Most indicators answer a question you did not need to ask. Moving averages tell you where price was; oscillators tell you momentum is stretched after the fact. The Donchian channel is different in one specific way: it is a direct statement of range, drawn from actual highs and lows rather than a formula applied to closes. That makes it the cleanest available definition of “price has gone somewhere it has not been before” — which is the entire premise of trend following.

This guide covers what the channel is, how the Turtle systems built on it were actually specified, what two decades of tests say about the edge, and how to adapt the rules to CFDs on UZFX without letting the platform’s simplicity hide the system’s real failure modes.

What the Donchian Channel Actually Is

Three lines, one rule each:

  • Upper band — the highest high of the last N periods.
  • Lower band — the lowest low of the last N periods.
  • Middle line — the midpoint between the two, which is a trend-strength read rather than a signal.

The default N is 20, inherited from Richard Donchian’s work in the 1950s and 60s and later fixed in place by the Turtle experiment. A 20-period channel on a daily chart means: the highest high of the last 20 trading days, and the lowest low of the same window.

Two properties matter for how you trade it.

First, the bands only move outward or reset. Price cannot be above the upper band of the channel that includes the current bar — by definition the current high is inside the range. So in practice you compare the close (or the high) against the previous N-period extreme, which is a properly forward-looking test.

Second, there is no smoothing. Every other common channel — Bollinger, Keltner — applies a moving average and a deviation or ATR multiple, which means the band drifts. The Donchian band is a step function: it holds flat until an old extreme drops out of the lookback window, then jumps. That is why breakout signals appear on the exact bar the level breaks rather than two or three bars later.

The cost of that responsiveness is noise. Every genuine trend begins with a Donchian break; most Donchian breaks are not the beginning of a trend.

The Turtle Specification, Precisely

The Turtle rules are frequently retold loosely. The actual specification is worth stating exactly, because the details are where the edge lives.

System 1 — the faster one:

  • Entry: a close above the 20-day high (long) or below the 20-day low (short).
  • Filter: skip the signal if the last breakout in that same market would have been a winner.
  • Failsafe: if a skipped breakout keeps running and price reaches the 55-day extreme, take that signal anyway.
  • Initial stop: 2N from entry, where N is the 20-day average true range. Longs: stop 2N below entry. Shorts: stop 2N above.
  • Exit: a close beyond the 10-day extreme in the opposite direction. No profit target.
  • Sizing: one unit sized so a 1N move changes equity by roughly 1 percent.
  • Pyramiding: add one unit every 0.5N in your favour, up to four units, raising the stop on the whole position each time.

System 2 — the slower one:

  • Entry: a 55-day breakout.
  • Filter: none. Every signal is taken.
  • Exit: a 20-day extreme in the opposite direction.
  • Stop and sizing: identical construction.

The filter in System 1 is the part most retail retellings drop, and it is the part that does the most work. Its logic is mechanical: if the last breakout in this market was profitable, the market is likely already trending, and a fresh breakout signal is more likely to be a late entry into an exhausted move.

What the Data Says

A 20/10 test across three liquid instruments from January 2006 to October 2026 — 499 completed trades, next-open fills, 0.05 percent cost per side — produced this:

Metric Result
Winning trades 36.1%
Average winner +4.29%
Average loser −2.46%
Winner/loser ratio 1.74
Median trade −1.24%
Median holding period 17 sessions

Read that honestly. The median trade loses money. More than six out of ten trades are losers. The system is profitable only because the average winner is 1.74 times the average loser, and because a minority of trades capture the large sustained trends that pay for everything else.

Longs won 44.4 percent of the time; shorts won 27.1 percent. That asymmetry is not a quirk of the test window — equity indices and gold spent most of the period in structural uptrends, so short breakouts fought the dominant drift. On a market with a persistent upward bias, the short side of a symmetric breakout system is close to a donation.

The 55/20 variant over the same window produced 236 trades, a 41.9 percent win rate, and a slightly positive mean trade. Fewer signals, longer holds, better hit rate — but also later entries and giving back more of each move before the exit triggers.

Adapting the Rules to CFD Trading

The Turtle rules assume futures with defined contract sizes and near-zero financing. CFDs change three things, and each one changes the maths.

