How to Read Bollinger Bands: Bandwidth, the Squeeze and %B

The formula behind the bands

Bollinger Bands consist of a middle line and two lines above and below it. The middle band is the 20-bar SMA of the close, the upper band is the middle band + 2 × standard deviation, and the lower band is the middle band − 2 × standard deviation. The standard deviation measures how widely those same 20 closes are scattered around their average: take the mean of the squared deviations, then the square root. The calculation divides by 20 (the population standard deviation), which matches TradingView's default Bollinger Bands. Using the sample standard deviation, which divides by 19, makes the bands about 2.6% wider. When price swings hard the bands widen; when it is calm they narrow. What is easy to miss is that the width reflects dispersion, not direction. If price climbed steadily in one direction for all 20 bars, the closes are still spread far from their average, so the bands widen. And on the day a large bar from 20 bars ago drops out of the window, the bands can suddenly narrow even though price barely moved. If two charts show slightly different bands, first check that the period, the multiplier, the standard deviation method and the input price (close or something else) all match.

The claim that '95% stays inside the bands'

The explanation that 95% of prices fall inside the bands holds only if you assume prices are normally distributed around the mean. Under a normal distribution, the probability of landing outside the mean ± 2 standard deviations is about 4.6%. But prices lean to one side as they follow trends, and crypto returns see extreme moves more often than a normal distribution would predict (this is described as having fat tails). Of the 29,946 daily bars of 10 Binance coins measured for this course, 2,353 closed above the upper band and 1,516 below the lower band, about 13% combined. That is far more often than the normal-distribution assumption suggests. The reason lies in how market participants behave. When news breaks or a wave of liquidations hits, orders in the same direction arrive all at once, and that move draws in more orders on the next bar. Because price changes are not independent of each other, closes far from the average come in runs. So reading a close outside the bands as 'something rare happened, so price will snap back' can be wrong from the start. A close outside the bands records a large move relative to the last 20 bars; it is not evidence of a statistically improbable event.

%B and bandwidth

%B shows where the close sits within the bands. %B = (close − lower band) ÷ (upper band − lower band), so 1 means the upper band, 0.5 the middle band, and 0 the lower band. Bandwidth is (upper band − lower band) ÷ middle band; it turns the standard deviation into a ratio of price, so it can be compared across coins trading at very different prices. The Bollinger Band Squeeze Scanner shows both values, along with where the current bandwidth ranks within the last 120 bars. Either value on its own tells only half the story. The same %B of 1.1 can come from a small move when the bands are very narrow, but needs a large move when they are wide. So the natural order is to check first whether the bandwidth is narrow or wide compared with its own past, and then use %B to see the position. The usual level of bandwidth differs by coin and by timeframe, so a threshold borrowed from another chart will not fit. %B is calculated from the close, so if only the wick pokes outside the band and the bar closes inside, %B does not exceed 1.

• %B > 1: the close is above the upper band
• %B = 0.5: the close is at the middle band
• %B < 0: the close is below the lower band
• Bandwidth = 4 × standard deviation ÷ middle band (with a multiplier of 2)

The squeeze: when the bands narrow

A squeeze is when the bandwidth has become very narrow compared with its own past. John Bollinger pointed to times when the bandwidth was at its narrowest in several months, while the approach known as the TTM Squeeze uses the bands moving inside a Keltner Channel as its test. This site's squeeze scanner sets the Keltner Channel at EMA20 ± 1.5 × ATR(20). Volatility does not stay at one level for long and tends to drift back toward its usual level, so very narrow bands tend to widen again at some point. But a squeeze does not tell you when price will move or in which direction, and price sometimes heads one way first and then reverses (see the article on sideways markets). While the bands narrow, buyers and sellers are usually facing off at similar prices, and many participants sit on their hands ahead of a big announcement, so volume often dries up. When something breaks that balance, the orders and stops that had built up pour out in one direction, and the bands open quickly, as in the chart. The order of checks is: is the bandwidth on the low side of its own history, is the bar that left the range a closed bar, and did volume on that bar rise above normal.

Walking the bands: what a strong trend looks like

In a strong uptrend, closes hug the upper band or finish above it again and again. This is called walking the bands. Because the bands are calculated from the same prices, the middle and upper bands rise along with price, and in the chart below %B stays above 0.7 throughout. If you read each touch of the upper band as a sign of a pullback, you will be wrong for the whole trend. In a downtrend the mirror image appears, with price walking down the lower band. While a band walk continues, even when a resting bar appears, buyers tend to step in near the middle band and the pullback ends shallow. That is why many people use the middle band rather than the upper band to judge whether the trend is still intact. If closes stay above the middle band and %B rarely drops below 0.5, they treat it as the same move; if a bar closes below the middle band and the bandwidth starts shrinking from its peak, they read it as a sign the move is weakening. Even that does not confirm the trend is over, and price sometimes dips below the middle band briefly before walking the bands again.

