What are Bollinger Bands?

Bollinger Bands are three lines, and each one is defined by a short formula. The middle band is a simple moving average of closing prices over N periods, where N is 20 by default. The upper and lower bands are that same average shifted up and down by K times the standard deviation of price over the identical N periods, where K is 2 by default. Standard deviation is the statistical yardstick of dispersion: when recent closes are scattered far from their average it is large, and when they cluster tightly it is small. That is the whole mechanism.

The consequence is the part worth dwelling on. The envelope is not drawn at a fixed rupee distance, and it is not drawn at a fixed percentage either. It is drawn at a distance proportional to how dispersed the last twenty closes have been, so it inflates and deflates on its own with no input from you. A fixed-percentage envelope would sit at the same width through a dead fortnight and a panic; the standard-deviation envelope cannot, because the same arithmetic that places it is fed by the very prices it is wrapping. The bands are an output of price, not an overlay on it.

Two details of the construction repay attention. The first is that the same twenty closes feed both the average and the dispersion, so the middle band and the band distance move together and are not independent readings. The second is that the standard deviation used is the population form, computed over the twenty closes in the window rather than treating them as a sample drawn from something larger. Neither detail changes how you read the picture, but both explain why the bands react the way they do: everything you see is a function of one short, rolling window, and the moment a violent bar enters that window the envelope widens, whether or not the violence continues.

The choice of two standard deviations is not arbitrary, and it is not a law either. For a variable that follows a normal, bell-shaped distribution, about 95 percent of observations fall within two standard deviations of the mean. Carry that statistic across to price and you get the familiar claim that price spends almost all of its time inside the bands, so a move beyond them is unusual and worth noticing. That is the intuition the default encodes, and it is where a great deal of loose teaching stops.

The caveat is that price is not normally distributed. Real returns have fat tails, meaning extreme moves happen far more often than a bell curve predicts, and they cluster: calm begets calm and turbulence begets turbulence. So the clean 95 percent does not survive contact with a real price series. On the standard 20 and 2 setting, published descriptions of the indicator note that the bands tend to contain closer to 88 to 89 percent of price action, not 95. That gap of six or seven percentage points is not a rounding error and it is not a flaw to be tuned away; it is the honest character of markets showing through a tool that borrowed its language from a distribution markets do not obey.

Read the containment figure as a rule of thumb that makes an out-of-band move worth a second look, never as a probability you can bank on. There is a further reason to hold it loosely: the bands are computed from a rolling twenty-bar window, so the reference against which "unusual" is judged is itself moving. An out-of-band close during a placid stretch and an out-of-band close during a panic are the same event by the indicator's definition and completely different events in the market. Anyone quoting a precise hit rate off the bands is over-claiming what a standard-deviation envelope can deliver.

The bands borrowed their vocabulary from a bell curve. Markets never agreed to the loan, and the six-point gap between 95 and 89 is the interest.

Read correctly, the bands answer two different questions, and they are not equally useful. The first is relative position: price near the upper band is high relative to its recent range, price near the lower band is low relative to that range. This is a statement about where price sits against its own short history, and nothing more. The tidy way to express it is %B, which places price on a 0 to 1 scale between the bands. It reads 1 at the upper band, 0 at the lower, 0.5 at the middle average, and it spills above 1 or below 0 whenever price closes outside the envelope entirely, which on real data it does often enough to matter.

The second question, and the more important one, is how wide the envelope is, because the width is the live measure of volatility. That is captured by bandwidth: the distance between the upper and lower bands, divided by the middle band to normalise it so that different instruments and different price levels can be compared on one scale. When bandwidth is high, the market is volatile and the bands are far apart. When bandwidth collapses to an unusually low reading, volatility has drained out and the bands have pinched together.

Here is the asymmetry that most explanations miss. Relative position is a weak reading, because it is defined against a window that is itself adapting, and because being high in a range and being high in a trend are opposite situations that produce the same number. Width is a strong reading, because it is measuring something real and physical: how much the market is actually moving. If you take one thing from the bands, take the width. The level of price inside them is the part that invites the mistake, and the width is the part that does honest work. Bandwidth is also the only Bollinger reading that must be judged against the instrument's own history rather than an absolute threshold, because a bandwidth of 3 percent is a coiled spring on one instrument and an ordinary day on another.