1. Financing costs punish long holds. A 20-day hold on a leveraged CFD position accrues swap daily. On a 1-lot EUR/USD long at typical retail rates, that can be USD 30–80 over the holding period — comparable to the entire per-trade edge. The overnight swap and financing guide walks through how to estimate it before entry, and the practical adjustment is to prefer the faster 20/10 variant, to favour instruments with positive carry in your direction, or to size down so swap is a smaller fraction of the position.

2. Spread is a fixed tax on a losing-majority system. With a 36 percent win rate, roughly two out of three trades pay the round turn for nothing. A 1.5-pip spread on EUR/USD costs about USD 15 per standard lot round turn; a 0.2-pip effective spread costs about USD 2. Over 100 trades that difference is USD 1,300 — larger than most annual returns on a small account. This is the single strongest argument for checking the full cost structure before tuning the indicator window.

3. Gap risk breaks the ATR stop. CFDs on indices and single stocks gap over weekends and earnings. A 2N stop offers no protection against a Monday open 3N below Friday’s close. The stop becomes a market order at the first available price. Our weekend gap risk note covers the position-sizing response: hold smaller size into the weekend, or avoid the setup entirely on gap-prone instruments.

Position Sizing: The Part That Decides the Outcome

The entry rules get the attention; sizing decides whether the system survives. The Turtle unit is defined so that 1N of adverse movement equals about 1 percent of account equity:

Unit size = (Account equity × 0.01) / (N × value per point)

Worked example on XAU/USD: equity USD 5,000, N (20-day ATR) = USD 28, contract value USD 1 per 0.01 move so USD 100 per USD 1.00 move.

Unit = (5,000 × 0.01) / (28 × 100) = 50 / 2,800 = 0.0178 lots

Round to 0.01 lots — the minimum on a standard account. Note what that means: at USD 5,000 equity, a properly sized gold breakout unit is below the minimum tradeable size. Traders who ignore this and trade 0.10 lots are running roughly 5.6N of risk per stop, which is 5.6 percent of equity on a single trade, in a system that loses 64 percent of the time. Ten consecutive losses at that size is a 44 percent drawdown.

This is the mechanism behind every “the Turtles’ rules did not work for me” post. The rules were not the problem. The position sizing was.

Filtering False Breakouts

Three filters improve the raw signal without changing its character.

Volatility contraction first. A Donchian break out of a tight range (narrow Bollinger width, low ATR relative to its own 50-day average) has a materially better follow-through rate than a break from an already-extended range. Fade the second kind.

Higher-timeframe agreement. A 20-day long signal on a market trading below its 200-day average is fighting the dominant flow. Requiring price above the 200-day for longs and below it for shorts cuts trade count by roughly a third and improves the win rate.

Session and liquidity timing. Breakouts that occur during the London or New York overlap have real volume behind them; the same break at 02:00 during thin Asian liquidity often reverses by the open of London. Our session timing guide covers which windows to trust.

What does not work as a filter: adding a momentum oscillator confirmation. By the time RSI or MACD confirms a Donchian break, the move is often half complete, and the exit rule then gives back the difference. Filters that reduce signal count without improving the winner/loser ratio only reduce sample size.

Common Implementation Errors

Using the current bar in the lookback. If the channel includes the bar being tested, price can never break it. Offset the channel by one bar, or compare the close to the prior N-period extreme explicitly.

Recalculating the stop. The initial 2N stop is set once at entry and never widened. Traders who “give it room” after a drawdown have converted a defined-risk system into an undefined-risk one.

Taking the exit as a stop order. The Turtle exit is a close beyond the 10-day extreme, not an intrabar touch. Wicks through the level during a liquidity sweep are not exit signals — but if you use a resting stop there, you will be taken out by them anyway.

Trading every market with identical parameters. A 20-day channel is roughly four trading weeks. On a market with strong weekly seasonality, that window sits exactly on the cycle and produces repeated false signals. Test the window, do not assume it.