Measured: after closes outside the bands

On the daily bars of 10 coins on Binance (each from its Binance listing date to September 2026), after bars that closed above the upper band, the share whose close was higher 20 bars later was 55.1%, above the 50.4% baseline measured over all bars, and the median was +2.35% (baseline +0.15%). A close above the upper band did not mean a pullback was coming. After closes below the lower band, the share higher 1 and 5 bars later was above the baseline, but 20 bars later it matched the baseline. Closes outside the bands also tend to come in runs, so keep in mind that the samples are not independent of each other. These results use the closes of finished daily bars; bars where only the wick crossed a band are not included. Because they come from daily bars, there is no basis for assuming the same on 1-hour or 15-minute charts, and they cannot be carried over to stocks or other coins as they are. The figures after closes above the upper band matter because they point the opposite way from the common reading that 'outside the band means overheated and about to turn'. Even so, flipping that around and reading a close above the upper band as a signal to expect gains goes beyond what these numbers allow.

• Baseline (all bars): 50.6% after 1 bar, 50.5% after 5, 50.4% after 20
• Above the upper band (2,353 times): 50.4%, 52.0%, 55.1%
• Below the lower band (1,516 times): 57.6%, 56.8%, 50.4%
• BTC daily, 20 bars later: above upper 59.8%, below lower 45.1%, baseline 53.7%

Limits and common misconceptions

Bollinger Bands are a picture of the average of 20 closes and how widely they are scattered; they are not a limit on how far price can go. Touching the upper band only summarizes that price is high relative to the last 20 bars, and a squeeze only says that movement may get bigger, not which way. Changing the period or the multiplier also changes how often closes land outside the bands. You can check past results, fees included, with the Bollinger reversal strategy in the coin strategy backtester, but there is no guarantee that one period's results will hold in another. One common mistake is treating the bands like support and resistance. Support and resistance form at prices where participants have actually stacked orders, but the bands are a statistical line recalculated on every bar, and there is no reason to think orders are clustered there. Another is expecting a big rally just because the bandwidth is narrow; narrow bands can resolve to the downside just as easily. Bollinger Bands are a ruler for whether the current movement is large or small compared with the recent past, and direction has to be judged together with other evidence such as trend, support and resistance, and volume.

How it looks different in crypto and stocks

Crypto trades around the clock, so a daily close is just the value cut at a time the exchange chooses (midnight UTC on Binance), and there are almost no gaps between bars. Stocks stop trading when the regular session ends, and the next day often opens away from the previous close, leaving a gap. When Samsung Electronics or SK hynix gaps open after an earnings report or a big overnight move in the US semiconductor index, that single day inflates the standard deviation and the bands widen in a step. Twenty bars also mean different things. Twenty daily stock bars are about a month of trading days, while 20 daily crypto bars are 20 calendar days. Large US tech stocks often go through a squeeze and an expansion around earnings, and because the dates are known in advance, it is easier to guess why the bands narrowed. Korean stocks have a daily price limit (±30%), so on a day locked at the upper or lower limit the close is pinned to the limit, and an extreme %B does not capture the full demand. Crypto bandwidth tends to be wider than that of large-cap stocks, so do not carry your sense of 'narrow' from one market to another; compare each with its own past.

A practical checklist

When you look at Bollinger Bands on a live chart, the first thing to separate out is that the current bar is still forming. The close of a forming bar is the current price, so the middle band, the upper and lower bands, %B and bandwidth all keep changing while price moves. A bar that seemed to break above the upper band during the session often returns inside by the close and leaves only a wick, and in that case it is not a close outside the bands. Past bars' bands are fixed, but the last bar's bands breathe with every tick; keeping this in mind reduces the confusion of 'the signal I just saw has disappeared'. All of this course's measurements use the closes of finished bars. If price moves sharply in the last few minutes before a bar closes, the last bar's bands and %B can change a lot, so it helps to look again at the confirmed values once the next bar has opened. Here is an order of checks for the screen.

• Check that the timeframe and settings (20, 2, close) are what you intended
• Start with whether bandwidth is narrow or wide versus that market's recent past
• Use %B for the close's position, remembering the forming bar's value is not final
• Judge closes outside the bands on finished bars, and tell them apart from wicks
• When a squeeze breaks, check that bar's volume and whether there was a gap
• Use whether the middle band holds to see if a band walk is continuing
• Open the same scene on a higher timeframe to see if the picture agrees