Turtle Trading Rules: The Published System, Worked Through
Turtle trading is the rule-based futures system Richard Dennis and William Eckhardt taught a group of trainees in 1983. It buys a break of the 20-day or 55-day high, sizes each position so one average day's move equals 1% of the account, and exits on a 2N stop or an opposite 10- or 20-day break.
The Turtle experiment is the most retold story in trend following, and unusually, the rules behind it were eventually published in full. This page works from that document rather than from the story.
How it works
The source. The Original Turtle Trading Rules was released free in 2003 by a group of the original trainees through OriginalTurtles.org, with Curtis Faith signing the foreword.
The experiment. The document says the trainees were recruited in 1983 after Richard Dennis and William Eckhardt disagreed about whether great traders are born or made. Dennis advertised, received over 1,000 applications, interviewed 80, and trained a group that grew to 13 for two weeks in Chicago at the end of December 1983. Every rule below is taken from that document, read on 25 Sep 2026.
The markets. The Turtles traded liquid US futures: Treasury bonds and notes, currencies, the S&P 500, gold, silver, copper, crude oil, heating oil, coffee, cocoa, sugar and cotton. They did not trade grains or meats.
N, units and position size
N, the volatility unit. N is the 20-day exponential moving average of the true range, which the document notes is what is now called average true range (ATR). True range is the largest of today’s high minus low, high minus the previous close, and the previous close minus low.
Position size. Positions were built in units, each sized so that a move of 1 N equals 1% of account equity:
unit = (1% of account) ÷ (N × dollars per point)
Entries, adds, stops and exits
Entries. System 1 buys when price exceeds the high of the preceding 20 days by one tick, and sells short one tick below the 20-day low.
The skip rule. A System 1 signal is skipped if the previous 20-day breakout would have been a winner, with a 55-day breakout used as a failsafe. System 2 enters on a 55-day breakout and takes every signal. These are Donchian channel breaks, which the document credits to Richard Donchian.
Adding. After the first unit, another unit is added every ½ N in the trade’s favor, up to four.
Stops. Every unit’s stop sits 2 N from its entry, which is 2% of the account. When a unit is added, the stops on the earlier units rise by ½ N, so the whole position sits 2 N below the latest entry.
Exits. System 1 exits on a 10-day low (for longs); System 2 on a 20-day low. The document calls these the most difficult rules to follow, because waiting for them can mean watching a large share of an open profit disappear.
Limits. No more than 4 units in one market, 6 in closely correlated markets, 10 in loosely correlated markets, and 12 in one direction overall.
Drawdown rule. Each time the account fell 10% from its yearly starting size, the Turtles traded as if it were 20% smaller. A $1,000,000 notional account down $100,000 was traded as $800,000; a further 10% loss cut it to $640,000.
A worked example
Both examples below are the document’s own, recomputed.
Sizing one unit of heating oil. On 4 Dec 2002, N was 0.0141 for March 2003 heating oil. A contract is 42,000 gallons, so one point is worth $42,000. For a $1,000,000 account:
- 1% of the account = $10,000.
- N × dollars per point = 0.0141 × 42,000 = $592.20.
- Unit = $10,000 ÷ $592.20 = 16.89 contracts, truncated to 16.
The same trade on a $100,000 account gives 1.689, which truncates to 1 contract. That is the document’s own warning about small accounts: sizing gets too coarse to spread risk evenly.
Adding units and moving stops in crude oil. With N = 1.20 and a 55-day breakout at 28.30:
| Unit | Entry | Stop on all units after this add |
|---|---|---|
| 1 | 28.30 | 25.90 (28.30 − 2 × 1.20) |
| 2 | 28.90 (28.30 + ½ × 1.20) | 26.50 |
| 3 | 29.50 | 27.10 |
| 4 | 30.10 | 27.70 (30.10 − 2.40) |
Each add raises every stop by 0.60, so at four units the whole position exits at 27.70. If the fourth unit fills at 30.80 on a gap instead, its own stop is 28.40 while the first three stay at 27.70.
The original data
Across the 24,971 unique videos in the site’s finance search study, 13 titles refer to the Turtle system, counted with a case-insensitive match on “turtle” or “turtles” and excluding 3 titles about “turtle soup”, a separate setup name used mostly in ICT videos. The 13 come from 13 different channels.
Their median is 4,652 views, and 8 of the 13 stay under 10,000. Two pass 100,000, at 213,881 and 184,831 views. Two titles lead with money, “$100 MILLION” and “Made Millions”, while the number the rules actually turn on, 1% of the account per N, appears in none of them.
The published rules also carry a performance claim, and it is reported, not audited. The document says the Turtles earned an average annual compound rate of return of 80% over four years. It also says many of the trainees did not make money, because they did not follow the rules.
When it fails
Most breakouts are losers, by design. The document says so directly: most breakouts do not become trends, so most Turtle trades lost money, and the system depended on a few large winners. A trader who skips signals after a run of losses can easily miss the one that pays for them.
Long losing stretches. The document warns that many months, at times a year or two, can pass between winning periods. That is when traders change the rules, and changing them is how several of the original Turtles failed.
Gaps and fast markets. A 2N stop is a plan, not a fill. The document’s own example of October 1987 describes losses of 20% to 40% of equity in a single day for some Turtles holding interest-rate futures when rates were cut overnight. The unit limits made it smaller; they did not prevent it.
Small accounts. As the heating oil example shows, a $100,000 account could only hold one contract, so volatility sizing loses its precision. Futures contract sizes make the method hard to copy with little capital.
Crowding and time. The rules describe markets of the 1980s traded by a small group. Anyone can now run the same channel breakouts, and a published rule set says nothing about how it performs today without a fresh test.
Related
Donchian channels are the Turtle entries and exits drawn on a chart, and that page explains the offset setting that makes a channel breakable. Average true range (ATR) is the N behind every unit and every stop. And trend following is the wider approach, where a handful of big winners have to pay for a long list of small losses.
The part of this system worth copying first is the sizing, not the entry. Set the position so an average day’s move costs the same fraction of the account in every market, and the breakout rule becomes something you can survive long enough to judge.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.