What multiplier should the channel use?

Two standard deviations is the common default, and with roughly normal residuals it contains most of the price action in the window. The multiplier scales the width linearly, so for comparing width across time it matters less than keeping it constant.

Why does the channel width change when I change the lookback?

Because both the fit and the residuals are window-dependent. A short window fits recent action closely and prints narrow; a long window straddles multiple swings and prints wide. Width readings are only comparable at a fixed lookback.

Is a narrow channel bullish?

It is neither bullish nor bearish; it says the current trend, whatever its direction, is orderly. Direction comes from the regression slope. A narrow rising channel and a narrow falling channel are equally clean trends.

How is this different from just using standard deviation of price?

Raw standard deviation of prices over a trending window is inflated by the trend itself. Measuring dispersion around the fitted regression line removes that drift, leaving only the fluctuation traders actually have to survive within the trend.

Standard-deviation Channel Width

Top Standard-deviation Channel Width indicator

The top custom implementation, built on the original standard Standard-deviation Channel Width formula.

1 total

Open the Standard-deviation Channel Width implementation below on Vela, then have Quant, our coding agent, turn it into a strategy you can backtest.

Open in Vela

Standard-deviation Channel Width

Indicator

What is standard-deviation channel width?

Standard-deviation channel width is the distance between the rails of a linear regression channel whose bands are set a multiple of the standard deviation of residuals above and below the fitted line. Where the regression line summarizes the trend over the lookback, the width summarizes how loosely price has traded around that trend: a wide channel means large excursions from the fit, a narrow one means price has hugged its trendline.

The reason it exists as a distinct reading is that ordinary volatility measures conflate two different things: movement along a trend and movement around it. A steadily climbing market can post large daily ranges yet stay pinned to its regression line, printing a narrow channel. Detrending through the regression first isolates the noise component, which is often the part that matters for stop distance and pullback depth within an established trend.

Traders watch the width both as a level and as a trajectory. Narrowing width shows an increasingly orderly, efficient trend; widening width shows the trend loosening, which often precedes either a regime change or a transition into a trading range. Comparing the current width to its own history, in percent-of-price terms, keeps the reading meaningful across instruments.

How it's calculated

Fit a linear regression over the lookback, measure the standard deviation of residuals, and offset the rails by a multiple of it.

fit: y_i = a + b * i, over the last n bars (least squares)

resid_i = P_i - (a + b * i)

sd = sqrt(sum(resid_i^2) / n)

Upper_t = (a + b * i) + k * sd

Lower_t = (a + b * i) - k * sd

Width_t = 2 * k * sd

NormWidth_t = 100 * Width_t / Mid_t

n: regression lookback in bars (commonly 100)

i: bar position within the lookback window

P_i: price at position i (usually the close)

a: regression intercept

b: regression slope per bar

resid_i: residual of price versus the fitted line

sd: standard deviation of the residuals

k: deviation multiplier (commonly 2)

Width_t: absolute channel width in price units

Mid_t: value of the regression line at bar t

NormWidth_t: width as a percent of the midline

Some platforms divide by n - 2 rather than n when estimating residual dispersion; values differ slightly.

The whole fit is recomputed each bar, so past channel values repaint as the window rolls forward.

How traders use it

As a trend-quality gauge: a narrow channel relative to its own history marks an orderly trend where pullback-to-the-line entries have tended to work; a widening channel warns that the trend is getting sloppy, similar in spirit to a falling Kaufman efficiency ratio.
For stop and target scaling inside a trend: because the width measures typical excursion around the trend rather than raw movement, it sizes stops that tolerate normal noise without also absorbing the trend's own drift.
As a compression screen on trending instruments: unusually narrow normalized width flags coiled trends, a complement to BandWidth, which measures compression around a moving average instead of a fitted line.
Its main caveats: the reading depends heavily on the chosen lookback, the fit repaints as new bars arrive, and a linear fit is a poor model of a curving or rangebound market, in which case the width mostly measures model error.

Standard-deviation channel width vs related measures

Standard Error Channel: Standard error channels offset the rails by the standard error of the regression estimate, which shrinks with sample size, so they hug the line far more tightly than standard-deviation rails at the same lookback. The two widths answer different questions: dispersion of price versus uncertainty of the fit.

BandWidth: BandWidth measures Bollinger Band spread around a moving average, so it mixes trend drift into the dispersion when price is trending. Channel width detrends through the regression first, isolating movement around the trend.

Donchian Width: Donchian width is simply the highest high minus the lowest low of the lookback, dominated by the two extreme prints. Standard-deviation channel width uses every bar's residual, so single outliers move it far less.

Standard-deviation Channel Width FAQ

Turn Standard-deviation Channel Width into a trading strategy.

Open the implementation from this page in Vela. Quant, our coding agent, builds the strategy with you and backtests it on real data.

Build a strategy in Vela

Browse this familyVolatility61 concepts mapped · 61 in the Library

Previous conceptSqueeze Release Direction Next conceptSTARC Bands