TURTLE TRADING / TREND FOLLOWING / EA004
Understand Turtle Trading: breakout entries, volatility-based sizing and disciplined exits. Explore published historical evidence, then explore EA004, our free, configurable MT5 Expert Advisor.
Get EA #004 Turtle Breakout — FREE MT5 robot →
Part 1 — The original Turtle Trading method
Richard Dennis and William Eckhardt taught the Turtles a systematic futures approach in the 1980s. Curtis Faith’s 2003 rules document is the primary reference here. [1]
| Component | Historical baseline |
|---|---|
| System 1 | 20-day high/low entry; opposite 10-day channel exit. Skip after a winning previous breakout, including hypothetical skipped signals; a 55-day failsafe prevents missing a major trend. |
| System 2 | 55-day high/low entry; opposite 20-day channel exit. No previous-winner filter. |
| Timing | Intraday breakout beyond prior days’ extremes; no daily-close confirmation. |
| Volatility N | Seed with 20 true ranges; then N = (19 × previous N + today’s TR) / 20. |
| Unit sizing | One N corresponds to 1% of account equity. A 2N stop therefore implies about 2% initial risk per unit. |
| Pyramiding | Add at 0.5N intervals from actual fills, up to four units; adjust earlier stops as units are added. |
| Portfolio controls | Unit caps: 4 per market, 6 closely correlated, 10 loosely correlated, 12 in one direction. Reduce sizing capital during drawdowns. |
Condensed baseline, not every execution exception or alternative stop method. Read the complete Original Turtle Rules before attempting a historical replication.
The idea: participate in a move rather than predict its turning point
A breakout system asks whether price has escaped its recent range. It does not require a story about why that happened. The channel gives the entry a reproducible reference; the exit gives the position room to develop. In a sideways market, the same mechanism can repeatedly buy a brief rise or sell a brief fall just before the reversal.
For EA LAB, the useful research question is how entry speed, exit speed, volatility and exposure interact. A faster entry can participate earlier but may increase turnover. A faster exit can retain more of a reversal move while also cutting off a trend that later resumes. Neither trade-off can be settled by changing one parameter on one attractive chart.
A worked example: breakout, additions and failure
Illustrative arithmetic only. Suppose a configurable long setup has an entry level of 200 and a frozen volatility value of 4. An initial fill at 201, with a 2N stop, gives a stop at 193. With additions spaced 0.5N apart, the next intended levels are 203, 205 and 207. Real fills can differ; the implementation must use executable quotes and record slippage.
When the move continues
Additions increase exposure only after favorable movement. The channel exit may follow the trend upward. A winning campaign can still surrender substantial open profit before that exit is reached.
When the breakout fails
Price can cross the entry and reverse directly into the stop. Repeated failures, correlated positions and an opening gap can create a larger loss than the nominal risk calculation suggests.
For EA004’s risk-per-stop mode, equity of 10,000 and a 0.5% unit budget allocate 50 currency units before costs. If one lot would lose 400 at the actual stop, the calculated volume is 0.125 lots, rounded down to the broker’s permitted step. This is an example of our implemented sizing mode, not the original Turtle 1%-per-N convention and not a recommended risk level.
Historical evidence across futures and currencies
External simulations, not EA004 results. These publications provide test descriptions and figures; we have not independently rerun them or audited the original Turtle accounts. Their different instruments, sizing and execution assumptions prevent a direct ranking. They are historical evidence about specific implementations, not forecasts.
Futures portfolio — a strong early period and a difficult later one
Garner tested System 1 across futures sectors, 1 January 1970–23 September 2009, starting at USD 1 million. Costs, interest and reduced unit risk were included. [2]
| Published portfolio test | CAGR | Max drawdown | Longest drawdown | Trades |
|---|---|---|---|---|
| Original-style System 1 | 72% | 66% | 85 months | 11,440 |
| Revised model, 100+ futures markets | 35.28% | 15.60% | 16.3 months | 3,205 |
The revision combined 90/45-day channels, a wider stop, scaling out, a trend filter and changed portfolio controls. The original-style curve later flattened. Neither simulation proves an isolated modification or represents audited live performance. [2]
Spot forex — the same method produced different outcomes
Daniel Fernandez’s The Turtle system: Forex performance analysis (Currency Trader, January 2010) tested seven pairs in MT4, 1 June 2000–1 June 2009, starting at USD 100,000, with quoted spreads of 2–5 pips. Its entry description uses closing-price extremes, a departure from the original high/low rule. [3]
| Pair | Published average yearly return | Max drawdown | Trades |
|---|---|---|---|
| EUR/USD | 27% | 23% | 147 |
| GBP/USD | 8% | 34% | 151 |
| USD/CHF | −1% | 40% | 185 |
| USD/JPY | 3% | 33% | 141 |
| NZD/USD | 22% | 66% | 168 |
| AUD/USD | 20% | 64% | 154 |
| EUR/JPY | −1% | 67% | 157 |
These are individual-pair results. The 11% return and 47% drawdown averages are cross-pair averages, not shared-capital portfolio statistics. “Average yearly return” is not relabeled CAGR. [3]
Three original source curves
Limited figure excerpts are retained for discussion with their original axes. We have not digitized, smoothed or reconstructed a balance ledger from these images. A report figure is not enough to verify every trade, funding flow or drawdown calculation.
The forex publication calls these equity curves; their vertical axes say “Balance (USD)” and their horizontal axes count trades. We preserve that distinction and do not claim they are continuous tick-by-tick equity series. Figure rights remain with their respective authors and publishers.
