Bollinger Bands: How to Read the Bands, Set Them and Stop Misusing Them

Bollinger Bands are the indicator most people have on their chart and fewest people have actually read. Three lines, an easy rule that sounds like it works (price hits the top, sell; price hits the bottom, buy), and a lot of blown trades from believing it. This post is the foundation: the formula worked out on numbers you can check, what %B and BandWidth measure, what it looks like when price walks the band, why a touch is not a signal, and what the 20 and 2 in the default are actually doing.

Quick Answer

Everything below is either the arithmetic behind that paragraph or the places it gets misused.

How Bollinger Bands Are Calculated

The StockCharts ChartSchool article on Bollinger Bands gives the construction in three lines. The middle band is the 20-day simple moving average. The upper band is that average plus the 20-day standard deviation of price times 2. The lower band is the average minus the same amount. If the simple moving average itself is new to you, the simple moving average guide covers the middle line in full.

Here it is on twenty hypothetical closes. They are made up so the arithmetic is easy to check, not market data. The 20 closes add to 2,046.8, so the middle band is 2,046.8 / 20 = 102.34. Next, each close's distance from that average, squared, as in the table. The squared distances add to 37.528, and dividing by 20 gives a variance of 1.8764. The square root of that is the standard deviation, 1.3698.

Now the bands. Two standard deviations is 2.74, so the upper band is 102.34 + 2.74 = 105.08 and the lower band is 102.34 - 2.74 = 99.60. The last close, 105.2, is above the upper band. That is what a band tag looks like: price finishing a little outside a line drawn from its own last 20 closes, while the whole series has been climbing.

A note on the divide-by-20 step. This example divides the squared distances by n (the population standard deviation). Dividing by n minus 1 gives a slightly larger sample standard deviation, here 1.4054 instead of 1.3698, which would put the upper band at 105.15 instead of 105.08. I am not claiming which version any particular charting platform uses. If your bands are a hair different from a hand calculation, check that first.

Because the standard deviation is built from squared distances, one big move pushes the bands out faster than a lot of small ones, and the bands snap back inward only once that bar leaves the 20-bar window. In plain terms, the bands widen when volatility rises and contract when it falls.

%B, BandWidth and What the Width Tells You

Two derived numbers turn the picture into something you can read off a value. Both are on the same bands, and they answer different questions.

%B: where is price inside the bands?

StockCharts defines %B as (price minus lower band) divided by (upper band minus lower band). It is above 1 when price is above the upper band and below 0 when price is below the lower band. Price on the middle band reads 0.5. On the worked series, the last close gives (105.2 - 99.60) / (105.08 - 99.60) = 1.02, a hair over 1.

BandWidth: how wide are the bands?

StockCharts defines BandWidth as (upper band minus lower band) divided by the middle band, times 100. It is the band gap as a percentage of the average, so you can compare it across prices and across time. On the worked series it is (105.08 - 99.60) / 102.34 x 100 = 5.35%. StockCharts notes that falling BandWidth reflects decreasing volatility and rising BandWidth reflects increasing volatility.

To see what the width is actually measuring, take two charts that both average exactly 100 over 20 bars. One alternates between 99.5 and 100.5, the other between 98 and 102.

Same center line, four times the width. That is the whole message of the band gap: it measures recent volatility, nothing about direction. Wide bands say price has been moving a lot over the last 20 bars. Narrow bands say it has been quiet. Neither says which way the next move goes. The narrow case is the one traders call a squeeze, which StockCharts describes as volatility falling to a low level. We give it its own page, so here it is in two sentences: a squeeze is the BandWidth reading low relative to its own history, and the open question is the direction and the quality of the breakout that follows. How to size up that setup, and where the false starts come from, is in the Bollinger Band squeeze guide.

