The experiment behind the name
Could successful trading be taught? In 1983, commodities trader Richard Dennis and his partner William Eckhardt turned that question into an experiment. Dennis favored teachable methods; Eckhardt put more weight on innate ability. They recruited trainees through newspaper advertisements, taught them a systematic approach, and supplied trading capital. Participant Curtis Faith describes the first group's training in Chicago in December 1983, followed by trading in early 1984. He also recounts Dennis comparing the project with raising turtles in Singapore—the source of the nickname. These details come from the introduction to the Original Turtle Trading Rules.
The legacy includes careers beyond the experiment. Original participant Jerry Parker founded Chesapeake Capital in February 1988. His firm's history of the program and its own founding traces its approach to diversification, cutting losses, allowing profits to run, and consistent execution.
That history makes the strategy interesting to study. It does not establish what a particular stock backtest should earn today. A historical training program, a diversified futures portfolio, and a daily stock simulation are different research settings.
What a breakout is trying to capture
A breakout happens when price moves beyond a previous range. Suppose a stock's highest traded price over its last 20 completed sessions was $100. A close of $101 is above that boundary. The question is whether the move will continue far enough to compensate for entries that reverse.
This makes a breakout strategy comfortable buying after a rise. Its testable premise is that some moves persist. It accepts that a new high can also be followed immediately by a decline. An exit rule decides when to abandon that attempt.
A Donchian Channel makes the range visible by plotting rolling highs and lows. TickRun's Turtle implementation uses separate entry and exit windows, plus a filter with memory. That memory distinguishes it from simply buying every new 20-session high.
Where System 1 fits
The original rules distinguish System 1's 20-day entries and 10-day exits from System 2's 55-day entries and 20-day exits. System 1 skips entries after a winning breakout but retains a 55-day fallback. The complete method also includes long and short futures positions, volatility sizing, stops, and adding units. Entries were triggered during the session rather than awaiting its close. See the entries, stops, and exits chapters of the original participant account.
A software label therefore needs a precise specification. Which windows does it use? Does it track skipped signals? What price creates an entry? How is position size determined? Two backtests can both say “Turtle” and answer those questions differently.
The fixed rules in TickRun
Select Experimental → Turtle Breakout (System 1). TickRun implements a daily-close, long-only adaptation with these fixed settings:
| Component | TickRun behavior |
|---|---|
| Normal entry | Close above the prior 20-session high, if the winner filter permits it. |
| Fallback entry | Close above the prior 55-session high, regardless of the winner filter. |
| Channel exit | Close below the prior 10-session low. |
| Volatility stop | Close at or below entry close minus twice the entry value of N. |
| Actual holdings | Long or cash; the filter also tracks hypothetical shorts. |
“Prior” excludes the current session. Including today's high in today's entry boundary would undermine a close-above-high test: the close cannot exceed its own session's high. A boundary built from earlier bars is also straightforward to audit.
TickRun allows the first eligible breakout when no completed hypothetical outcome exists. Normal entries need 20 complete prior sessions; fallback entries need 55. Those are trading-session counts, so weekends and market holidays do not advance them.
The filter follows trades you did not take
TickRun maintains a separate hypothetical 20-session breakout trade. It can point upward or downward and is tracked independently of the actual long position. It closes on its opposite 10-session boundary or after an adverse move of 2N, using daily closes.
A profitable hypothetical channel exit marks the last completed breakout as a winner. That blocks ordinary 20-session long entries. A stopped trade, a losing channel exit, or a breakeven exit removes the block. While a hypothetical trade remains open, the filter retains the last completed result.
The crucial detail is that a skipped breakout still gets tracked. Otherwise the filter could remain stuck on an old winner. Likewise, a hypothetical short winner can block a later long entry even though TickRun never held a short position. Using only actual completed trades would produce a different strategy.
A skipped trade and a fallback entry
Consider an illustrative sequence with invented prices. Assume the last completed hypothetical breakout was profitable, and the strategy currently holds cash.
- The prior 20-session high is $100 and the prior 55-session high is $110. A $101 close creates a normal breakout, but the filter skips the actual entry.
- TickRun still opens a hypothetical long at $101. If N is $2 at that entry, its fixed 2N stop is $97.
- If no channel exit occurs first and a later close reaches $97, the hypothetical loss clears the filter. A subsequent eligible 20-session breakout can enter.
- Alternatively, suppose the skipped move continues upward. A close above the then-current 55-session high can trigger the actual fallback entry while the filter remains blocked.
These are alternative paths, not a claim about observed performance. Recalculate channel boundaries each session: the $110 fallback level in the example does not remain fixed forever. The fallback is still part of System 1; it does not switch the exit rule to System 2's longer window.
What N measures—and what it does not
TickRun calculates N from true range, a measure that captures both a session's high-to-low movement and gaps relative to the previous close:
True range = max(high − low, |high − previous close|, |low − previous close|)
For the first available bar, the range is high minus low. TickRun seeds N with the mean of the first 20 true ranges, then updates it with:
N = (19 × previous N + current true range) / 20
The entry value is frozen for that trade's stop. If an actual entry occurs at $101 with N of $2, its stop remains $97 even if later volatility changes. The separate 10-session low can rise as older bars leave its window, providing another exit condition.
A 2N distance is a price distance, not an account-risk percentage. TickRun's all-in/all-out position model does not calculate Turtle units or reduce the position when N rises. An N-based stop alone does not recreate volatility position sizing.
Daily closes change the experiment
A session might trade above the entry boundary at noon and close below it. TickRun creates no entry from that intraday excursion. Similarly, touching the stop during the session and closing above it creates no close-based stop exit.
When a close does confirm a signal, the changed position affects the following close-to-close return. A stop checked at the close is not an order guaranteed to fill at its threshold. A close below the threshold can mean a larger loss than the nominal stop distance.
TickRun also omits pyramiding—adding to an existing winning position—and actual short holdings. Its results describe this specific adaptation. For the execution model shared by other strategies, see signal timing and look-ahead bias and the Turtle section of the user manual.
How to assess a backtest
Start with the fixed strategy rather than searching for replacement windows. TickRun exposes no adjustable Turtle parameters; Find best parameters evaluates one configuration. Selecting a favorable stock or date range can still introduce selection bias even when the rules stay fixed.
Inspect the trade list alongside the chart. Check whether entries followed valid prior highs, whether a filtered entry remained in cash, and whether exits followed a channel break or close-based stop. A quiet chart may reflect filtering rather than missing data.
Then compare return with drawdown, completed trades, transaction costs, and time in the market. Repeated failed breakouts can make sideways periods expensive. A strong result dominated by one long trade deserves inspection across additional periods and stocks chosen for a stated reason.
Reserve later data for evaluation after making research choices. If Turtle wins a comparison against 51 other strategies, that comparison is itself a selection step. The guide to out-of-sample testing explains how to keep a later evaluation useful.
The historical experiment supplies context; the implementation supplies a reproducible question. Use both to understand what was tested before deciding how much weight to place on its result. You can test Turtle System 1 in TickRun or review its complete signal specification.