ATR vs Standard Deviation
Michael Whitman
·
Trading since 2017, full-time since 2020
·
Published 2026-09-08
·
Updated
·
Reviewed
Average true range measures the typical size of a bar, including any gap from the previous close, so it captures everything price did within that bar. Standard deviation of closes instead measures how spread out recent closing prices have been, which ignores wicks entirely.
Two measures of how much a market moves. One looks at the whole bar and the jump from the previous close; the other looks only at where bars closed. On thirty-two years of SPY daily bars they came out about the same size and still disagreed often about when volatility was rising.
What each one is
Average true range measures the typical size of a bar. J. Welles Wilder set it out in his 1978 book New Concepts in Technical Trading Systems: the true range is the largest of the high minus the low, the high minus the previous close and the previous close minus the low, and the average is smoothed over 14 bars, each new value being 13 parts the old average and one part the new range. Average true range covers it.
Standard deviation measures how spread out closes have been. StockCharts’ version takes the last N closes, their mean, and the square root of the average squared distance from that mean, dividing by N (the population form); its default N is 10. Standard deviation covers it.
One includes gaps and wicks; the other does not. That single design choice decides which questions each can answer.
Where they differ
What each can see. A day that swung widely and closed flat is a large reading on one and almost nothing on the other, because only one of them looks inside the bar.
How gaps are handled. Average true range counts the jump from the previous close directly. A standard deviation of closes sees a gap only as far as it survives into the close.
What a trend does to each. Closes that climb in even steps are spread across the whole 14-close window, so the standard deviation stays well above zero in a steady trend even when every bar is small. Average true range does not care about direction at all, only bar size.
How a shock moves each. One violent close changes a 14-close standard deviation sharply, then drops out all at once 14 sessions later. Wilder’s average takes a fourteenth of a big bar and lets it fade.
Where range and dispersion agree
Both measure movement rather than direction. Neither has an opinion about which way price is going, which is why both are used as inputs rather than as signals.
Both are backward-looking. Each reading is built from finished bars, so neither can see a change in conditions coming.
Both need a length. The lookback decides how quickly the reading responds, and a value chosen for one instrument does not transfer to another.
And they are about the same size on a broad index. On SPY the two 14-session readings had similar medians, so the choice between them is about behavior, not scale.
A wide day that closed flat
2 March 2026 shows the split in one bar. SPY had closed at $685.99 the session before. On the day it traded down to $678.02 and up to $688.62, a true range of $10.60 against a prior average true range of $8.01, and then closed at $686.38, up 39 cents. Average true range rose to $8.20. The 14-close standard deviation fell, from $4.26 to $3.84, because the close landed right where recent closes had been.
Both readings were right about what they measure. The bar was wide and the close was ordinary. A stop placed from the second number would have been sized for a market that, on that day, had just traded through more than $10 of range.
Which one to use: ATR or Standard Deviation?
Use average true range for stops and sizing. A stop has to survive the full bar including the gap, which is what this measure reports.
Use standard deviation when the question is statistical. Comparing dispersion, building an envelope or filtering on how unusual a close is are all jobs it was designed for.
Prefer average true range for anything prone to gaps. Instruments that open away from the previous close leave a close-based measure partly blind to their movement.
And when you are not sure which, use average true range. The questions most traders are asking, where to put a stop and how large a position should be, are range questions.
Checks before trusting either volatility reading
The lookback length. It decides how quickly the reading responds, and StockCharts’ default for the standard deviation is 10, not 14. On SPY a 10-close version put one deviation even closer to price.
Whether your instrument gaps. If it does, a close-based measure is missing a meaningful share of the movement.
Whether you need comparability. Raw values are in price units, so dividing by price is what makes readings transferable between markets.
And what the number is for. A stop, a size, a filter or an envelope are four different jobs, and only some of them suit each measure.
The original data
The sample. SPY daily bars from Yahoo Finance, highs, lows and closes as traded, 29 January 1993 to 2 October 2026; measurement starts on 3 January 1994, giving 8,243 sessions. Opens are not used. Average true range: Wilder’s 14-session version as above. Standard deviation: population form of the last 14 closes, set to 14 so both cover the same bars; the 10-close default is reported as a check.
What each sees. On 2,647 sessions (32.11%) the previous close sat outside the day’s own high-low range, so the true range was bigger than the bar. Summed over every session, 9.37% of all true range came from that gap part. Wide days that closed flat were not common: 158 sessions (1.92%) had a true range above the prior average and a close within 0.10% of the previous one.
Size. As a share of the close, average true range had a median of 1.20% and the 14-close standard deviation 1.15%. The ratio of the second to the first had a median of 0.90, a tenth of sessions below 0.58 and a tenth above 1.43; the standard deviation was the larger of the two on 3,224 sessions (39.11%).
Timing. Measured as a share of price, the two readings had a correlation of 0.82, but their day-to-day percentage changes only 0.28. Of the 1,648 sessions in the top fifth by average true range, 1,063 (64.50%) were also in the top fifth by standard deviation.
How often the next session reached each distance. Starting from each close, the table counts how often the next session’s low fell below, or its high rose above, the close minus or plus a multiple of each measure. 8,242 sessions had a next session; the next low fell below one and two average true ranges on 1,287 and 162 of them, and below one and two standard deviations on 1,632 and 383.
| Distance from the close | Next low below | Next high above |
|---|---|---|
| 1 × ATR (14) | 15.62% | 13.63% |
| 2 × ATR (14) | 1.97% | 0.63% |
| 1 × SD (14 closes) | 19.80% | 19.01% |
| 2 × SD (14 closes) | 4.65% | 3.15% |
The standard deviation was the nearer distance on about six sessions in ten (it was the larger on only 39.11%), which is why it was reached more often. With the 10-close default, the next session’s low fell below one standard deviation 24.97% of the time.
The tail. The true range topped twice the previous day’s average true range on 319 sessions (3.87%) and three times on 45 (0.55%). The largest was 6 May 2010, a true range of $12.00 against an average of $1.65, 7.25 times. Every session’s true range and both measures are in the SPY ATR and standard deviation file.
When it fails
The failure is sizing a stop from a close-based measure on an instrument that gaps. The dispersion reading is low because closes have been steady, so the stop is placed close to price. The instrument then opens beyond it, the fill happens wherever liquidity allows, and the loss is larger than the stop suggested. On SPY one standard deviation was reached by the next session’s low on about one session in five.
The second failure is setting raw readings from two different markets side by side. They are in price units.
A third is treating either as a direction signal. Both measure movement only.
A fourth is changing the lookback after a losing run. That is fitting.
A fifth is reading a low value as a sign of safety. It means recent bars were small, and on SPY single bars reached seven times the prior average.
And a sixth is using a fixed distance instead of either. The instrument decides the scale.
Related
Average true range explains the bar-size measure in full. Standard deviation covers the close-based one. And the ATR trailing stop covers the most common use of the first.
If you are sizing a stop, you care about how far price can travel within a bar, gap included. That is what average true range measures and what a close-based statistic cannot see directly. Pick the measure by the question, and size the stop so one ordinary bar does not reach it.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.