Keltner Channels vs ATR
Keltner channels draw lines a multiple of average true range either side of a moving average, so the channel is that measure made visible. Average true range is the underlying number, which is what a stop distance or a position size actually needs.
These are not two tools. One is a measure of how far price typically travels in a bar, and the other is that measure drawn around a moving average.
What each one is
Average true range measures the typical size of a bar, including any gap from the previous close. Average true range covers the calculation.
Keltner channels plot that measure around an average. Lines sit a chosen multiple of it either side of a moving average. Keltner channels covers the construction.
So one contains the other. Everything the channel does is the measure made visible, plus a centre line and a multiplier you chose.
Where they differ
Whether it is a picture or a number. The channel is readable at a glance; the raw measure is a value a rule can use directly.
What plotting adds. A centre line and a multiplier. Both are choices layered on top, and both change what a touch of the edge means.
What each is used for. The channel for judging stretch on the chart; the number for stop distances, position sizes and volatility filters.
How comparable readings are. The raw value is in price units and not comparable between instruments until you normalise it; the channel does not do that for you.
Where they agree
They are the same measure. Nothing in the channel originates outside it, so the two cannot disagree in any meaningful way.
Both read the whole bar. Wicks and the gap from the previous close are included, which is the advantage this family has over close-based statistics.
Both are backward-looking. Every value comes from bars that have already printed and neither anticipates a change in conditions.
And both need a length. The lookback decides responsiveness in each case and neither convention was chosen for your instrument.
Which one to use
Use the number when a rule needs it. Stop distances, position sizes and filters all take a value, and reading one off a chart line is a worse version of having it.
Use the channel when you want to see stretch in place. How far price sits from its average relative to normal movement is genuinely easier to read as a picture.
Use the number when comparing instruments. Divided by price it becomes transferable, which the channel cannot show.
And when both are on the chart, that is one measure twice. The channel is already reporting it, so a separate panel adds nothing.
Why the multiplier is the only real choice
Because the underlying measure is fixed. Once the length is set, the only thing left to decide is how many multiples out the lines sit, which is what makes touches common or rare.
And because that choice has measured consequences. On this site’s data trailing stops at 1, 2, 3 and 4 average ranges survived a median of 3, 10, 22 and 32 bars across 562 trials.
What the measured figures are here
Average true range over fourteen bars has a median of 0.5994, with a ninetieth percentile of 0.7954.
Bar ranges themselves are wider. Median 0.493, ninetieth percentile 1.101, largest 2.338 — the average smooths a distribution with a long tail.
Stop survival scales with the multiple. 3, 10, 22 and 32 bars at 1, 2, 3 and 4 average ranges across 562 trials, which is the trade-off written as numbers.
And a wider stop costs more when it goes. Which is why the multiple is a decision rather than a default.
What to check before using either
The lookback length. It is the only parameter the measure has and it decides how quickly the reading responds.
The multiplier, if you are plotting it. It sets how often price reaches an edge, which is the whole of what the channel reports.
Whether your instrument gaps. Both handle it, which is the family’s advantage — but the size of the effect differs by market.
And what the number is actually for. A stop, a size or a filter are different jobs and only some of them want a picture.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two
directly — this pair is constructed from two subjects the corpus covers separately. Separately,
average true range appears in 307 titles at a median of 9,432 across 220 channels, and Keltner channels
in 119 at a median of 3,360 across 96. The counts come from site/corpus_count.py.
307 videos on the measure at 9,432 and 119 on the channel at 3,360. Nearly three times the coverage and nearly three times the audience per video for the raw number — unusually, the underlying statistic is better covered than the indicator built from it.
The answer to the question on that chart is to use the number. The channel is that number times a multiplier, drawn — so reading a distance off the line is an indirect way of using a value you already have.
When it fails
The failure is treating a channel touch as a signal when it is a distance. Price reaches the upper edge, which reads as stretched, so a fade is taken. The edge is a stated multiple of typical bar range from an average — in a trend price can sit there for many bars while the envelope moves up beneath it. Nothing about the touch says a reversal is due, and the multiplier you chose decided how often it would happen.
The second failure is running the channel and the raw measure together. One measure twice.
A third is comparing raw values across instruments. They are in price units.
A fourth is leaving the multiplier at a default. It is the only real choice.
A fifth is stopping exactly at the edge. The largest bar range here was 2.338.
And a sixth is expecting either to give direction. Both measure movement only.
Related
Keltner channels covers the plotted version. Average true range covers the measure underneath. And the ATR trailing stop covers its most common practical use.
Once you see that the channel is the measure with a multiplier on it, the choice becomes obvious: if you want to look at it, plot it; if you want to calculate with it, use the number. There is no third option and nothing is lost either way.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.