Bollinger Bands — volatility and the squeeze
The answer
Bollinger Bands are a 20-period moving average with 2 lines drawn above and below it, each placed 2 standard deviations away. They measure how much this instrument has been moving recently — the bands widen when it moves a lot and narrow when it moves a little. Price touching the upper band is not a sell signal. In a strong trend, price rides the upper band for weeks.
Why this costs you money
The most common Bollinger Band strategy in retail trading is: sell at the upper band, buy at the lower band. It is presented as a mean-reversion system and it sounds reasonable, because 2 standard deviations sounds statistically extreme.
Two things make it expensive.
The first is that the bands adapt. They are calculated from recent price. When a stock starts trending hard, the standard deviation rises, so the bands widen, so price stays inside them. When a stock is trending and price "touches the upper band", it is not reaching an extreme — the band moved to meet it. In a sustained advance price can walk along the upper band for 20 or 30 sessions, touching it almost every day. Every touch is a sell signal by the rule. Every one of them loses.
The second is the statistics. The "2 standard deviations means 95% of observations" idea comes from a normal distribution. Daily stock returns are not normally distributed. They have fat tails, which means extreme moves happen far more often than a normal distribution predicts. The band is also calculated from only the last 20 days, so the standard deviation itself changes constantly. In practice, the share of closes falling outside 2-standard-deviation Bollinger Bands is often close to 5% but it is not a stable property and it varies by instrument and period.
The cost of both errors is the same and it is specific. A trader shorts a stock that is starting the strongest advance of its life, because a line that moved to follow the price told them the price was extreme.
A band touch is a statement about recent volatility, not about value, and the band was drawn using the very move you are trying to judge.
How it works
Three lines. The construction is simple and the middle line matters more than people expect.
The middle band is a 20-period simple moving average of the close. It is a plain moving average and it does everything a plain moving average does — defines direction, lags, sits in the middle of a range.
The upper band is the middle band plus 2 standard deviations of the last 20 closes.
The lower band is the middle band minus 2 standard deviations of the same 20 closes.
Standard deviation is a measure of spread. If the last 20 closes were all close together, the standard deviation is small and the bands sit close to the average. If the last 20 closes were scattered, the standard deviation is large and the bands sit far apart.
That is the whole indicator. The distance between the bands is a volatility measurement. The position of price inside them is a position measurement. They are 2 separate readings that happen to be drawn on the same 3 lines, and most misuse comes from confusing them.
Band width and the squeeze
Band width is the distance between the upper and lower bands, usually expressed as a percentage of the middle band.
When band width falls to a low level compared with its own recent history, that is called a squeeze. It means the last 20 closes have been unusually tightly clustered.
The squeeze is the most useful thing Bollinger Bands offer, and it is worth being precise about what it does and does not say.
A squeeze says: volatility has been low, and volatility tends not to stay low forever. It says nothing at all about direction.
Volatility clustering — quiet periods following quiet periods, violent periods following violent periods — is one of the better-established statistical properties of financial prices. That property is what gives the squeeze its content. It is also why a squeeze cannot tell you which way price will go: the statistics say the size of moves will grow, not their sign.
Riding the band
When price touches or hugs the upper band for many consecutive sessions, that is called riding the band. It is the signature of a strong trend, not of an extreme.
John Bollinger himself has been explicit that a band touch is not a signal by itself and that the bands should be combined with other, independent information . The person who built the tool says the popular use of it is wrong. That is worth more than any argument on this page.
What it tells you, and what it does not
Bollinger Bands tell you 2 things: how much the instrument has moved recently (band width) and where the last close sits relative to that movement (position).
Five limits.
They do not tell you value. There is no valuation input. A stock at the upper band can be cheap and a stock at the lower band can be expensive.
They do not tell you direction. Neither the squeeze nor the position tells you which way price goes next. The squeeze in particular is direction-neutral, and every squeeze breakout that "failed" was a trader supplying a direction the indicator never offered.
The 95% figure is not reliable. It assumes a normal distribution and daily returns are not normal. Treat the bands as descriptive, not probabilistic.
They are self-referential. The band that price is touching was calculated using price. A big move widens the band that is supposed to judge the big move. This is not a flaw to fix. It is what "adaptive" means, and it is why the tool cannot flag an extreme during the extreme.
They break in a range with a different failure than they break in a trend. In a trend, price rides a band and mean-reversion signals fail continuously. In a very quiet range, the bands narrow so far that ordinary noise touches them daily, and the tool produces a signal almost every session. Both failures produce a stream of losses and they come from opposite conditions.