The squeeze is the most celebrated Bollinger idea, and the one most often mistold. It is simply the state where bandwidth falls to a local extreme low: volatility has compressed so far that the two outer bands draw into a tight neck around the average. The logic that makes it interesting is the empirical tendency for volatility to revert and cycle. Extended calm is unusual, it does not persist indefinitely, and it tends to resolve into a burst of movement. So a pronounced squeeze historically precedes a volatility expansion, a decisive move that throws the bands back apart.

That much is genuinely useful, and it is worth being precise about what kind of information it is. A squeeze is a statement about timing and likelihood: that a move is more probable than usual, and probably soon. It is completely silent on direction. The same tight coil can erupt upward or downward, and nothing in the arithmetic that produced the neck contains a preference between the two. Standard deviation is computed from squared deviations, so it discards the sign of every move that fed it. A tool built by squaring away direction cannot hand direction back at the end.

There are two further honest limits. Squeezes produce false starts: price poking out of the neck one way, pulling traders in, then reversing and running the other way, which is a recognised enough pattern that Bollinger discusses the head fake explicitly. And a squeeze has no clock. It tells you the spring is compressed, not what day it releases, and a coil can stay coiled far longer than a position sized for an imminent move can comfortably wait. Using the squeeze well means treating it as a reason to pay attention and prepare for both directions, then turning to price structure for the direction, rather than asking the bands for an answer they do not contain.

Here is the mistake that separates using Bollinger Bands from misusing them. The bands are drawn where price is statistically stretched, so it is tempting to read a touch of the upper band as "too high, sell" and a touch of the lower band as "too low, buy". John Bollinger himself rejects this. In his own published rules he states that tags of the bands are just tags, not signals: a tag of the upper band is not, in and of itself, a sell signal, and a tag of the lower band is not a buy signal. The bands mark relative position. They do not mark a turn.

The reason the fade fails is walking the band. In a strong trend, price does not oscillate politely between the bands; it clings to the outer one. In a powerful uptrend price rides the upper band, tagging it again and again for many bars while the whole envelope drifts higher, and in a downtrend it walks the lower band down. Every one of those tags looks like an overbought sell to someone fading the band, and in a trend almost every one of them is wrong, because the tag is confirming strength rather than exhaustion. Selling each upper-band touch in a trend means shorting a market that keeps going up, which is precisely how a mean-reversion reflex gets run over.

The figure below is deliberately not a one-sided argument, because the honest version is stronger. Of the twelve upper-band tags in that trend, nine were followed by more upside and three sat at the high. That is the actual lesson, and it is sharper than "band touches always continue". The bands really do become stretched near the end of a move, and a fader will occasionally be right. What the indicator cannot do is tell the nine from the three, because the tag itself is identical in both cases. A signal that fires on every one of twelve occasions and is right on three of them is not a signal; it is a description of where price is, which is exactly what Bollinger says it is.

The reason Bollinger Bands confuse people is that the identical event, a tag of a band, means opposite things in the two market regimes, and the bands themselves do not label which regime is in force. In a range, the envelope behaves like a rubber container and price tends to revert from the edges back toward the middle average. In a trend, the envelope is a moving conveyor and price rides its leading edge. Knowing which story you are in sits upstream of the indicator, and it is the entire game.

Notice what this does to the idea of a band-touch rule. A rule that says "sell the upper band" is not one rule with a mixed record; it is two different rules wearing the same clothes. In a range it is a mean-reversion trade with some logic behind it. In a trend it is a counter-trend trade against a market that is telling you, through the very tag being used as the trigger, that it is moving with conviction. The two are not variations on a theme. They are opposites, and the indicator hands you the same number for both.

This is why Bollinger Bands are a context tool rather than a trigger. They are excellent at one honest job: showing you at a glance how volatile an instrument is right now, and where price sits inside that volatility. They are silent on the question that actually decides a trade, which regime you are in and which way the next move breaks. That upstream judgement, reading trend and structure before reading any indicator, is exactly what the method we teach is built around. The indicator is the easy part. The context is the skill, and no setting on the bands will supply it for you.

