Richard Dennis and the Turtles: what happened when a trading system was given away
Most trading stories end with a secret being kept. This one ends with the secret being printed. In 1983 a Chicago futures trader recruited twenty-three people with almost no trading experience, taught them a complete mechanical system in two weeks, handed them his own money and set them loose. Within a decade the rules had leaked; by 2007 they were published in full by one of the students. The system still worked on paper. Almost nobody could run it — including, in the end, the man who wrote it.
KEY TAKEAWAYS
- The bet was about teachability, not about a strategy. Dennis believed trading could be taught; his partner William Eckhardt believed it could not. The Turtles were the experiment that settled it.
- Volatility set the position size, not the trader. A unit was sized so that one average daily range equalled 1% of the account, and the stop sat two ranges away — so every unit risked 2%, in every market.
- The system got smaller after losses, automatically. Every 10% of drawdown cut the notional account by 20%. At a 50% drawdown a unit risked about 0.66% — roughly a third of the risk allowed at the high.
- The teacher stopped following his own rules. Dennis reportedly lost about $50m across 1987–88, stopped managing client money in spring 1988, and in 1990 his firm settled investor complaints that he had failed to follow his own rules, for more than $2.5m.
- Publishing the rules did not destroy them. Dennis said in 1989 that he could print them in a newspaper and nobody would follow them. That claim has aged better than the system’s returns.
Who was Richard Dennis?
Richard Dennis was born in Chicago in January 1949 and was working as an order runner on the floor of the Chicago Mercantile Exchange at seventeen. Too young to trade a pit himself, he hired his father to stand in it and traded through him.
The capital story is usually told wrong. He borrowed $1,600 from his family; $1,200 of it bought a seat at the MidAmerica Commodity Exchange, leaving $400 to actually trade with. By 1970 the $400 was $3,000. By 1973 it was over $100,000. In 1974 he made about $500,000 in soybeans and was a millionaire a few months short of his twenty-sixth birthday. By 1976 The New York Times Magazine was calling him the Prince of the Pit.
One detail matters more than the arithmetic. Dennis rose in the commodity markets of the 1970s, and those markets were unusually kind to his method. The 1972 Great Grain Robbery — Soviet buyers quietly taking roughly 30% of the American wheat crop in a matter of weeks — was followed by years of crop failures and inflation. Prices trended hard, in both directions, for years at a time, while Dennis bought new highs and held them. Which is the first thing to hold in mind when reading anybody’s track record: a result is produced by a method and a market, and only one of the two is under your control.
What was the Turtle experiment?
Dennis thought trading was a skill that could be written down and taught to anyone. His trading partner William Eckhardt thought Dennis had a talent that could not be transferred. Rather than keep arguing, they ran the test.
Twenty-three people were chosen — twenty-one men and two women — across two intakes, December 1983 and December 1984. Their backgrounds included a game designer, an accountant, a security guard and a professional gambler. Most had little or no futures experience, which was the entire point.
The training lasted two weeks. In January 1984 each trainee got a small account and permission to trade up to twelve contracts per market for a month. Those who followed the rules were then handed between $250,000 and $2m of Dennis’s own money to manage. They were called Turtles; where the name came from is genuinely disputed, and since no version is documented we are not going to pick one.
Five years later the group had reportedly earned an aggregate profit of about $175m. That figure is repeated everywhere and audited nowhere, so read it as reported rather than confirmed. Several Turtles — Jerry Parker, Liz Cheval, Paul Rabar, Tom Shanks, Howard Seidler, Jim DiMaria — went on to manage money professionally for decades afterwards.
“I always say that you could publish my trading rules in the newspaper and no one would follow them. The key is consistency and discipline.”Richard Dennis, in Jack D. Schwager, Market Wizards (1989), answering whether he worried about teaching his methods to other people.
He said this four years before the rules began to leak and eighteen years before they were published in full. It is the rare boast that was later put to the test.
What did the Turtles actually learn?
