Turtle Trading Strategy Rules, and What a CFD Account Adds

The turtle rules are unusual among famous trading methods in that the full rule set was published, free and complete, by one of the traders who was taught it. Nothing has to be inferred from anecdotes.

What has not been published is the translation. Every number in the document is written for a futures contract with an exchange-set value per point, and almost nobody reading it today is trading one.

What the Turtle Rules Actually Were

The document describes a complete trading system in the strict sense: it fixes what to trade, how much, when to enter, when to cut, when to leave and how to place the orders. No step is left to the reader.

Two systems run side by side. One is short term and built on a 20-day breakout, the other longer term and built on a 55-day breakout, and the traders were free to allocate whatever share of equity they wished to either.

A breakout is defined as price exceeding the high or the low of a stated number of preceding days, taken intraday rather than waiting for a close. That is the whole entry condition, and it makes the method systematic trading in the reproducible sense: two people reading the same chart reach the same decision.

What the rules do not contain is any forecast. Nothing in them estimates where a market is going, which is why the risk instructions occupy far more of the document than the entries do.

N: The Volatility Unit Everything Else Is Built On

N is the average true range over 20 periods, where true range is the largest of the current high minus the current low, the current high minus the previous close, and the previous close minus the current low.

The smoothing is specific. Each day, N is 19 parts of the previous day’s N plus the current day’s true range, divided by 20, seeded from a plain 20-day average of true range for the first value.

That detail matters more than it looks. An average true range indicator on a platform takes its averaging period as a parameter, as the MQL5 reference for iATR shows, so a chart left on a default setting is not producing N. It is producing a different number that happens to share the name, and the Donchian channel page makes the same point about lookback windows drawn two different ways.

From N comes everything else. Dollar volatility is N multiplied by the value of one point of the contract, and a Unit is one per cent of account equity divided by that dollar volatility. Sized this way, one N of adverse movement costs one per cent of the account, whatever the instrument.

Translating a Futures Unit to a CFD or Spot Account

The Unit formula needs one input the document treats as fixed: the value of one point. On a futures contract that value is set by the exchange and is identical for everyone trading it.

On a spot forex or CFD account it is not. Point value follows the broker’s contract size for the instrument and the currency the account is denominated in, which is why lot size has to be resolved before any of this arithmetic runs.

The consequence is that the same rule set, applied to the same pair with the same account balance, produces different Units at two brokers whose contract specifications differ. The rule has not changed. Its input has.

None of this breaks the method. It means the arithmetic is a step in the setup rather than a constant to be copied, and a Unit figure taken from any example in the document describes that example alone.

The Two Entry Systems, and the Filter That Connects Them

System 1 buys one Unit when price exceeds the previous 20-day high by a single tick, and sells one Unit when it drops a tick below the 20-day low.

Then comes the filter that most summaries state only in half. A System 1 signal is ignored when the last breakout in that market would have been a winning trade. The test is one of order: did the 2N adverse move arrive first, ahead of any profitable exit at a 10-day extreme. Whether that earlier breakout was long or short makes no difference to it.

Two details inside that rule are routinely dropped. The last breakout means the last breakout in the market whether or not it was actually taken, so a skipped signal still counts as the reference. And the test is hypothetical: the question is what the trade would have done, not what the account did.

The failsafe is the other half. When a System 1 entry is skipped because the previous breakout would have won, the 55-day breakout is taken instead, so the method cannot sit out a major move entirely. Presented without that clause, the filter reads as a rule that removes trades, when its actual job is to move one class of trade onto a slower entry.

The exits are equally mechanical. System 1 leaves on a 10-day low for a long position and a 10-day high for a short one; System 2 uses 20 days in the same way, and the whole position leaves together.

Adding Units, and the Caps That Limit Them

A position is built rather than taken. After the first Unit, further Units are added at half-N intervals, and the interval is measured from where the previous order actually filled rather than from the theoretical breakout level, so slippage carries forward into every subsequent add.

Risk is held down by the stop rule. No trade may risk more than 2 per cent, and since one N equals one per cent of equity, the stop sits 2N from entry. As Units are added, the stops on the earlier Units are raised by half an N each time, which keeps the whole position inside the same risk budget.