The mechanics of the bands are universal, but the conditions an Indian retail trader applies them in are not, and a few of them bite harder than the textbook suggests. The first is the gap. Bollinger Bands are computed from closes, and an index or a stock that closes inside the envelope and opens well outside it the next morning has not violated anything the indicator can see. Overnight and weekend risk lands as a single jump between two closes, and the band simply widens afterwards to accommodate it. On instruments that regularly gap on global cues or on results, the envelope is describing a continuity that the tradeable session does not have.

The second is the volatility calendar. Index and index-derivative activity clusters heavily around expiry, and volatility that is scheduled behaves differently from volatility that arrives at random. A squeeze that resolves into an expiry-week expansion is not evidence that the coil predicted anything; the calendar predicted it, and the bands reported it afterwards. The same applies around scheduled events. A bandwidth reading tells you what has already happened to dispersion over the last twenty closes, so on a known event date it is close to the last thing that will inform you about the day ahead.

The third is scale. Bandwidth is normalised by the middle band precisely so that instruments at different price levels can be compared, but that does not make a single threshold portable. A broad index is structurally calmer than a mid-cap constituent, so the bandwidth that counts as a squeeze on one is unremarkable on the other, and any number carried over from a foreign chart or a stranger's screenshot is meaningless. Squeeze is a percentile against an instrument's own recent history, not an absolute. The fourth is leverage, and it is the one that converts a reading error into a capital event: derivatives magnify both the loss and the emotion attached to it, and the regulator's own study found that about 93% of individual traders in equity derivatives made net losses over FY22 to FY24, aggregate net losses exceeding ₹1.8 lakh crore (SEBI, September 2024). An envelope that describes volatility is a poor foundation on which to take leveraged directional risk, because describing volatility is all it does.

A fair summary is that Bollinger Bands do one thing precisely and are routinely asked to do a second thing they cannot. The thing they do is measure volatility and relative price. The width tells you how turbulent the market is, %B tells you where price sits inside that turbulence, and an extreme low in bandwidth flags that calm has stretched unusually far. Used that way, as a volatility lens read alongside trend and structure, they earn their place on a chart. Like every indicator built from past prices, they summarise what has happened; they do not forecast what will, and John Bollinger frames them not as a source of buy and sell signals but as a framework for identifying setups where the odds may be more favourable.

The mislead is always the same shape: treating a band as a line the market is obliged to respect. The band touch is not a sell. The lower band is not a buy. The squeeze does not name a direction. The clean 95 percent is a rule of thumb that real, fat-tailed prices violate roughly one close in nine. Every one of those errors comes from the same root, which is reading an output of price as though it were an authority over price. The bands are a mirror held up to the last twenty closes. A mirror is genuinely useful, and it has never once told anyone what happens next.

So the honest ledger is short. Read plainly, Bollinger Bands hand you a well-calibrated picture of volatility and stretch, and then hand the actual decision straight back to your reading of price. That is not a weakness of the tool, and it is not a gap to be filled by a second indicator bolted alongside it. It is an accurate account of what a standard-deviation envelope can and cannot know, and a trader who accepts that boundary will get more out of the bands than one who keeps asking them for a verdict they were never built to give.

• John Bollinger, the 22 rules of Bollinger Bands. The originator's own rules: Rule 6 states tags of the bands are not signals, so an upper-band tag is not a sell and a lower-band tag is not a buy; Rule 9 sets the defaults of 20 periods and 2 standard deviations; Rules 15 and 18 define %B and BandWidth; Rule 22 frames the bands as a framework, not continuous advice. bollingerbands.com
• Bollinger Bands construction and containment. The middle band as a 20-day simple moving average, the upper and lower bands at plus and minus two standard deviations, the note that the 20 and 2 setting contains roughly 88 to 89 percent of price action, and the descriptions of walking the bands, the Bollinger Band Squeeze and the head fake that can follow it. stockcharts.com
• %B and the two-standard-deviation intuition. The %B formula, price minus lower band over upper band minus lower band, and the statistical basis for the two-standard-deviation default. fidelity.com
• Volatility clustering and fat tails. The empirical regularities that make the 95 percent figure a rule of thumb rather than a law: large changes tend to be followed by large changes and returns are not normally distributed. Benoit Mandelbrot, The Variation of Certain Speculative Prices (Journal of Business, 1963). jstor.org
• Indian retail derivatives outcomes. The regulator's study of individual traders in the equity derivatives segment, the source of the loss figure quoted above. sebi.gov.in