What Are Bollinger Bands?
Bollinger Bands are a moving average with a channel drawn two standard deviations above and below it, usually over 20 bars. The width of the channel measures recent volatility, so the bands widen when price is moving and narrow when it is quiet.
Two lines above and below a moving average, set by how much price has been moving. Everything useful about them is in the width, and most of what is written about them is about the edges.
How it works
Three lines. The middle one is a 20-bar simple moving average of the closes — verified against a plain 20-bar average on this data, they are the same number.
The outer two are that average plus and minus two standard deviations of the same closes. On the last bar of this chart the standard deviation is 0.55, so the bands sit 1.09 either side.
That is it. There is no second idea.
Width is the whole point
Standard deviation is a measure of spread, so the channel is wide when price has been moving and narrow when it has not.
On this chart the width ran from 0.82 to 2.66 — the widest reading is 3.2 times the narrowest, on the same instrument a few bars apart.
That is genuinely hard to see by eye and it is the one thing these bands give you that nothing else on a plain chart does.
The squeeze
A squeeze is the channel at a local minimum, and the observation attached to it is that quiet periods tend to be followed by loud ones.
It contains no direction. A squeeze is a statement about size, and the move out of it goes whichever way it goes — which is the same honest limit as the compression section on the price action page.
Anyone telling you a squeeze is bullish is adding something the arithmetic does not contain.
The mistake everybody makes
The bands are not a ceiling and a floor. They are a description of the last twenty bars.
On this chart price closed above the upper band 6 times and kept going. Selling the first touch would have been wrong, and so would the next five.
Here is why it happens. Two standard deviations means “unusual relative to the recent past.” In a trend, closing near the top of the recent range is the recent past — the band moves up to meet price rather than stopping it.
The settings
Two numbers: the length of the average and how many deviations out to draw.
The deviation setting decides how often anything happens. Measured on this chart: 6 closes outside the two-deviation band, and 1 outside the three-deviation band.
So “price is outside the bands” is a statement about your settings at least as much as about the market. Six events or one, same data.
The default 20 and 2 is worth keeping for exactly the reason on the support and resistance page — it is what everyone else is looking at.
The two derived readings
Bollinger published two numbers alongside the bands, and both are more usable than the bands themselves.
%B says where price sits inside the channel, as a number: (price − lower) / (upper − lower).
One is on the upper band, zero is on the lower, and 0.5 is the basis. On the last bar of this chart
it is 0.51 — dead centre, which is a fact the picture makes you squint to get.
Bandwidth normalises the width: (upper − lower) / basis. That matters because a raw width of
2.66 means something completely different on a $10 stock and a $4,000 one, and only the normalised
version can be compared across instruments or across years. On this chart the basis sits near 100,
so the two happen to read almost the same — which is exactly the coincidence Bandwidth exists to
stop you relying on.
A worked example
The bands narrow. Volatility has fallen. No trade, no direction, just a note that something is building.
Price closes outside the upper band with the channel expanding. Both halves matter: outside a narrowing band is stretched, outside a widening one is a trend starting.
The basis becomes the reference. In a trend, pullbacks to the middle line are where being wrong is cheap — that is what the middle line is good for, and it is a moving average doing an ordinary moving average job.
The invalidation is a close back below the basis. Not a touch of the lower band, which in a strong move may not happen for weeks.
The original data
Across our study of 24,971 trading videos, 384 cover Bollinger Bands. The median one gets 3,516 views, 85% never pass 50,000, and the median length is 9.2 minutes.
That 85% puts it among the more saturated topics measured here — behind Chaikin money flow at 96% and Hull moving averages at 92%, and still very high: 384 videos explaining a formula that has not changed since John Bollinger published it in the 1980s.
The corpus carries description text for 73 of those 384, and across those 73, four mention invalidation, failure, or what a bad read looks like.
When it fails
Price walks the band
Covered above and repeated here because it is the failure that costs money. Six closes outside the band on this chart, and the move did not end at any of them.
The band is not resistance. It is a distance from an average.
The squeeze breaks the other way
Quiet then loud is the observation; quiet then up is a guess. A squeeze resolving downward is exactly as consistent with the tool as one resolving upward, and no amount of staring at the narrowing changes that.
You changed the setting until it agreed
Two deviations gave six events here and three gave one. If you tune the number until the chart shows what you already believed, you have not measured anything — the argument on the technical analysis page, in its most tempting form.
You looked once it had resolved
Every squeeze is obvious in retrospect and directionless at the time. Cover the right-hand side and there is nothing here that says up.
Related
Moving average is the middle line, and understanding it is most of understanding this.
Trading range is the condition a squeeze usually describes, read off price instead of arithmetic.
And breakout is what the expansion out of a squeeze usually turns out to be, including how often it fails.
The only thing I take from these is the width, because that genuinely tells me something I cannot easily see otherwise, which is whether the last twenty bars were quiet or wild. What I do not do is fade the bands. I lost money early on selling every touch of the upper one in what turned out to be a strong trend, and the reason is now obvious to me - in a trend, closing near the top of the recent range is the normal condition, not an extreme.
— Michael Whitman, from this video
This page is educational, not financial advice. Test every idea on your own charts before risking money.