The decision rule
Read band width first, and position second. Never position alone.
If band width is low compared with its own recent range — a squeeze — expect the size of moves to increase. Take no direction from the bands. Get direction from price structure or a trend indicator.
If band width is high and price is touching a band, that touch is information about trend strength, not an extreme. In an uptrend, repeated upper-band touches mean the trend is working.
Use band touches as a mean-reversion signal only when the middle band is flat and price has turned near the same 2 levels more than twice — the same range condition that makes any oscillator usable.
Unless the instrument has been at a price band or circuit limit. Then the standard deviation was computed from truncated data and the bands understate the real volatility.
Try this now
This takes 5 minutes. It uses 2 charts side by side, which is the fastest way to see an indicator's failure mode.
- Open 2 daily charts set to 1 year. From your own watchlist, pick 1 stock that trended strongly in one direction and 1 that moved sideways between roughly the same 2 levels.
- Add Bollinger Bands (20, 2) to both.
- On the ranging stock, count every session where the close was above the upper band or below the lower band. For each one, check whether price moved back toward the middle band within the next 5 sessions. Write the fraction.
- On the trending stock, do exactly the same. Write the fraction.
- Now find the narrowest band width of the year on either chart — the point where the 2 bands are closest together. Look at the following 20 sessions. Write down the size of the move that followed, and the direction.
What you should see. In step 3, on a genuine range, a decent share of band touches were followed by a move back toward the middle. In step 4, on the trending chart, you will find long runs of consecutive upper-band closes where price did not revert at all. Some of those runs will be 10 or more sessions. Every one of those sessions was a sell signal under the popular rule.
In step 5 you will usually find that the narrow point was followed by a larger move. Now look at your note on direction. Do this on 5 different charts and the directions will be roughly split. That is the honest content of the squeeze: the size of the move is somewhat predictable and the direction is not.
Bonus, 60 seconds. Change the setting from (20, 2) to (20, 1). Count the band touches on the trending chart again. You will get many times more. Change it to (20, 3) and you will get very few. The number of signals is a setting you chose.
Three real cases
1. John Bollinger, early 1980s (United States) — the author disagrees with the users John Bollinger developed the bands in the early 1980s while working as a market analyst, and the name became standard after he used them on financial television . In Bollinger on Bollinger Bands, published in 2001, he set out rules for their use, and he has stated repeatedly that a tag of the upper band is not a sell signal and a tag of the lower band is not a buy signal. The most widely taught Bollinger Band strategy is one the developer of the tool has publicly rejected. That is the single most useful fact about this indicator.
2. The NIFTY 50, April 2020 to October 2021 (India) — riding the upper band for months After the March 2020 low the Indian index advanced for roughly 18 months with only shallow interruptions. Across long stretches of that advance the index closed at or above its upper daily Bollinger Band repeatedly, in runs of many consecutive sessions. Under the popular mean-reversion rule, that period produced dozens of sell signals in the strongest Indian equity advance in recent memory. The bands were not wrong. They were reporting sustained high volatility with price at the top of it, which is what a powerful trend is.
3. The S&P 500, February and March 2020 (United States) — the bands widen after the move, not before Volatility on the S&P 500 was low in the weeks before the COVID crash. Bollinger Bands were narrow. Then the index fell roughly a third in about a month. The bands widened enormously — but they widened because of the fall, not in advance of it. The squeeze beforehand was real and it correctly said that volatility was compressed. It gave no warning of direction, and anybody who had guessed a direction from it had a 50/50 chance and nothing more. This is the clearest available demonstration that the squeeze predicts magnitude and not sign.
The question that resolves it
A novice sees price at the upper band and asks: is this stretched?
An expert asks: is the band wide or narrow, and is the middle band sloping?
Price at the upper band with narrow bands and a flat middle line is price at the top of a quiet range, and it may well revert. Price at the upper band with wide bands and a steeply rising middle line is a trend in progress, and the band moved up to meet the price rather than the price reaching up to the band.
The touch looks identical. What produced it does not.
What would make this wrong
The claim is that Bollinger Bands measure volatility rather than value, that band touches in a trend are signs of strength, and that the squeeze predicts magnitude but not direction.
You would falsify the direction claim with the exercise in step 5. Do it on 20 charts and record the direction after each squeeze. If the direction was predictable from the squeeze itself — not from price structure, not from a trend indicator, from the squeeze alone — this article is wrong.
