Volatility indicators: ATR, Bollinger Bands, Keltner channels
Volatility measures how much price moves, never which way. This is the family True Range Research is built on, so it gets the most detail.
Volatility indicators quantify movement without regard to direction. A market falling hard and a market rising hard can produce identical readings, and that is correct behaviour rather than a shortcoming.
True range
Ordinary high-minus-low undercounts whenever a bar opens away from the previous close. True range fixes this by taking the largest of three distances:
The second and third terms capture gaps. A stock that closes at 50 and opens at 47 has travelled three points that a naive high-minus-low calculation would miss entirely.
Average True Range
ATR is a smoothed average of true range, conventionally over 14 periods. It is expressed in the instrument's own units — points, ticks, dollars — not as a percentage.
That unit choice has a consequence worth internalising: ATR is not comparable across instruments. An ATR of 12 on one contract and 0.4 on a stock tells you nothing about which is more volatile in any meaningful sense. ATR is for comparing an instrument to its own recent history, and for very little else.
What ATR is actually answering. “Is this instrument moving more or less than it typically has recently?” A move of two points means something different when ATR is 1 than when ATR is 6. That normalisation — measuring a move against the instrument's own normal — is the useful part.
Bollinger Bands
A moving average with bands placed a number of standard deviations above and below:
Typical settings are n = 20 and k = 2. Because standard deviation is computed from closes, the bands react to the distribution of closing prices and are relatively insensitive to intrabar extremes.
Note what the bands do not mean. Price touching the upper band is not a sell condition. In a strong trend price can ride the upper band for many bars, and the band will widen to accommodate it.
Keltner channels
Structurally similar, but built on ATR instead of standard deviation:
Because ATR accounts for gaps and intrabar range while standard deviation of closes does not, Keltner channels tend to be steadier through gappy conditions and slower to expand on a single violent close.
Choosing between them
| Bollinger | Keltner | |
|---|---|---|
| Width driven by | Std deviation of closes | Average true range |
| Sees gaps | Only via the close | Yes, through true range |
| Reacts to one wild bar | Sharply | Moderately |
Compression and expansion
Volatility clusters. Quiet periods tend to follow quiet periods and violent ones follow violent ones, and this is among the more durable observations in market data.
What it does not give you is direction. A period of unusually narrow range says a larger move becomes more likely; it says nothing whatsoever about which way. Any tool or write-up that treats compression as directional has added an assumption the measurement does not contain.