Bollinger Bands Explained: What the Inventor's Rules Say

Bollinger Bands are three lines drawn on a price chart: a moving average in the middle, and an upper and a lower band set a number of standard deviations away from it. The bands widen when volatility rises and narrow when it falls.

John Bollinger, who invented them, publishes 22 rules for using them on bollingerbands.com, and those rules say a tag of a band is not in itself a signal and a close outside the bands is a continuation signal rather than a reversal one. Plenty of guides teach the opposite, so settle that disagreement before you test anything.

What are Bollinger Bands?

The middle line is an ordinary moving average. What separates the outer two from any other channel is the standard deviation, which measures how far price has been straying from that average. A quiet market gives a small deviation and tight bands, and a violent one gives a large deviation and wide ones.

Rule 1 of the 22 says what the bands are for: "Bollinger Bands provide a relative definition of high and low. By definition price is high at the upper band and low at the lower band."

The page is signed "John Bollinger, CFA, CMT", and the rules are introduced as "assembled in response to questions often asked by users and our experience over 30 years of using the bands".

Bollinger Bands explained: the formula

The middle band is a simple moving average of closing prices, 20 periods by default. The upper and lower bands are that same average plus and minus a multiple of the standard deviation of those same periods, 2 by default.

TradingStrategyGuides, dated June 18, 2025, writes it out: "The middle band consists of a 20-period moving average. The upper and lower are two standard deviations below and above the moving average in the middle." TrendSpider's FAQ adds what platforms actually ship: "Most charting platforms use a 20-period simple moving average with bands set two standard deviations above and below it." LiteFinance, updated 18Sep.2026, lists the same three components, notes that the classic version calculates from closing prices, and adds: "The bands are recalculated whenever a new candlestick forms, allowing them to adjust dynamically to changes in price and volatility."

Rule 12 gives the reason for a simple average rather than an exponential one: "Traditional Bollinger Bands are based upon a simple moving average. This is because a simple average is used in the standard deviation calculation and we wish to be logically consistent."

Bollinger Bands is one of the indicators in the ChartLabs Indicators menu, under Standard, drawn on price, so you do not need to build it to test it.

Does a touch of the band mean sell?

Not by itself. Two of Bollinger's rules cut against that reading, and the one between them says why.

Rule 6: "Tags of the bands are just that, tags not signals. A tag of the upper Bollinger Band is NOT in-and-of-itself a sell signal. A tag of the lower Bollinger Band is NOT in-and-of-itself a buy signal."

Rule 7: "In trending markets price can, and does, walk up the upper Bollinger Band and down the lower Bollinger Band."

Rule 8: "Closes outside the Bollinger Bands are initially continuation signals, not reversal signals. (This has been the basis for many successful volatility breakout systems.)"

Guides teach it the other way round. TradingStrategyGuides: "They are mainly used when determining when there are overbought or oversold levels. Selling when the price touches the upper band and buying when the price touches the lower band."

TrendSpider's learning centre lists "Upper Band Touch = Potential Overbought" and "Lower Band Touch = Potential Oversold", and in a combinations list "Upper Band + RSI Overbought → Stronger sell signal". Its own limitations section then says "Overbought/oversold readings can persist during strong trends", which is rule 7 in different words.

LiteFinance, a broker's education blog, states the caution outright: "the price touching either band does not signal a reversal on its own", and the first entry on its list of common mistakes is "Treating a touch of the upper band as an unconditional sell signal, or a touch of the lower band as a buy signal, particularly during a strong trend."

The caution comes from the indicator's inventor, so start there and treat the overbought reading as something to prove on your own market and timeframe.

Do Bollinger Bands contain 95% of price?

Bollinger says do not assume that, and gives a different number. Rule 14: "Make no statistical assumptions based on the use of the standard deviation calculation in the construction of the bands. The distribution of security prices is non-normal and the typical sample size in most deployments of Bollinger Bands is too small for statistical significance. (In practice we typically find 90%, not 95%, of the data inside Bollinger Bands with the default parameters)"

The 95% figure comes from the normal distribution, not from price. LiteFinance says both halves: "Assuming a normal distribution, a range of two standard deviations above and below the mean theoretically encompasses about 95% of all values. However, actual price movements over a given period may not follow a normal distribution." Its FAQ calls the figure "a simplification based on the assumption of a normal distribution".

TradersAgency, dated June 9, 2026, prints it with no qualifier: "Statistically, assuming a normal distribution, the bands contain roughly 95% of recent price closes."

Rule 14 is the figure to trust, because 90% is what the inventor reports finding on real charts rather than what the normal distribution predicts in theory. Band containment describes the chart in front of you and tells you nothing about the odds on the next bar.

What settings should I use?

Rule 9 says the defaults are only a starting point: "The default parameters of 20 periods for the moving average and standard deviation calculations, and two standard deviations for the width of the bands are just that, defaults. The actual parameters needed for any given market/task may be different."

Rule 11 adds that the period and the deviation move together: "For consistent price containment: If the average is lengthened the number of standard deviations needs to be increased; from 2 at 20 periods, to 2.1 at 50 periods. Likewise, if the average is shortened the number of standard deviations should be reduced; from 2 at 20 periods, to 1.9 at 10 periods."

TradingStrategyGuides tells its readers that John Bollinger recommends 9 to 12 periods at 2 standard deviations. His rules page carries no such recommendation. Its only period and deviation pairing is rule 11's, and rule 11 moves the deviation down to 1.9 when the average is shortened to 10 periods, rather than leaving it at 2. The 9-to-12 suggestion is TradingStrategyGuides' own.