Part 2 — EA #004 Turtle Breakout for MT5
Available free, version 1.01. EA004 is a configurable, time-bar breakout robot for MT5 hedging accounts. Open the free EA004 product and download package. It is an EA LAB adaptation; a single CFD chart does not reproduce the complete original futures portfolio.
System 1 and System 2
S1 uses fixed 20/10 channels with virtual previous-winner tracking and a 55-bar failsafe. S2 uses fixed 55/20 channels without the previous-winner skip. D1 is the source default; the supplied H4 configurations measure H4 bars, not days.
Your research configuration
Custom mode enables editable entry and exit lookbacks. Optional favorable-move additions, SMA filtering, spread restrictions, ATR trailing and risk settings let you compare controlled variations. Entry is a tick breakout, not a selectable candle-close confirmation.
| Implemented control | Behavior and limit |
|---|---|
| Channels and timeframe | Prior completed-bar high/low; the active bar is excluded. S1/S2 periods are fixed; Custom periods are editable. |
| Volatility N | Smoothed completed-bar true range. Campaign N sets the stop and addition spacing. |
| Position sizing | Risk at the actual stop, original-style 1N sizing or fixed lots. Broker tick values and volume steps are used. Source default stop-risk budget is 0.5%. |
| Pyramiding | Optional additions after favorable movement, up to four units, with own-symbol campaign exposure checks. No cross-symbol correlation coordination. |
| Exits | Broker hard stop plus opposite channel exit; optional ATR trailing never loosens a stop. No fixed take profit. |
| Filters and execution | Optional fast/slow SMA ordering on completed bars and spread limits; direction, broker stop-distance and margin checks. |
| Chart interface | Dark native OHLC candles, compact CLEAN panel, BUY/SELL/EXIT labels and trade links in v1.01. |
Supplied EA004 historical results
Both H4 reports cover 1 January 2020–1 October 2026 on OANDA TMS data, starting at USD 50,000 with 1:100 leverage. They identify v1.00; v1.01 supplies the new interface. These figures are separate from the external historical publications above.
| Measure | GOLD.pro | US100.pro |
|---|---|---|
| Net profit (USD) | 121,798.15 | 152,829.17 |
| Profit factor | 1.92 | 1.63 |
| Maximum relative equity drawdown | 39.58% | 39.86% |
| Trades | 109 | 140 |
Both use S2 55/20, 3% unit risk, a 4% campaign cap, up to four units and 0.5N spacing. GOLD uses N20 / 2.5N stop; US100 uses N15 / 1.5N stop. MA filtering, ATR trailing and drawdown risk reduction are off. The presets are historical examples, not recommended live risk settings.
History quality is reported as 99%; explicit tick-model selection and fixed spread assumptions are not provided. GOLD swap totals USD −33,259.45 with zero commission; US100 reports zero commission and swap. View both balance curves and download the original reports on the EA004 product page.
Broker execution and documentation
This build requires a hedging account. Broker stops can remain active while MT5 is offline, but channel exits require the running EA and a connection. The campaign cap covers its own symbol; separate chart instances are not a coordinated portfolio. Read the English manual with all 41 actual inputs, installation and license steps.
The compiled EX5 was supplied by the author. Packaging did not include independent recompilation, activation-matrix testing or a fresh v1.01 broker rerun. Validate your exact feed, costs and configuration on demo.
Testing and configuring EA004
- Verify the rules. Test channel exclusion, signal state, unit sizing, additions, stop updates and exits with small deterministic examples.
- Establish a baseline. Save the exact build and SET. Record symbol, broker feed, period, tick model, spread, commission, swap, deposit and leverage.
- Compare one change at a time. Begin with channels and exits; then additions and one optional filter. Keep sizing and costs consistent.
- Look beyond the best run. Check unseen periods, neighboring parameters, different market regimes, losing streaks and time under water.
- Test exposure jointly. A portfolio needs overlapping trades and shared capital. Adding individual returns or averaging drawdowns does not produce a valid portfolio test.
- Reconcile the output. Build EA004’s balance curve from actual deals, retain equity drawdown from the report and verify demo behavior before using a configuration.
Useful comparisons include an index, a currency pair and a metal where suitable broker data exist. These are research candidates, not a claim that the robot works on every market.
Questions before you test
Is this a Renko strategy?
No. EA004 uses time-based price bars and breakout channels. Renko Liquidity is a different method with different signal construction.
Is EA004 free?
Yes. EA004 is available free in the Robot Library, with compiled EX5 v1.01, two H4 presets, manual, Quick Start and original historical reports. Broker trading costs still apply.
Do the historical curves predict EA004’s performance?
No. They come from external implementations and historical data. EA004 requires its own reports, configurations and validation.
Does a wider stop automatically improve a system?
No. It changes sizing, holding time and risk. Evaluate it with the same costs and risk definition; a lower drawdown in another published model is not proof for your configuration.
Sources and further research
- Curtis Faith / OriginalTurtles.org — The Original Turtle Trading Rules (2003). Primary rules publication; the landing page links the free PDF.
- Anthony Garner — Tuning up the turtle, Active Trader, February 2010, pp. 34–38. Tables 3–5 and Figures 2/4; externally hosted publication PDF.
- Daniel Fernandez — The Turtle system: Forex performance analysis, Currency Trader, January 2010, pp. 20–25. Tables 1–3 and Figures 2/3; author-hosted publication PDF.
Read our MT5 backtesting checklist, guide to backtest reports and drawdown and optimization and overfitting guide. Browse Trading Strategies or the Robot Library.