ATR bands: formula, multiplier, breakout or fade, and an ES test

Contents

1. How are ATR bands calculated?
2. What ATR length and multiplier should you set?
3. Should you trade ATR bands as breakouts or as fades?
4. How do ATR bands differ from Keltner channels and Bollinger Bands?
5. What does an ES daily-bar test of an ATR rule show?
6. Where do ATR bands fail, and how do you check for it?
7. Frequently asked questions
8. References

How are ATR bands calculated?

ATR bands are a middle line (the close or a moving average) plus and minus a multiplier times the Average True Range, so they widen when the market moves more and narrow when it goes quiet. The band has no direction; the rule you attach and the costs you charge decide whether it earns anything in a test.

Average True Range (ATR) is the average, over a set number of bars, of each bar's true range. True range is the largest of three distances: the high minus the low, the high minus the previous close, and the previous close minus the low. The second and third distances matter on futures, because a gap between sessions is real movement that a plain high-minus-low measure misses. The definition and its history are on the Average True Range indicator page, so this article spends its words on the bands built from it and on how to test them.

ATR bands are three lines: a middle line, an upper band and a lower band. The one formula:

upper band = middle + m × ATR(n), lower band = middle − m × ATR(n)

Here middle is the close of the bar or a moving average of the close, m is the multiplier (how many ATRs away the band sits) and ATR(n) is the Average True Range over the last n bars. Because ATR changes on every bar, the band distance changes on every bar too. A fixed 50-point channel is 50 points wide in a calm market and in a wild one. An atr channel moves out when ranges grow and pulls in when ranges shrink. That is the whole reason to use it, and it is what separates it from the fixed-distance channels in the guide to technical analysis tools and indicators you can test.

One choice has a large effect. A middle line of the close gives bands that jump with every bar, so one large bar can move the band past the next bar's price. A middle line of a moving average moves more slowly, and it is the form most tested strategies use.

What ATR length and multiplier should you set?

The ATR length sets how fast the bands react and the multiplier sets how often price reaches them, so test a grid of both and keep a region of neighbouring settings that behave alike, not the single best pair.

The length is a memory. ATR(10) on daily bars reflects a change in volatility within about two trading weeks; ATR(50) takes about ten weeks. A short length follows volatility closely but the bands wobble from bar to bar. A long length gives steady bands that are late to widen after a shock, so the first bars of a volatile stretch close far outside the band and produce a cluster of signals at once.

The multiplier is a distance. At 1 ATR, price closes outside the bands often: entries come early, and many of them reverse. At 3 ATR, closes outside the bands are rare: entries come late, well after the turn, and fewer of them are false. Figure f2 shows the same price stretch at all three settings.

A parameter plateau is a region of neighbouring settings whose results are all acceptable and similar. Pardo (2008) argues that the middle of such a region is a safer choice than the single best cell, because a cell whose neighbours are poor describes noise in one stretch of history. The steps:

1. Choose a grid before you run anything. For example lengths 10, 14, 20 and 50, and multipliers from 1.0 to 3.0 in steps of 0.5. That is 20 pairs.
2. Run every pair with the same commission, slippage and session settings. A pair tested with different costs cannot be compared with the others.
3. Lay the results out as a table with lengths down the side and multipliers across the top, one number per cell (net profit after costs, or drawdown relative to profit).
4. Look for a block of neighbouring cells that are all positive and close in value. That block is the plateau.
5. Pick the middle of the block. If the only good cell has poor neighbours, the grid found luck, not a setting.

Should you trade ATR bands as breakouts or as fades?

Write both rule sets out, run each once with the same costs and sessions, and compare their time in market and their drawdown relative to profit; the chart alone cannot tell you which one holds up. A breakout rule is a rule that enters in the direction of a close outside the band, expecting the move to continue. A fade is a rule that enters against a touch of the band, expecting price to return to the middle line. The two rules read the same line and do the opposite thing.