The published suggestions spread wide, and they are their authors' own.

• TradingStrategyGuides' FAQ gives a per-timeframe table that pulls against rule 11 in the other direction: a 20-period average with 0.5 or 0.75 deviations on a 1-minute chart, 1 or 1.5 on a 5-minute chart, and 2.5 or 3 on a 15-minute chart.
• TradingWithRayner, "Last Updated: July 31, 2023", goes wider still: "You can adjust your Bollinger Bands settings to 3 standard deviations (or higher) to identify even more overbought/oversold levels to trade-off."
• LiteFinance suggests reducing the average to 14 or 10 periods for shorter time frames and increasing it to 50 for longer-term trading.
• Audacity Capital, a prop firm, updated Aug 3, 2026, gives no number: "Treat the default 20-period, two-deviation values as a testing baseline rather than a universal best."

Rule 9 and Audacity Capital agree on that much: the setting is one you measure on the market you are actually trading.

What is a Bollinger Band squeeze?

A squeeze is the bands drawing close together, which is BandWidth getting small. BandWidth is Bollinger's own measure of how far apart the bands are.

Rule 18: "BandWidth tells us how wide the Bollinger Bands are. The raw width is normalized using the middle band. Using the default parameters BandWidth is four times the coefficient of variation."

Rule 19: "BandWidth has many uses. Its most popular use is to identify "The Squeeze", but is also useful in identifying trend changes..."

What a squeeze means next is unsettled. TradersAgency gives a fixed trigger: "When the bandwidth drops to a six-month low, a major breakout is usually imminent." Its worked setups use round numbers on a stock the page itself labels "a hypothetical example with XYZ stock", so they illustrate the arithmetic rather than report a market.

Audacity Capital rejects fixed numbers for the same measure: "There is no fixed BandWidth number that works across every market, so define "unusual" for the chart you are trading."

Do Bollinger Bands work?

The inventor's own answer is rule 22: "Bollinger Bands do not provide continuous advice; rather they help identify setups where the odds may be in your favor."

The costs are named by the people teaching the indicator. TrendSpider's limitations section: "They react to price-they do not predict it", and "In sideways markets, the signals may be frequent and unreliable." LiteFinance's mistakes list includes "Expecting the price to always bounce off the outer bands. In a trending market, price may "walk" along one of the bands for an extended period."

You will also find accuracy percentages and win rates attached to band strategies. We publish none, for Bollinger Bands or anything else. A win rate detached from the market, the period, the settings, the costs and the exit rules is not a number you can use, and a number from someone else's test is not evidence about yours.

How to test a Bollinger Bands rule on past data

Write one rule down before you look at a chart, then step through history bar by bar so the next bar is not visible while you decide.

1. Pick one market and one timeframe, and keep them fixed for the whole test.
2. Write the entry trigger in words that leave nothing to judgement. "A close outside the upper band" and "a tag of the upper band" are two different tests. Rules 6 and 8 are the reason they are different.
3. Write the invalidation point and the exit before the first trade, not after a losing one.
4. Record the settings beside the result. The same rule at 20 periods and 2 deviations is a different test from the same rule at 10 and 1.9, and rule 11 says so.
5. Wait for the bar to close before you act on a band value. Audacity Capital puts the reason plainly: "Completed historical bands do not repaint in the usual sense. Live values do continue to update as long as the current bar is unfinished, so a reading you see mid-candle can shift by the close. If your rules depend on a specific band value, wait for the bar to complete before acting."
6. Include the stretches you would rather skip: the flat months, the gap-ridden weeks, the run where your rule kept getting stopped out.

Audacity Capital publishes a checklist that reads like a test specification. Under its "Risk Management and Backtesting" heading it asks you to define, in observable terms: "Instrument and timeframe", "Market condition filter that qualifies a chart for the setup", "Entry, stop, exit, and re-entry rules" and "Realistic transaction costs, including spread and slippage". Then: "Test the rules across favorable and unfavorable regimes, not only the periods where the setup felt easy", and "Reserve an out-of-sample window or a forward test before you rely on the results."

Stepping bar by bar is the part software has to do for you. In ChartLabs, replay runs at 3x and 7x as well as step by step, on daily bars as well as shorter ones. Each moving average you add gets its own colour and line style, and editing one instance leaves the others alone, which matters when a test has a 20-period average on the chart next to a 10-period one. BandWidth is built in as well: Bollinger BandWidth is in the Indicators menu under Standard and draws in its own pane, at Length 20 and StdDev 2 to match the bands, with lines for its highest and lowest reading over the last 125 bars.

For the general method, start with our beginner's guide: how to backtest a trading strategy. For why the fill price matters more than the signal, see backtesting accuracy, slippage and data quality. If you are weighing this against running the same rule through code, manual versus automated backtesting sets out what each one actually tests.

What to record

Per trade: the date, the direction, which band state you traded (a tag, a close outside, a squeeze breakout), the settings in use, your entry, your stop, your exit and the result in R. Then the count, because a handful of trades will not tell you much.

Audacity Capital asks for the same shape of evidence from its own readers: "Report trade count, average win and loss, expectancy, and maximum drawdown rather than curating screenshots."

Record the settings on every line. Shorten the average and the deviation should move with it, so two logs with the same entry wording can turn out to be two different tests.

Next step

Pick one band rule, write it down, and step through a year of your own market bar by bar. Start a free trial and run it on cleaned historical data, sourced from Dukascopy at tick level where it is available and built into 1-minute candles.