Donchian channel breakout: the prior high is the signal

Research status: untested hypothesis. No backtest, paper or live results exist. Every threshold, period, R value and risk fraction is an untested candidate parameter, not a recommendation. This is not the same playbook as Breakout and retest, which waits for a volume-confirmed close and a retest of a 20-bar range.

The idea

Richard Donchian’s price channel — often called a Donchian channel — is just the highest high and lowest low of a lookback window. ChartSchool notes that the window should exclude the bar being tested, otherwise a close can never break the high of a channel that already includes that bar’s high. This playbook uses that construction as a breakout rule: leave the recent range on a completed close, then stay in until a shorter channel says the move has given back. It is a definition to test, not evidence that breakouts persist. Whipsaws are expected and must be counted.

IF–THEN rules

Channels are built from prior completed 1-hour bars only.

1. Let H20 be the highest high of the previous 20 completed bars, and L10 the lowest low of the previous 10 completed bars. Neither includes the current bar.
2. IF a completed 1-hour close is above H20, and H20 is at least 1.0 × ATR(14) above L10 (a minimum width so the stop isn’t a rounding error),
3. THEN enter at the next bar’s first executable price, if all shared checks pass.

• Initial stop: L10 at the signal bar, minus max(2 ticks, 0.05 × ATR(14)). Frozen as a floor: it may never be moved farther from the market.
• Written trailing exit: after entry, on each new completed bar, recompute the lowest low of the previous 10 completed bars. If that value, minus the same tick buffer, is above the current stop, raise the stop to that level. Never lower it. This is a pre-written tightening rule, not an improvised runner, and it is the only target this playbook uses. There is no fixed 2R take-profit.
• Time stop: none in the base rule. A 120-bar cap is a separate research variant to test later, not part of this definition.
• The stop is live intrabar. If a later bar’s high and the trail are both inside one bar, assume the stop filled first (same shared ambiguity rule).
• Size from the initial stop distance with the position-size calculator, at 0.25% of equity (untested modeling parameter). Do not add size as the trail tightens. Results are counted in R against that initial risk, including costs.
• One position at a time. No second entry while a trade is open. After an exit, wait for a fresh close above a newly computed H20.

When NOT to trade

• Channel width is below 1.0 × ATR(14), or costs eat more than 0.3R of the initial risk.
• Price is already above H20 at the open of the would-be signal bar because of a gap: skip; don’t chase an unfinished break.
• Any shared no-trade filter fails, including a daily loss pause.
• A mirrored short on a close below the 20-bar low isn’t covered here.

Common mistakes

• Building the 20-bar high including the signal bar, which makes a close-through mathematically impossible or platform-dependent.
• Waiting for a retest. That is Breakout and retest, a different hypothesis. Missed pauses are part of this rule.
• Widening the 10-bar exit after a shock because “trend following needs room.” The floor does not move down.
• Judging the rule on one spectacular hold and ignoring the cluster of small false breaks. Count every exit.
• Treating a Turtle or other published story as a result for this site. Those are other people’s tests, under other costs and markets.

Hypothetical example (made-up prices)

Item	Value
H20 (prior 20-bar high)	2,000
L10 (prior 10-bar low)	1,940
ATR(14) 20; width 60 ≥ 20	Width ✓
Signal close	2,006
Entry (next open)	2,008
Initial stop: 1,940 − 1	1,939
Risk per unit incl. 0.8 costs	69.8 (this distance is 1R)

If it works: later completed bars raise the 10-bar low, the stop is moved up in steps, and an exit at 2,148 is about +2R after costs. A large hold is geometry plus time, not a sign the next breakout is likely. If it fails: price trades back through 1,939 on the next bar and fills at 1,936: a little more than −1R. An educational chart of a break that later failed — not a result for this rule — is in technical examples.

Ratigan Capital rules in this playbook

• Early Is Wrong: buying as the high is first touched, before the close, is a different strategy.
• Never Average Down: a failed break is an exit, not a cheaper entry.
• Patience as a Position: most hours do not close through H20.
• Price Is the Only Truth: the channel is a description of recent extremes; the stop is the stop.

Why it can fail

Close-throughs of a 20-bar high fail often, especially in two-sided ranges, and a trailing 10-bar exit can hand back a large open gain or get shaken out of a lasting trend. Costs and slippage on the stop matter more here than in tight, 2R-target playbooks because some holds last many bars. Parameter choice (20/10 vs other windows) is an easy place to data-snoop (Sullivan, Timmermann & White, 1999).

Pre-trade checklist

1. H20 and L10 computed from prior bars only, and width ≥ 1.0 × ATR(14)?
2. Completed close above H20, not a wick and not a gap chase?
3. Shared filters clear; no open position?
4. Initial stop, trailing rule, and size written; 1R defined from the initial stop after costs?

Sources

• StockCharts ChartSchool: Price Channels (Donchian)
• StockCharts ChartSchool: Average True Range (ATR)
• Sullivan, Timmermann & White (1999), Journal of Finance (RePEc listing)
• Core rules for every playbook