How to Use Bollinger Bands
To use Bollinger Bands, read the width between them as a measure of how volatile the instrument currently is relative to its own recent history. A touch of a band is not a signal — in a trend price walks along the outer band for many bars at a time.
Bollinger Bands are a moving average with a volatility measure drawn either side of it. Almost every useful thing they do involves the distance between the bands rather than the bands themselves.
Before you start
A length and a deviation multiple you chose rather than inherited at 20 and 2. Those defaults are a starting point. Choosing them again deliberately is fine; never looking is not.
A decision about whether you are trading the squeeze or the touch, made in advance. They are different strategies with opposite logic, and having both available means always having a reason.
The instrument’s ordinary bar range, so band width can be compared to it. On this site’s shared series the median bar range is 0.493, the tenth percentile 0.17 and the ninetieth 1.101 — a 6.5-fold spread.
The steps
1. Read the width before anything else
Wide bands mean recent movement has been large; narrow bands mean it has been small. That is the indicator’s actual output and everything else is derived from it.
2. Classify the market from the width
Narrow relative to its own recent history is a squeeze. Wide is an active market. The strategy that applies differs completely between the two.
3. In a range, treat a touch as a location and not a signal
Price at the upper band in a range is at the top of recent movement. That is a place to look for a reason, not a reason on its own.
4. In a trend, expect price to walk the band
A strong move stays against the outer band for 10 bars or more. Selling every touch in that condition is fighting the only thing the indicator actually detected.
5. Trade the squeeze’s resolution, not the squeeze
A squeeze says volatility is low and will not stay low. It contains no directional information at all, so the entry has to come from the break, not from the compression.
6. Confirm against structure before acting
The bands mark no price anybody is trading against. A touch that coincides with a level that already mattered is a different event from a touch in open space.
7. Take the stop from the chart, not from a band
Bands recalculate every bar, so a stop placed on one moves with the indicator. Where the idea is wrong is a fixed structural question, and that distance sets the size.
How to tell it worked
You read the width before you read the position, every time. If the touch came first, the market type was decided after the signal.
A trend that walked the upper band for 10 bars produced no short trades. That is the classification step working.
Your settings are the ones you chose. Reviewing 20 and 2 and keeping them counts; inheriting them does not.
And every stop came from a level rather than from a band, so it did not move when the indicator recalculated.
What they cannot do
They cannot tell you direction. On this site’s shared series a round trip measures about 2% of the
median bar range of 0.493, so a strategy trading every band touch pays that repeatedly for an
observation about location. The figures are in research/series-measurements.json.
They also assume a distribution price does not have. Standard deviation describes a normal distribution, and real price movement has larger extremes than one — so band touches are more common than the statistics imply.
Reading width against the instrument’s own history
Width only means something relative to the same instrument’s recent width. A band separation of 2 points is wide on something that usually moves 0.5 and narrow on something that usually moves 5.
Express it as a ratio to the moving average and it becomes comparable across instruments. That is what a normalised width measure does, and it is the version worth watching rather than the raw distance.
On this site’s shared series the average true range has a median of 0.5994 against a tenth percentile of 0.2823 — a 2.82-fold spread within one instrument. Width has to be judged against that range, not against a fixed number.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 75 have an instruction-shaped
title about these bands, at a median of 5,505 views across 60 channels. The relative strength index
appears in 154 videos at 4,398 and the average true range in 135 at 11,637. The counts come from
site/rank_howto.py, which deduplicates by video id.
75 videos at a 5,505 median. Almost all of that coverage teaches the touch as a reversal signal, which is the interpretation that only holds in a range — and nothing in the indicator tells you which condition you are in.
The answer to the question on that chart is that the bands cannot answer it. They report that price is at the top of recent movement, which is true in a reversal and in a breakout. On this site’s series direction runs average 2.01 bars and the longest ran 11 — a touch is the start of a run as often as the end of one.
When it fails
The failure is a trending market traded as a range, and it repeats until the account notices. Price touches the upper band, a short is taken, price walks further up the band, the next touch looks like an even better opportunity. Every entry is stopped and the indicator was accurate throughout — it reported that price was at the top of recent movement, which it was, and it never claimed the move was finished.
The second failure is trading the squeeze directionally. It says when, not which way.
A third is placing a stop on a band. It recalculates every bar and the stop moves with it.
A fourth is inheriting the defaults. 20 and 2 are a starting point.
A fifth is treating the bands as levels. Nobody is trading against them.
And a sixth is expecting the statistics to hold. Real price has fatter extremes than a normal distribution.
Related
Bollinger Bands explains the calculation and what the deviation multiple does. The Bollinger squeeze is the width condition and how it resolves. And standard deviation is the statistic the bands are built from.
The habit worth building is reading the width first and the position second. Width tells me what kind of market I am in, which decides whether a touch means anything at all. Reading the touch first and the width afterwards is how the indicator gets used as a reversal signal in a trend, which is where it does the most damage.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.