Running It on a Proprietary Platform

The Donchian channel is a standard indicator on MT4, MT5 and TradingView, and it is straightforward to reproduce on a proprietary terminal because it needs no calculation beyond highest-high and lowest-low over a window. On UZFX, whose Web Terminal and mobile apps run without MetaTrader, the practical workflow is:

  1. Add a 20-period channel for entry and a 10-period channel for exit.
  2. Add ATR(20) and read it before sizing every trade.
  3. Calculate unit size by hand from the formula above, not from a fixed lot default.
  4. Set the 2N stop at entry and leave it.
  5. Review the trade only on a close beyond the exit channel.

The limitation is real: no expert advisor, so the system cannot be automated natively, and no backtesting engine inside the platform. If automated execution matters more than platform simplicity, the no-MetaTrader platform comparison sets out the alternatives honestly. For discretionary traders who want to run the rules manually with fewer moving parts, the UZFX $10 minimum deposit account is a low-cost way to test the system with real money at a size where mistakes are survivable.

Which Market Suits It

Donchian breakouts need markets that trend and do not mean-revert aggressively.

  • Work well: equity indices (US500, NAS100, GER40), gold, crude oil, and major FX pairs during policy-divergence regimes. These produce sustained directional moves and have deep liquidity at the breakout level.
  • Work poorly: range-bound crosses, low-volatility FX pairs, and anything with heavy central-bank intervention at obvious round numbers — the break gets absorbed repeatedly.
  • Avoid entirely: illiquid single-stock CFDs and crypto CFDs outside the top few by volume, where the spread at the breakout is wide enough to erase the advantage.

FAQ

What is a Donchian channel and what does it measure?

Three lines drawn from the highest high and lowest low of the last N periods, plus a midpoint. Default N is 20. It measures range expansion directly, with no smoothing, so a new high or low registers on the exact bar it happens.

What are the classic Donchian channel rules?

Long on a close above the 20-period high, short on a close below the 20-period low. Stop 2 × ATR(20) from entry. Exit on a close beyond the opposite 10-period extreme. No profit target. System 2 substitutes 55 for entry and 20 for exit and takes every signal.

Does the Donchian channel strategy actually work?

As a positive-expectancy trend-following system, yes — but only with the stop and exit rules intact. A 2006–2026 test across three liquid instruments showed 36.1 percent winners with a 1.74 winner/loser ratio. The median trade lost 1.24 percent. Remove the 2N stop or the exit channel and the edge disappears.

Should I use a 20-period or 55-period Donchian channel?

20/10 for swing trading on daily bars — more signals, shorter holds, lower financing drag. 55/20 for position trading and for accounts where swap costs dominate, at the cost of later entries and fewer trades. Test the window on the specific market rather than assuming the default is optimal.

How do trading costs affect a Donchian breakout system?

Severely, because two out of three trades lose and each still pays the round turn. A 1.5-pip spread versus a 0.2-pip effective spread on EUR/USD is roughly USD 1,300 per 100 standard-lot round turns. On a breakout system that difference can exceed the gross edge, which is why cost-per-trade matters more here than in a high-win-rate strategy.

Verdict

The Donchian channel is not a signal generator. It is a definition — the cleanest one available of “price has left its recent range”. The Turtle rules turned that definition into a complete system by adding the three things a definition cannot supply: a volatility-scaled stop, a volatility-scaled position size, and a mechanical exit that does not ask for a target.

Everything that makes the system work is in the sizing and the exit. Everything traders change first is the sizing and the exit. Run the 20/10 default on a trending market, size so 1N equals 1 percent of equity, keep the stop where it was set, and accept that six of ten trades will lose — that is the system, not a malfunction.

Risk Disclaimer

Trading leveraged CFDs and forex carries a high level of risk and may not be suitable for all investors. You can lose more than your initial deposit. Backtest results quoted in this article are historical, use next-open fills and a stated cost assumption, and do not account for swap, slippage during news events or platform execution differences. Past performance is not indicative of future results. Verify every strategy on a demo account before risking capital, and confirm that any broker you use holds a verifiable licence — for UZFX that is ASIC AFSL 001291473, checkable through the ASIC professional registers search. This article is for informational purposes only and does not constitute financial advice.

Last reviewed: 15 November 2026 by the MarketCFD editorial team. Related reading: ATR volatility and position sizing, Bollinger Bands complete guide, and breakout trading strategy for CFDs.