You would falsify the trend claim with the exercise in step 4. If upper-band closes on trending charts were followed by reversion as often as they were on ranging charts, believe your count.
The honest limits are 4.
First, the squeeze does have some documented directional content when combined with the direction price breaks out. That is not the bands predicting direction. That is price telling you, and the bands telling you the move is likely to be large.
Second, "ranging" and "trending" are labels applied after the fact. The exercise is easier in hindsight than the decision is in real time.
Third, this article does not establish that Bollinger Bands make money in any condition. It establishes what they measure and where the popular reading of them is structurally wrong.
Fourth, 20 and 2 are conventions, and everything said here about signal counts is specific to them. A different setting is a different tool.
In India
Indian platforms ship Bollinger Bands at 20, 2 and Indian traders use them widely on daily charts and on 5-minute intraday charts.
There is 1 Indian condition that matters more than anything else in this article, and it is specific to volatility indicators.
Price bands truncate the data that standard deviation is computed from. Shares outside the derivatives segment carry daily price bands, commonly 2%, 5%, 10% or 20%. Shares in the derivatives segment have dynamic price bands that can be flexed during the session.
Here is what that does. A stock wants to fall 30% on bad news. The exchange permits 20%. It closes at the lower band. The standard deviation of the last 20 closes therefore records a 20% move, not a 30% one. Bollinger Bands are calculated from a censored sample, and censoring the extremes is exactly the way to make a volatility measure understate volatility.
The effect is worst in the situation where you most need the measure. A stock in distress hits its band repeatedly. Each day the recorded move is capped. The bands stay narrower than the real risk. A trader sizing a position using band width, or setting a stop at 1 band width, is using a number the exchange manufactured.
How to check for it in 10 seconds. Look at the daily candles. If you see several consecutive days where the close equals the high exactly, or the close equals the low exactly, with no wick, the stock was probably locked at a band. Any volatility reading from that period is understated.
Two smaller Indian points. Thin volume outside the largest few hundred NSE names means a single order can set an extreme close and widen the bands on its own. And Indian equities have no regular overnight session, so all 20 closes in the calculation came from continuous auctions, which is a genuine advantage over US data.
In the United States
The same 20 and 2, and 2 differences.
Extended-hours moves are excluded from the closes. Most US companies report after the close and shares trade before 9:30 and after 16:00 Eastern time. The standard daily candle uses only regular hours. So the volatility that Bollinger Bands measure excludes the extended-hours session, where a large part of the real movement on results days occurs. The close-to-close change is captured, but the intraday range that day is not, and any band width calculation understates the actual risk around results.
No daily price bands means no truncation. A US stock can fall 40% in one session and the whole move enters the standard deviation. American Bollinger Bands therefore widen far more dramatically than Indian ones on comparable news, and any threshold about "wide" or "narrow" bands taken from an American source will be miscalibrated for Indian charts.
The US has Limit Up-Limit Down pauses, which halt trading for 5 minutes when a stock moves too fast, and market-wide circuit breakers at index level. Those interrupt the path but they do not cap the day.
Where they differ, and what that tells you
The difference is direct and it is the sharpest India/US contrast in this whole cluster.
India censors the tails. The United States does not.
Bollinger Bands are a measure of the spread of recent closes. Indian price bands remove the largest closes from that sample by making them impossible. The result is a volatility indicator that reads calm during the most dangerous periods.
What that tells you is practical and it is about position sizing. If you size positions or set stops using band width — a common and otherwise sensible approach — the Indian number is systematically too small on exactly the stocks where being wrong is most expensive. A stock that has been locked at circuit limits has a narrow-looking band and unlimited real risk, because the risk that did not fit into today's session is still waiting for tomorrow.
There is a second consequence that is easy to miss. A squeeze in India can be manufactured. If a stock has been trading in a narrow band because it is illiquid and nobody is trading it, the bands narrow. That looks identical on a chart to a genuine volatility compression before a large move. In the US, deep liquidity means a narrow band is more likely to be real information about the market's current state.
Check volume before you believe an Indian squeeze. A quiet chart and an empty chart look the same through a volatility indicator, and only one of them is telling you something.
Carry this
- Bollinger Bands measure volatility. The width is the reading. The touch is a position, not a verdict.
- A squeeze predicts the size of the next move, not its direction. Anybody who gets direction from a squeeze got it from somewhere else.
- In India, check for circuit locks before trusting any band width. Truncated data makes a dangerous stock look calm.