Bollinger Bands vs ATR
Bollinger Bands draw an envelope from the dispersion of closing prices, ignoring wicks and gaps. Average true range measures the typical size of a whole bar including the gap from the previous close, which is what a stop distance actually has to survive.
Both are described as volatility tools and they measure different things. One reads closing prices; the other reads the whole bar. That decides which one is useful for what.
What each one is
Bollinger Bands draw an envelope from the dispersion of closes, a chosen number of standard deviations either side of a moving average. Bollinger Bands covers it.
Average true range measures the typical size of a bar, including the gap from the previous close. Average true range covers it.
One sees only where bars finished. The other sees everything they did, which is a genuinely different view of the same market.
Where they differ
What each can see. A day that swung widely and closed flat is a large range reading and almost nothing to a dispersion measure.
How gaps are handled. Average true range counts the jump from the previous close explicitly. A close-based statistic treats it as an ordinary change in level.
How outliers behave. Dispersion is defined by extremes, so a single violent close moves it a great deal. Range is an average of bar sizes and absorbs one event more gently.
What each is for. Judging stretch and reading squeezes on one side; stop distances and position sizes on the other.
Where they agree
Both measure movement rather than direction. Neither has an opinion about which way price is going, which is why both belong as inputs rather than as signals.
Both are backward-looking. Every value comes from bars that have already printed and neither anticipates a change in conditions.
Both expand after a shock rather than before it. The move that would have hurt you is the one that widens the reading.
And both need a length. The lookback decides responsiveness in each case, and neither convention was chosen for your instrument.
Which one to use
Use average true range for stops and position sizes. A stop has to survive the full bar including any gap, and that is precisely what the measure reports.
Use Bollinger Bands when width itself is the reading. The squeeze is a genuine feature of the statistical version and it does not exist in the same form on a range measure.
Use average true range on anything that gaps. Instruments that open away from the previous close leave a close-based measure blind to a large share of their movement.
And when both are on the chart, decide which one sizes the trade. Using one for the picture and the other for the arithmetic is coherent; using both for both is not.
Why the bar matters more than the close
Because a stop is touched intrabar. Price does not wait for the close to reach your level, so a measure that only sees closes is measuring the wrong thing for that job.
And because wide bars that close flat are common. On this site’s shared series the ninetieth percentile bar range is 1.101 against a median of 0.493 and the largest was 2.338.
What the measured figures are here
Average true range over fourteen bars has a median of 0.5994, with a ninetieth percentile of 0.7954.
Bar ranges run wider than the average suggests. Median 0.493, ninetieth percentile 1.101, largest 2.338.
Trailing stops sized from range survive predictably. At 1, 2, 3 and 4 average ranges the median survival was 3, 10, 22 and 32 bars across 562 trials.
And that spread is the decision. A wider stop lasts longer and costs more when it goes, which is a choice you can make from figures rather than from feel.
What to check before using either
The lookback length. It is the main parameter on both and it decides how quickly each responds.
The band multiplier. It sets how often a touch happens, which is the entire content of that reading.
Whether your instrument gaps. If it does, a close-based measure is missing a meaningful part of the movement.
And which measure sizes the trade. Only one number can set the stop, so decide in advance which it is.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two
directly — this pair is constructed from two subjects the corpus covers separately. Separately,
Bollinger Bands appear in 259 titles at a median of 5,178 across 197 channels, and average true range in
307 at a median of 9,432 across 220. The counts come from site/corpus_count.py.
259 videos on one at 5,178 and 307 on the other at 9,432. Similar coverage and nearly double the audience per video for the range measure — unusually even, and the tool used for sizing draws more interest than the one used for looking.
The answer to the question on that chart is that narrow bands describe closes. Bars can still be wide while closes cluster — so the stop should be sized from the range measure, not from the envelope.
When it fails
The failure is sizing a stop from a close-based envelope on an instrument with wide bars. The bands are narrow because closes have clustered, so the stop is placed close to price. Bars continue to swing — on this site’s shared series the ninetieth percentile bar range is 1.101 against a median of 0.493 — and the stop is touched intrabar repeatedly. The envelope was accurate about closes and silent about the part that mattered.
The second failure is treating a band touch as extreme. It is a stated distance.
A third is comparing raw values across instruments. They are in price units.
A fourth is reading the squeeze as directional. It says nothing about which way.
A fifth is leaving both lookbacks at defaults. They came from elsewhere.
And a sixth is using both to size one trade. Only one number can set the stop.
Related
Bollinger Bands covers the close-based envelope. Average true range covers the whole-bar measure. And standard deviation covers the statistic the bands are built from.
A day that swung two percent and closed flat barely registers on a close-based measure and is a large reading on a range-based one. If you are placing a stop, that day is the whole problem — so the measure that saw it is the one to use.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.