Popular retellings compress the Turtle system into an entry signal: buy a twenty-day high, sell a twenty-day low. That is in there, and it is the least important part.
Eckhardt taught that a complete system has to answer five questions before the market opens, so that nothing has to be decided under pressure: what to trade, how much, when to enter, when to get out of a loser, and when to get out of a winner. Traders obsess over question three. The Turtle rules spend most of their weight on two and five.
The engine is a number the Turtles called N: a twenty-day average of the true daily range, which today you would call the average true range, or ATR. It answers one question — how far does this market usually travel in a day?
Everything else is denominated in N:
- Position size. A unit was sized so that a move of one N equalled about 1% of the account. Quiet market, bigger position; violent market, smaller position.
- The stop. Two N from entry — which, with that sizing rule, makes a stopped-out unit cost 2% of the account, whether the market is gold, sugar or the Swiss franc.
- Adding. Every half-N in your favour, add a unit and raise the stops on the ones you hold. Four units maximum in any one market.
- Exits. No profit targets, ever. Get out when price breaks the ten-day extreme against you, or the twenty-day on the slower version.
Read that sizing rule against what Lesson 42 on position sizing teaches — that the stop decides the size — and you can see exactly what the Turtles added. Lesson 42 leaves one question open: where should the stop distance come from in the first place? The Turtle answer is that you do not choose it. The market’s own volatility hands it to you, and the size falls out of the division. The consequence is one most people get backwards: when a market turns wild, the Turtle position gets smaller. A trader who keeps buying the same quantity while volatility doubles has doubled the bet without ever deciding to.
Then come the caps: four units in one market, six across closely correlated markets, ten across loosely correlated ones, and twelve in any one direction. Twelve units at 2% each is the number worth remembering.
Twenty-four per cent of an account exposed at one time is a great deal of risk — and it was the maximum the rules would permit, reached only when every market was trending at once. Getting back from a 24% loss requires a gain of 31.6%. The cap exists because somebody did that arithmetic in advance, on a calm afternoon, instead of in the middle of a bad week.
Why does the system shrink when you lose?
Here is the rule that separates the Turtle system from almost everything sold as a trading strategy today.
For every 10% the account fell from its high, the Turtles were required to cut their notional trading capital by 20%. Lose $10,000 on a $100,000 account and you size every trade from then on as though you had $80,000. Fall another 10% and you cut again. Normal size returns only when the account makes a new equity high. Compounded, that produces a ladder nobody has to think about:
At a 50% drawdown a unit is risking 0.66% against the 2.00% it was allowed at the high — about one third. The system has quietly moved from aggression to survival, and no act of willpower was required to make it happen.
Compare that with what a losing trader tends to do: widen the stop, add to the position that is already wrong, size up so that one good trade repairs the damage. That is not a character flaw peculiar to beginners — it is what pressure does to almost everyone, which is precisely why the Turtles were not permitted to decide.
The principle reaches well beyond this one system: risk control that depends on you feeling calm will fail on the day you are not. Risk control built into the arithmetic does not care how you feel.
PRACTICE CORNER
The Turtle rule was that the stop distance is read off the market’s volatility rather than chosen by the trader — and that it is set the moment the position opens, not later. ATR is a standard indicator on every charting platform, and every major exchange lets you attach the stop on the same screen where you open the trade, while you are still calm.
Referral links — they never change our assessment. Education only; most retail traders lose money.
What did it cost the teacher?
A profile that stops at the $175m teaches nothing. Dennis had been managing pools of outside money alongside the experiment. In the Black Monday crash of October 1987 he reportedly lost about $10m; across 1987 and 1988 the reported total was around $50m. His clients were hurt badly enough that in the spring of 1988 he stopped managing outside money altogether. He returned to fund management in the 1990s and closed those operations too, after losses in the summer of 2000.
Then comes the detail that makes this profile worth writing. In November 1990 his firm settled investor complaints for more than $2.5m, without admitting or denying wrongdoing. The complaint was not that his system was flawed. It was that he had failed to follow his own rules.
