Bollinger Bands Settings: Is 20, 2 Right for 5-Min Charts?

You open a 5-minute chart, add Bollinger Bands, and price is poking out of the bands every few candles. So you start fiddling with the numbers until the picture looks tidy, and call that a setting. The 20 and 2 are defaults, not laws. What

You open a 5-minute chart, add Bollinger Bands, and price is poking out of the bands every few candles. So you start fiddling with the numbers until the picture looks tidy, and call that a setting.

The 20 and 2 are defaults, not laws. What matters is that length and width move together, and that a band touch tells you price is stretched, not that it will turn.

The 20 and 2 are defaults and the inventor says so

Bollinger Bands are a simple moving average with a line above and below it, each placed a set number of standard deviations away. Standard deviation (a measure of how spread out recent prices are) is what makes the bands widen when the market gets wild and shrink when it goes quiet.

TradingView's Bollinger Bands support page lists 20 as the default length and 2 as the default standard deviation multiplier. The bands are the 20-bar simple moving average plus and minus two standard deviations.

The person who invented them is blunter. In Bollinger's own rules, rule 9 calls the 20 periods and two deviations "just that, defaults," and adds that the parameters needed for any given market or task "may be different." So the question is not whether 20 and 2 are correct. It is what changes when you leave them.

Shorter averages need narrower bands, but by how much

Rule 11 of the same page is the one most settings guides skip. For consistent price containment, if the average is lengthened the number of standard deviations needs to go up, from 2 at 20 periods to 2.1 at 50. If it is shortened, the deviations should come down, from 2 at 20 periods to 1.9 at 10.

Look at how small those adjustments are. Going from 20 to 10 periods only takes you from 2 to 1.9.

Scalping guides online often pair a 10-period average with 1.5 deviations. That is not Bollinger's number. It is a much tighter band than his containment rule suggests, so it makes a different promise: more touches, more often.

A simulation shows what each setting does to touch frequency

I did not want to argue this from vibes, so I ran the numbers. This is a simulation, not market data, and it only isolates one question: how often does a close land outside the bands for a given length and width?

The setup: 1,000,000 simulated bars per run, price as the running sum of random returns, a simple moving average, and population standard deviation. I ran three versions of the returns: normally distributed, fat-tailed (Student's t with 3 degrees of freedom), and one with volatility that drifts up and down over time. The generator is seeded, so the run is repeatable, and the share of closes inside the bands came out like this.

Length / deviations	Share of closes inside the bands
10 / 1.5	about 72%
10 / 1.9	about 88%
20 / 2.0	about 87% to 88%
20 / 2.5	about 96% to 97%
50 / 2.1	about 87% to 89%
50 / 2.5	about 94% to 96%

Two things jump out. First, Bollinger's containment pairs (10 with 1.9, 20 with 2.0, 50 with 2.1) all land within a point or two of each other. His rule 11 holds up on a pure random walk, which is a decent sign the idea is sound.

Second, 10 with 1.5 is a different animal. Roughly 28% of closes end up outside the bands, against about 12% at the default.

Put that on a clock. A regular US stock session is 6.5 hours, which is 78 five-minute bars. At 20 and 2 you would expect around 9 closes outside the bands per session in this kind of random path. At 10 and 1.5, around 22. If every one of those looks like a "signal," you have a signal machine that fires constantly, which is not the same as a machine that finds good trades.

The 95% figure is wrong even on perfect random data

You will see claims that 95% of prices sit inside two standard deviations. That is the textbook rule for a bell curve, and price does not behave like one.

Bollinger addresses this in rule 14. He says to make no statistical assumptions from the standard deviation calculation, because price distributions are non-normal and the typical sample is too small for statistical significance. In practice, he writes, about 90%, not 95%, of the data falls inside the bands at default settings.

