Donchian Channels vs ATR

Michael Whitman · Trading since 2017, full-time since 2020 · Published 2026-09-08 · Updated 2026-10-04 · Reviewed 2026-10-04

These two are often lined up as rival volatility tools. They answer different questions. The channel gives a price on the chart; the range measure gives a distance in dollars. A breakout trade needs both, and SPY’s own history shows how the two fit together.

What each one is

Donchian channels mark the top and bottom of the recent range, here the previous 20 sessions, the usual setting for Richard Donchian’s price channels. The edges are prices that traded. Donchian channels covers it.

Average true range measures a typical session’s travel, including any jump from the prior close. The version here is Welles Wilder’s 14-session average. Average true range covers it.

Neither substitutes for the other. The channel has no opinion about distance, and the range measure has no opinion about where anything is.

Where they differ

What each produces. A price against a size. The channel is a coordinate on the chart; the range measure is a number used in arithmetic.

How each moves. The upper channel line changes only when a new high enters or an old one drops out of the window. Average true range changes a little every session as ranges widen and narrow.

How they relate. On SPY the 20-session channel was a median 4.66 average true ranges wide, from 3.38 to 6.61 for the middle 80% of sessions. It was narrowest at 2.07 ranges on 6 July 2004 and widest at 11.02 on 27 November 2023, after a fast November rally with the late-October low still inside the window.

What each says about a breakout. The channel says one happened. The range measure says how far price usually travels in a session afterwards, which is the distance a stop has to live with.

Which one to use: Donchian Channels or ATR?

Use the channel when you need a level that others can see. A 20-session high is the same number on every platform, whereas a range-based level depends on the multiple each trader chooses.

Use average true range when the question is how far away a stop goes or how many shares to hold. Both need a number in dollars, and the channel does not supply one.

Use both when trading a breakout. Mark the level with the channel, then set the stop some number of ranges beneath it rather than a few cents under it.

And when the channel is narrow in range terms, expect the level to be retested. A channel only a few ranges wide means the breakout close is close to ordinary noise.

How much room a breakout needed on SPY

The test. Here a breakout means a session closing higher than every high of the 20 sessions before it. From 3 January 1994 to 2 October 2026 there were 962 with 10 sessions still to come. To keep them separate, a fresh breakout is one with no such close in the 10 sessions before: 193 of them. For each, the next 10 sessions’ deepest low was set against the broken level, in that day’s average true range.

The breakout close was usually close to the level. Fresh breakouts closed a median of under three tenths of a range above it, and 87 of the 193 (45.08%) closed less than a quarter of a range above.

A return to the level was the normal case. The next 10 sessions traded below the level after 148 of the 193 (76.68%); half a range below after 104 (53.89%); a full range below after 71 (36.79%); two ranges below after 37 (19.17%). Counting all 962 breakouts, overlapping runs included, the shares were 82.74%, 62.16%, 44.28% and 23.60%.

One breakout after the April 2025 slide

A worked case. On 2 May 2025 SPY closed at $566.76, above the prior 20-session high of $564.07, with an average true range of $14.45 after April’s turmoil. Within the next 10 sessions the low reached $556.04, 0.56 ranges below the level. A stop half a range under the level ($556.85) was reached; one a full range under ($549.62) was not.

The original data

The sample. Daily SPY bars from Yahoo Finance, high, low and close as traded, beginning 29 January 1993 and counted from the first session of 1994 through 2 October 2026. Opens are not used. Channel: the top of the prior 20 sessions’ highs and the bottom of their lows. Average true range: Wilder’s 14-session smoothing; the width uses the prior session’s value, the room uses the breakout day’s.

The headline figure. After 76.68% of the 193 fresh breakouts the next 10 sessions traded back under the level, and after 36.79% a full range under it.

The file. Every session’s channel, range, width and, on breakout days, the next 10 sessions’ lowest low and its distance from the level are in the SPY channel and range file.

What it does not show. Whether breakouts went on to rise, any cost, or any stop rule’s result. A low below a level says a stop there was reached; it does not say the trade lost, because nothing here follows what came after. The Bollinger Bands vs Donchian channels page counts how often each kind of break happened.

When it fails

The failure is a stop a few cents under the breakout level. The close clears the 20-session high, the stop goes just beneath it, and two sessions later an ordinary down day trades through it. On SPY that return to the level followed about three fresh breakouts in four, so the stop was set where the normal case would reach it. The level may well hold on a closing basis; the position is gone either way. Measuring the stop in ranges from the level is the fix.

The second failure is sizing from the channel’s width. The width is several ranges; the stop is not.

A third is a range measure with no level. It says how far, never where.

A fourth is trusting a narrow channel. In range terms it is close to noise.

And a fifth is counting overlapping breakouts as separate cases. They are one run of strength.

Related

Donchian channels covers the level. Average true range covers the distance. And breakout covers the event the two are used to trade.

This page is educational, not financial advice. Test every idea on your own charts before risking money.