Turtle trading is a mechanical trend-following system for futures markets, named after a group of novice traders recruited and trained in the 1980s by the Chicago commodity trader Richard Dennis and his partner William Eckhardt. The system enters positions when prices break out of recent ranges, sizes each position according to the market's volatility, and exits when the trend reverses by a set amount. The same name is used for the experiment itself and for the trading rules, which later became public.
The experiment is often cited in debates about whether trading skill can be taught or depends on innate talent, and it helped popularise the systematic approach used by many commodity trading advisors and managed futures funds. The rules are an example of breakout trading combined with strict risk control, and they have been widely studied and tested since.
The experiment
According to accounts by the participants, Dennis believed that successful trading could be taught as a set of rules, while Eckhardt thought it depended more on personal aptitude. In 1983 they settled the argument with an experiment. They placed newspaper advertisements, interviewed applicants and selected a small group with little or no trading experience, who were trained for about two weeks and then given money from Dennis's own accounts to trade. A second group was recruited in 1984. Dennis reportedly named them after turtle farms he had seen in Asia, saying he would grow traders the way turtles were raised.
Several of the trainees went on to strong results, and a number later ran their own trading firms. The programme ended in the late 1980s. For years the rules were kept confidential, but they were published by former participants in the 2000s, and one of them, Curtis Faith, described the programme in detail in a book in 2007.
The trading rules
The published rules cover every part of a trade, leaving little to discretion:
- Markets. A diversified portfolio of liquid futures, including currencies, interest rates, metals, energy, agricultural products and equity indices.
- Entries. One system bought when the price exceeded its 20-day high and sold short when it fell below its 20-day low; a slower system used 55-day highs and lows. This is the logic of the Donchian channel.
- Exits. Positions in the faster system were closed on a break of the opposite 10-day extreme and in the slower system on a 20-day break. A protective stop was also placed a fixed multiple of volatility from the entry.
- Adding to positions. Further units could be added as the price moved in the trade's favour, up to a set maximum, with stops moved up accordingly.
- Portfolio limits. Caps applied to the number of units in a single market and in groups of closely related markets.
Volatility-based position sizing
The most influential part of the system was its method of sizing positions. Volatility was measured by a quantity called N, essentially the 20-day average true range. Each unit was sized so that a one-N move would change the account's value by about 1 percent:
- is the average true range of the market, in price points.
- Dollars per point is the change in a contract's value for a one-point move in price.
- Account equity is the capital allocated to the system.
As a worked example, take an account of 1,000,000 dollars and a crude oil contract of 1,000 barrels, so each one-dollar move is worth 1,000 dollars. If N is 1.50 dollars, one N is worth 1,500 dollars per contract, and the unit size is , rounded down to 6 contracts. With a stop at 2N, or 3.00 dollars below the entry, the planned loss on that unit is dollars, about 1.8 percent of the account. This approach equalised risk across markets of very different size and volatility.
Results and evidence
The original Turtles traded through a period of strong trends in commodity, currency and interest rate markets, which suited breakout systems. Later tests of the published rules have shown long periods of losses and deep drawdowns, particularly in markets with few sustained trends. Trend-following strategies of this kind tend to produce many small losses and a few large gains, so their results depend on staying with the system through losing runs.
Researchers and practitioners also note that rules that worked in the 1980s may be less effective after becoming widely known and copied, and that results from backtesting depend on assumptions about costs, slippage and contract rolls.
Significance and criticism
The experiment is frequently presented as proof that trading can be taught, though critics point out that the trainees were carefully selected, that the sample was small and that not all of them succeeded. The lasting contribution of the system is less its specific entry signals than its principles: trading many markets, following trends mechanically, cutting losses quickly and sizing positions by volatility. These ideas remain central to systematic algorithmic trading and managed futures. As with any strategy, past results do not indicate future returns.