Walking the Band and the With-Trend Read

This is the part the beginner rule gets backwards. StockCharts, relaying Bollinger, says that in a strong uptrend it is common for price to walk the band, with numerous touches of the upper band, and that in an uptrend it is also common for price to never reach the lower band. The diagram above is a hypothetical series built to show it. After sixteen quiet bars, price trends up. From bar 20 to bar 27 (eight bars in a row) the close finishes above the upper band, with %B between 1.00 and 1.22. An entry rule that sold the first tag of the upper band would have been fighting the move for eight bars.

The with-trend read uses the same three lines the other way round. In an uptrend, the upper band tells you the trend is strong (price keeps reaching it), and the middle band is the place to watch for a pullback. In the series, price falls from 110.8 on bar 27 to 105.9 on bar 31, against a middle band of 104.73, a %B of 0.58. The lower band never comes into play: the lowest %B in the whole stretch is 0.58. Then the series resumes up and %B climbs back toward 1.

That pullback is a place to look, not a buy button. Whether it holds depends on the same things any pullback depends on: a higher low against the prior swing, volume drying up on the way down and returning on the bounce, and room to the next level. The support and resistance guide covers the levels, and the EMA day trading strategy post covers the faster averages that traders commonly use for the same job on intraday charts. The bands add one thing those lines do not: a width that scales with how fast the move has been.

Why a Band Touch Is Not a Signal

StockCharts puts it in one line: moves that touch or exceed the bands are not signals, they are tags. It also relays Bollinger's figure that the bands should contain 88 to 89 percent of price action, which is why a close outside them stands out. Standing out is not the same as reversing. In the worked series, price was outside the upper band for eight straight bars and the thing that followed was more of the same.

The tempting shortcut is the band fade: buy when price tags the lower band, sell when it tags the upper band, on the theory that price returns to the middle. That bets against the move, and in a trend the move keeps going. It is the low-grade setup, and this post does not recommend it as an entry. A tag of the lower band in a downtrend is price continuing down, the mirror image of the diagram. Entries in the direction of the trend, taken on a pullback, are the better-graded version of the same chart. The band only tells you what the market looks like. It does not give you the reason to enter.

Bands are one of several envelope-style indicators, and they are not interchangeable. The Keltner channel guide covers the version built from average true range, and the Donchian channel guide covers the one built from the highest high and lowest low with no volatility maths at all. Momentum oscillators such as RSI answer a different question, and the RSI trading strategy post covers the oscillator side. For where bands sit among the other tools, see the roundup of indicators for day trading.

The 20/2 Default and When to Change It

The default is 20 periods and 2 standard deviations, and StockCharts lists that as the standard configuration. Twenty is a window of roughly a month of daily bars. On an intraday chart the same 20 means a much shorter stretch of time, so the first decision is the window you want the bands to describe, not the number.

If you do change the period, StockCharts relays Bollinger's advice to adjust the multiplier in small increments, for example 2.1 for a 50-period average and 1.9 for a 10-period one.

So what are the best Bollinger Bands settings for day trading? No source used here shows that one set beats another, and this post does not claim it. A faster setting reacts sooner and touches the bands more often, which means more tags that mean nothing. A slower setting is calmer and later. Pick the window that matches how long you hold, keep the multiplier near 2, and judge the chart by structure and volume rather than by whether a band got touched.

Bollinger Bands on a Screenshot

The precise version, since this is easy to oversell. SnapPChart reads one chart screenshot you upload. Bollinger Bands are not among the fields it extracts, and it does not compute them. Its instructions name EMAs, VWAP, MACD and volume as the indicators to read. If you plot the bands before you take the screenshot, they are in the image, but nothing in the grader's instructions tells it to read them, and it never calculates a band or signals off a tag.

It also grades momentum continuations only: a long is a pullback in an established uptrend, a short is a rally in an established downtrend, and it does not take reversal or counter-trend setups. That is why a band fade is not something it grades as a trade. The with-trend read in this post, a pullback in a trend that has been walking the band, is the shape of chart it is built for. A neutral description of what a single screenshot read covers is on the AI chart analysis page.