The document also records an alternative the traders were told about, in which stops sit at half an N for half a per cent of risk and a stopped-out Unit is re-entered if price returns to its original entry. It produces more losing trades, and it removes the need to move earlier stops at all, since four Units can never exceed the 2 per cent ceiling.

That is where the connection to maximum drawdown sits: the ceiling constrains a single position, not a sequence of them.

Why the Correlation Caps Bite Hardest on a Forex Account

Above the per-trade stop sit four ceilings on how many Units may be open at once: 4 in a single market, 6 in one direction across closely correlated markets, 10 in one direction across loosely correlated markets, and 12 in one direction in total.

The examples given for closely correlated markets are pairs such as heating oil and crude oil, or gold and silver, and among currencies the Swiss franc against the Deutschmark. The grouping is about shared drivers, not about the instrument sharing a name.

Apply that definition honestly to a retail forex account and the arithmetic changes. Four dollar-quoted pairs are not four markets under this rule; they are one closely correlated group, so the binding limit is 6 Units in a direction rather than 4 Units in each of four pairs. Our page on currency correlation covers how tight those relationships get when conditions turn stressed.

This is the single most common way the rules are followed and breached at the same time. The trader counts Units per pair, sees 4 in each and believes the cap is respected, while the rule as written was counting the group.

The Cost the Original Turtles Never Paid

The turtles traded futures. A futures price already contains the cost of carrying the position to expiry, so holding one for six weeks incurs no separate nightly charge.

A trend-following position on a CFD or spot account is charged or credited every night it stays open, and the sign depends on the instrument and the direction. Deriv’s published trading specification, checked on 12 August 2026, lists AUD/USD at 0.09 points for a long position and minus 2.42 points for a short one, which is not symmetrical and not negligible over a position measured in weeks.

The rules contain no provision for it, because there was nothing to provide for. That leaves a decision the reader has to make rather than inherit: whether the holding cost is deducted from the expected result of the method, or treated as a reason to run the rules on an instrument that does not charge it.

The practical check is short. Multiply the nightly figure by a realistic holding period for a 20-day or 55-day breakout system, compare it with the 1 per cent of equity that one N is meant to represent, and the size of the omission is visible immediately.

What a Backtest of These Rules Can and Cannot Tell You

The rules are fully specified, so they can be tested, and backtesting a rule set is the right instrument for the question.

Two things will not come out of the test. The filter depends on breakouts that were never taken, so the simulation has to track hypothetical trades alongside real ones or the entry logic is wrong from the first bar. And the historical results belong to a market set of futures contracts, not to the pairs on a retail platform, so a result carried across instruments is a different experiment wearing the same name.

Who This Is Not For

Anyone looking for a method to run unmodified will not find one here, because the point-value input has to be derived for each instrument and broker before the first Unit is sized.

Anyone unwilling to hold through the exits will find the exits are the method. Waiting for a 10-day or 20-day extreme means returning a visible part of an open gain, by design, and the document says plainly that this is the hardest part of the system to follow.

Frequently Asked Questions

What are the two entry systems in the turtle trading rules?

System 1 enters on a breakout of the previous 20 days and System 2 enters on a breakout of the previous 55 days. Traders were free to split equity between them in any proportion, including running only one of the two.

What does N mean in the turtle system?

N is the average true range measured over 20 periods, updated each day as 19 parts of the previous value plus the current true range, divided by 20. It is expressed in the points of the contract, and a position sized at one Unit loses about 1 per cent of account equity for every N the market moves against it.

How many units could a turtle hold in correlated markets?

Four Units in any single market, six in one direction across closely correlated markets, ten in one direction across loosely correlated markets, and twelve in one direction across everything held at that moment.

Why did the turtles skip some breakout signals?

A System 1 signal was ignored when the previous breakout in that market would have been profitable. That is judged by order of events, namely whether the 2N adverse move landed ahead of a profitable exit at a 10-day extreme. When a signal was skipped this way, the 55-day breakout was taken instead so a large move could not be missed entirely.

What changes when the turtle rules are applied to a CFD account?

Two inputs change. The value of one point becomes a broker setting rather than an exchange specification, so the Unit has to be recalculated per instrument, and a position held overnight is charged or credited a daily swap that a futures position never paid.