Our simulation agrees on the direction even with a perfectly bell-shaped return distribution: about 87% to 88% inside, not 95%. I have not isolated the exact cause. One plausible reason is that the bands are built from the same 20 prices they are judging, so price levels drift away from their own recent average more than independent draws would. Either way, a band touch happens more often than "two sigma" sounds like it should.

A band touch is a tag and not a signal

This is the part that wrecks accounts. Rule 6 on the same page says: "Tags of the bands are just that, tags not signals." A tag of the upper band is not in itself a sell signal, and a tag of the lower band is not in itself a buy signal.

Think of the band as a speed camera reading, not a prediction. It tells you price is moving fast relative to its recent range. In a strong trend, price can "walk the band" for many bars while every touch looks like a reason to fade it.

Rule 22 on that page puts it well: the bands do not provide continuous advice, they help identify setups where the odds may be in your favour. That framing is why most serious uses add a second piece of information, such as regime or structure, and why the plain "touch the band, fade it" idea needs a filter. We went through the regime question in our post on mean reversion on forex and indices.

Changing your Bollinger settings until the chart looks clean is curve fitting with extra steps. The bands will always look clean on the stretch you tuned them on.

— The Tradoki desk note

Twenty bars is not a lot of memory on a 5-minute chart

On a daily chart, 20 bars is about a month of trading. On a 5-minute chart it is a hundred minutes. The bands are then describing how volatile the last hour and a half was, nothing more.

That has a practical consequence for the squeeze (when the bands pinch tight after a quiet stretch). On a low timeframe, a squeeze can be a lunch-hour lull that says nothing about the next session. The indicator is doing its job. The question is whether a hundred minutes of range is the context you want your rule built on.

It also means width interacts with costs. If the band is only a few ticks wide, spread and commission eat a large share of any move you are trying to capture. Check your instrument's real spread against your typical band width before deciding a tight setting is tradeable. Our post on indicator values that differ between platforms is a good reminder that the same defaults can also be calculated slightly differently from one platform to the next.

How to pick settings without fooling yourself

Start from the defaults and change one thing at a time. If you shorten the length, bring the deviations down in step (Bollinger's 1.9 at 10 periods is the reference point), then see what changes in how often price closes outside.

Decide the rule before you look at results. Write down the length, the multiplier, the average type, and what you do on a touch. Then test it on enough trades that the answer means something, because how many trades a backtest needs is usually more than people expect.

Also check what the platform's tester is allowing. A fill at the band price on the exact candle that touched it is the kind of thing that flatters a result, as we showed when looking at why the TradingView strategy tester looks unrealistic.

Finally, run it on a demo account for a few weeks. Real-time tags feel different from historical ones.

● FAQ

What are the default Bollinger Bands settings?

20 periods and 2 standard deviations, around a simple moving average. TradingView's support page lists 20 and 2 as the defaults. John Bollinger's own rules page calls them defaults and says the right parameters for a given market or task may be different.

What Bollinger Bands settings work best on a 5-minute chart?

There is no tested best setting, and anyone who names one is guessing. Bollinger's guidance is to shorten the average and reduce the deviations together, for example 1.9 at 10 periods. A popular scalping variant of 10 and 1.5 is narrower than that guidance, which means price closes outside the bands far more often.

Do 95% of prices fall inside the Bollinger Bands?

No, and Bollinger says so himself. He writes that in practice about 90%, not 95%, of the data sits inside the bands at the default settings. In our simulation of random price paths, it was about 87 to 88%.

Is touching the lower Bollinger Band a buy signal?

No. Bollinger's rules say a tag of the lower band is not in itself a buy signal, and a tag of the upper band is not in itself a sell signal. A touch tells you price is stretched relative to its recent range, not that it will reverse.

Should I use an exponential average in Bollinger Bands?

Traditional Bollinger Bands use a simple moving average, because the standard deviation calculation also uses a simple average. An exponential version is a variant, and Bollinger says the exponential average must then be used in both the middle band and the deviation calculation.