
  • The filter can backfire: System 1 skips the signal after a winner, and the skipped breakout can be the big trend. The 55-day failsafe exists for that reason.
  • Drifting from the rules: skipping signals after a losing streak, or doubling size to win it back, turns a tested system into an untested one.
  • No promise of an edge: the rules are public, markets change, and a backtest of the past says nothing certain about next year.

Turn your phone sideways to read the drawing. Tap to close.

A hand-drawn horizontal range between a support zone at the bottom and a resistance zone at the top, with price bouncing between them; buy triangles near support, sell triangles near resistance, the middle third lightly shaded as a no-trade zone, and dotted stops just outside the range.