The man who had just shown that a mechanical system could be taught to strangers in a fortnight — and who had told an interviewer the rules were so hard to follow he could publish them safely — faced a complaint that he had not followed them himself. On that account, he became his own best evidence.
There is no audited drawdown percentage for his funds in the public record, only the reported dollar losses, so we will not print one. But the arithmetic of recovery is fixed: a 24% loss needs +31.6% to repair, a 50% loss needs +100% — and under the drawdown rule the account digging out is allowed only a third of its original risk per trade. The hole is deeper than it looks, and the shovel is smaller.
Where the legend and the record disagree
The story is usually told as a fable: twenty-three ordinary people were handed a system and all got rich. The record is more useful, in three places.
The results diverged sharply. Every Turtle received identical rules, from the same teachers, in the same room. Some built firms that ran money for decades. Others were cut from the programme. Individual performance figures are contested between former Turtles and their biographers, and we are not going to adjudicate a dispute between living people using partisan sources. What every side agrees on is the shape: same rules, very different outcomes. That divergence is the experiment’s most valuable finding, and the fable erases it.
The aggregate profit is a reported number. The $175m has no audited public source. It may well be right; it is not established, and repetition is not evidence.
Publication and decay are two separate things. Backtests of the published rules show performance falling off sharply after 1986. It is tempting to conclude that giving the rules away killed them; the duller explanation is likelier, that the sustained commodity trends of the 1970s did not repeat, and a system which only pays during long trends stops paying when they stop.
Where this stops working
2% per unit is not a recommendation. It only works inside the rest of the machine — the unit caps, the correlation limits, the drawdown rule, a portfolio spread across a dozen unrelated markets. Lifted out and put on a single crypto position, 2% is simply a bigger bet than a beginner should make. Our default stays at 1% per idea, as in Lesson 42.
Diversification was doing quiet work. The Turtles traded bonds, currencies, metals, energy and softs at once, assuming these would not all move together. Crypto offers far less of that: when Bitcoin falls hard, most of the market falls with it. Four “different” altcoin positions are frequently one position wearing four names.
A stop is an instruction, not a guarantee. The Turtles traded deep futures markets. In a thin one a 2N stop fills where it fills, which in a fast move can be well past your level — see slippage and liquidity and spread.
The win rate is uncomfortable. Most breakouts fail. This is a method that loses more often than it wins and relies on a few large trends to pay for the rest. If you cannot sit through a long losing stretch without changing the rules, it will not save you — and that, on Dennis’s own account, describes nearly everyone.
Common mistakes when learning from the Turtles
Copying the entry and ignoring the sizing. The twenty-day breakout is the famous part and the disposable part. The volatility-normalised unit and the drawdown rule are what made the results possible.
Reading the story as proof that anyone can do this. As written, the experiment shows that a system can be transferred and that outcomes still scatter widely. Those two findings sit together uncomfortably, and the second is the one likelier to apply to you.
Treating the teacher’s record as a warranty on the method. The rules are worth studying. The biography is not an endorsement.
FAQ
How many Turtles were there?
Twenty-three: twenty-one men and two women, recruited in two intakes beginning December 1983 and December 1984. Figures of thirteen or fourteen usually refer to one intake only.
Are the Turtle trading rules public?
Yes. They leaked in pieces during the 1990s and were published in full by original Turtle Curtis Faith in Way of the Turtle (2007).
How much did each Turtle trade risk?
About 2% of the account per unit. A unit was sized so that one N — a twenty-day average true range — equalled roughly 1% of the account, and the stop sat 2N from entry.
Does Turtle trading still work?
Backtests of the published rules show performance dropping sharply after 1986. The sizing and drawdown principles have aged far better than the breakout signals.
Did Richard Dennis lose money?
Yes. He reportedly lost about $10m in the October 1987 crash and about $50m across 1987–88, stopped managing client money in spring 1988, and in November 1990 his firm settled investor complaints that he had failed to follow his own rules for more than $2.5m, without admitting or denying wrongdoing.