Keltner Channels vs Standard Deviation
Keltner channels draw an envelope a multiple of average true range away from a moving average, so the width reflects whole-bar movement including gaps. Standard deviation instead measures how dispersed closing prices have been, which leaves out wicks and gaps entirely.
An envelope drawn from bar range against a statistic computed from closes. They both describe movement and they are looking at different parts of the same bars.
What each one is
Keltner channels place lines a multiple of average true range from a moving average. The width reflects how large bars have been, gaps included. Keltner channels covers it.
Standard deviation measures how dispersed closing prices have been around their own average, in price units. Standard deviation covers it.
One reads whole bars and the other reads closes. That is the difference, and everything else follows from it.
Where they differ
What each can see. A day that swung widely and closed flat is a large reading on the range side and almost nothing on the dispersion side.
How gaps are treated. Average true range counts the jump from the previous close explicitly. A close-based statistic treats it as an ordinary change in level.
How outliers behave. Dispersion is defined by extremes and moves sharply on one violent close. A range average absorbs a single event more gently.
Whether it is drawn. One is an envelope on the chart by default; the other is a number that usually lives inside a rule rather than on the screen.
Where they agree
Both measure movement rather than direction. Neither has an opinion about which way price is going, which is why both are inputs rather than signals.
Both are backward-looking. Every value comes from bars that have already printed, and neither anticipates a change in conditions.
Both need a length. The lookback decides responsiveness, and a value chosen for one instrument does not transfer to another.
And both are in price units. Neither is comparable between markets until divided by price, which is a step people frequently skip.
Which one to use: Keltner Channels or Standard Deviation?
Use the range-based envelope for anything involving a stop. A stop is touched intrabar, so the measure that counts the whole bar is the one measuring the right thing.
Use dispersion when the question is statistical. Comparing how unusual a close is, or building a model on closing prices, are jobs the measure was designed for.
Use the range version on anything that gaps. Instruments that open away from the previous close leave a close-based measure blind to a real part of their movement.
And when you want it drawn, use the channel. The envelope is the range measure made visible, which is easier to read in place than a number in a panel.
Why the input decides the answer
Because a stop does not wait for the close. Price reaching your level ends the trade regardless of where the bar finishes, so a close-based measure is answering a different question.
And because wide bars that close flat are common. A bar can swing across a large range and finish almost where it opened, and a close-based measure records nearly nothing of that trip.
What to read off your own chart
Average true range over fourteen bars has a normal level and a busy-market level on any instrument, and both are worth reading off your own chart before a multiplier is chosen.
Bar ranges themselves run wider than their average. Most bars are modest, a minority are much larger, and the occasional one is several times the norm — a distribution with a long tail that an average smooths.
Trailing stops sized from range survive predictably. Wider multiples of average true range last longer before ordinary movement reaches them, and the pattern is easy to confirm by counting on your own instrument.
And that spread is the actual decision. A wider stop lasts longer and costs more when it goes, which is a trade-off you can size rather than guess.
What to check before using either
The lookback length. It is the main parameter on both and it decides how quickly each responds.
The channel multiplier, if you are plotting it. It sets how often price reaches an edge.
Whether your instrument gaps. If it does, a close-based measure is systematically understating movement.
And whether you need comparability. Both are in price units, so dividing by price is what makes a reading transferable between markets.
The original data
Of the 24,971 unique videos in this site’s search study, 1 names standard deviation in its title, while Keltner channels appear in 126 titles at a median of 3,163 across 101 channels. This pair is built from subjects the corpus covers very unevenly.
126 videos on the channel and 1 on the statistic. The tools built on that measure are taught constantly, and the measure itself is what decides how they behave.
Which reading matters depends on the job. For a stop, the wide bars are the fact; for a statistical filter on closes, the steady closes are — and using the wrong one is how a stop ends up too tight.
When it fails
The failure is sizing a stop from a close-based measure while the bars swing. Dispersion is low because closes have clustered, so the stop is placed close to price. Bars keep reaching well beyond it intrabar — wide ranges that finish near where they started barely register at all in a dispersion of closes — and the position is stopped again and again by movement the measure never counted.
The second failure is comparing raw values between instruments. Both are in price units.
A third is treating either as directional. Neither has a sign.
A fourth is leaving the lookback at a default. It came from elsewhere.
A fifth is assuming a low reading means safety. It means recent bars were small.
And a sixth is using a fixed stop distance instead of either. The instrument sets the scale.
Related
Keltner channels covers the range-based envelope. Standard deviation covers the close-based statistic. And average true range covers the measure the channel is built from.
The input is the whole story. One of these counts everything a bar did, including the jump from yesterday’s close; the other counts only where bars finished. For a stop that is not a preference, it is a different question.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.