Standard Error Bands: Fit, Not Volatility
Standard error bands sit a multiple of the standard error either side of a linear regression line fitted to recent prices. The width measures how well that line fits, not how far price moves. So an orderly trend narrows them, and a quiet, directionless market widens them.
Covered on this page: TradingView.
How it works
Standard error bands are drawn around a fitted line, not an average. The center is a linear regression through the recent closes, and the bands sit a multiple of that fit’s standard error either side of it.
The width is the error of the regression itself. Standard error summarizes how far the actual closes sit from the line drawn through them. Nothing else enters the number.
Which is not the same as price volatility. Bollinger Bands widen when price moves a lot; these widen when price stops following the line, and a fast, orderly trend can produce narrow bands and a large daily range at once.
A good fit gives narrow bands even in a fast market. If every close lands near the sloping line the residuals are small and the channel is tight, however far price has traveled in the meantime.
So they narrow when the trend is orderly. That is the reverse of what a volatility band does, and it is the single fact that makes the indicator worth learning at all.
Reading the width the right way round
And widen when price stops following the line. A quiet, directionless market fits a straight line badly, so the channel can be at its widest while the bars themselves are small.
Most versions smooth the result, which adds lag. Published implementations run a moving average over both the regression line and the error, which steadies the picture and delays every change in it.
Participation is nowhere in the calculation. Volume enters neither the regression nor the error, so a tight channel says the closes were orderly, not that anyone was trading them.
In practice
A longer window makes everything steadier and later. The lookback is a free parameter, and lengthening it flattens the slope and calms the width at the cost of reacting a good deal later.
And one gap inflates the error for the whole window. A single opening gap leaves a large residual that widens the channel for as long as that bar stays inside the lookback.
The opposite band is the only level on offer. Nothing in a regression channel marks where orders rest, so any stop loss other than the far band is a number you chose yourself.
Every round trip costs the spread and any commission. A band that is touched often invites frequent entries, and each one pays that toll whether or not the channel holds.
It is a goodness-of-fit measure wearing a channel. The whole regression is recomputed every bar, so the historical portion of the channel is not sitting where it sat yesterday.
Reading both band families at once
The two families answer different questions, so read them side by side. Bollinger Bands and Keltner channels both scale with movement — one from standard deviation, one from the average true range, or ATR. Standard error bands scale with disorder.
Volatility bands widening while error bands narrow is the clearest reading available. Price is moving hard and moving straight, which is what an orderly trend looks like from both angles at once. It is also the point at which the Bollinger squeeze idea inverts.
Both widening is the genuinely disorderly case. Large bars going nowhere in particular: historical volatility is up and the fit is poor, which is the condition that punishes trend analysis and flatters mean reversion.
And a trader treating the two as interchangeable is reading different things off the same-looking picture. A narrowing means a coming move on one and settled order on the other. The picture is identical; the meaning is inverted.
What standard error bands are not
- Not a volatility measure. The width scores fit quality, which is a different quantity entirely.
- Not a squeeze indicator. Narrowing here describes an orderly trend, not a move about to arrive.
- Not a support and resistance tool. The lines are computed each bar and nobody is defending them.
- Not a fixed channel. The fit is redrawn every bar, so the history you are looking at moves.
How this connects
A channel drawn around a linear regression line rather than around a moving average, which is the whole difference from Bollinger bands.
The center line is fitted, not averaged. A regression finds the best-fit slope through the window, so the center has direction built into it.
Standard error rather than standard deviation sets the width, which measures how well the line fits rather than how spread the data is.
Narrow bands mean price has been tracking a straight line closely - a trend-quality reading rather than a volatility one, which is a genuine difference.
And it repaints as the window moves, so the historical channel you see was never the channel at the time.
When it fails
- In a range the fit is poor and the bands are wide. A straight line through sideways trade leaves large residuals in both directions, so the channel is at its widest exactly where there is least to trade. Read as expanding volatility, that is precisely backwards.
- When the window is chosen after the event. The lookback has no natural value, and a channel that fits beautifully at one length is ragged at another. Nothing in the indicator says which length is right.
- When a single outlier dominates. Standard error is built from squared residuals, so one violent bar can widen the entire channel and hold it wide until that bar leaves the window.
- When the historical portion is read as fixed. The fit is recomputed every bar, so the channel on screen is not the channel that existed when those bars printed. Testing it by eye flatters it.
- When smoothing hides the change you needed. The lag applied to both line and error means the bands look steadiest at the moment an orderly trend is coming apart.
- When direction does not persist. Daily direction on a broad index is close to even, with 53.1% of S&P 500 sessions closing up from 1950 to 2026, so the orderly condition these bands reward is a short-lived one.
The original data
The coverage figures come from the site’s video study. A scan of the 24,971 videos collected for this site’s search study finds “bollinger” in 268 titles from 204 channels, median 4,842 views, against a single “standard error” title, 2,309 views, one channel. Both draw a channel; only one has a memorable name.
The more useful comparison is between the two measurements. Average true range measures how far price moves per bar; a standard error band measures how well a straight line fits the recent closes. Before acting on any band width, ask which quantity it measures — if the answer is fit quality, a narrowing describes order, not a coming move.
Related
Linear regression is the line these bands are drawn around, and the width is nothing more than how badly it fits.
Bollinger Bands measure dispersion around a moving average instead, which is why the two families widen at opposite moments.
Standard deviation is the quantity that separates them, since one applies it to price and the other to the residuals of a fit.
I do not read these like Bollinger Bands, and that is the mistake to avoid. A pinch here is not a coiled breakout; it is a trend behaving itself, because the width is scoring the fit rather than the movement. Not all bands measure the same thing, and assuming they do makes the indicator look like it is lying to you.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.