	Breakout rule	Fade rule
Entry	Close above the upper band → long; close below the lower band → short	Touch of the upper band → short; touch of the lower band → long
Exit	Close back through the middle line, or a close through the opposite band	Return to the middle line
Works best	Trending stretches	Sideways stretches
Typical loss	A false break inside a range	A trend that runs along the band
Time in market	Often high: the position is held until the middle line is crossed	Often lower: the trade ends when price returns to the middle

Time in market is the share of bars during which the strategy holds a position. A breakout rule that exits at the middle line is often positioned most of the time, so part of its result is simply the market's drift over the test period. A fade rule is flat more often, so its result depends less on drift and more on the band touches themselves. Read this number before the net profit, because it tells you what the profit is made of. The second number to read is the maximum drawdown next to the net profit: a drawdown close to the profit means the rule gave back almost everything it made at some point. AlphaPrime (Valynox) charges the commission and slippage you set and fills on the next bar in both runs, so when you place the breakout and the fade rule side by side in backtesting, the only difference between the two results is the rule.

A useful comparison for the breakout side is a channel that does not use volatility at all. The Donchian channel breakout rules enter on a new highest high or lowest low over a lookback. Running the ATR breakout and the Donchian breakout on the same data shows whether the volatility scaling of the band adds anything, or whether any band that moves with price gives a similar trade list.

How do ATR bands differ from Keltner channels and Bollinger Bands?

The three channels differ in the middle line and in the width measure, and those two choices decide how often price reaches the bands. A Keltner channel is an exponential moving average, often 20 bars, with bands a multiplier times ATR above and below, often 1.5 to 2. Bollinger Bands are a simple moving average, often 20 bars, with bands a multiplier times the standard deviation of the close, often 2. Standard deviation is a measure of how far the closes in the lookback spread around their average. Plain ATR bands around the close have no smoothing in the middle at all, so they hug price more tightly than either.

	ATR bands	Keltner channel	Bollinger Bands
Middle line	Close or any moving average	Exponential moving average, often 20	Simple moving average, often 20
Width measure	ATR × multiplier	ATR × multiplier, often 1.5–2	Standard deviation × multiplier, often 2
How width reacts	Follows true range, including gaps	Follows true range, including gaps	Follows close-to-close dispersion
Touches in a slow drift	Rare when the middle is the close	Moderate	Frequent: closes bunch together and the bands tighten

The width measure is the practical difference. The Keltner channel is a specific ATR band: an exponential middle line and an ATR width. Any ATR bands strategy with that middle line is a Keltner strategy under another name. Bollinger Bands measure something else: a market that drifts slowly upward in small bars with large intraday ranges shows a narrow Bollinger width (the closes are close together) and a wide ATR width (the ranges are large). The two bands then disagree about whether the market is quiet. The middle line matters too: the HMA Bollinger Bands test shows how much the trade signals change when only the midline is swapped for a Hull moving average and the width measure stays the same.

ATR also appears as a divisor rather than a band. MACD-V divides MACD by ATR so that the oscillator reads the same across markets with different price levels. The idea is the same as in ATR bands: measure distance in units of recent movement instead of points.

What does an ES daily-bar test of an ATR rule show?

The run behind this article kept one candidate out of the 1,024 it tried on ES daily bars from 2019-01-02 to 2025-06-30, and its numbers describe that one selected rule over that period, not a forecast. AlphaPrime tested long and short trades with one contract, charged $20 commission and $25 slippage per trade, and used an Average True Range rule as the only entry and exit, with no stop, no target and no filter. This is a stand-in rule: an ATR entry and exit alone, not a complete ATR bands breakout or fade strategy with a middle line and two bands.

Read the table in the order from the previous section. Time in market is 82.9%, so the rule held a position on more than four bars out of five. Over a period in which the S&P 500 rose in most calendar years, a rule that is almost always positioned carries part of the index's own movement, and the net profit of $117,750 cannot be separated from that drift without a longer test or a flat benchmark. The max drawdown of $57,590 is a little under half of the net profit, which means the rule gave back about half of its result at one point along the way. The profit factor of 1.41 says that gross wins were 1.41 times gross losses, a margin that a change in costs or a different six years could narrow. With 165 trades over about six and a half years, roughly 25 a year, a handful of trades moves the $713.64 average trade.

The candidate was kept out of 1,024 because its return looked best against its drawdown. Choosing the best of 1,024 after seeing all of them flatters the one kept: Bailey, Borwein, López de Prado and Zhu (2014) show that the best backtest in a large search is expected to be inflated even when no rule has an edge. So the table is a reading of one selected rule on one market, not evidence that an ATR rule earned this money, and not an amount a reader can expect. It is a starting point for the checks in the next section.

Where do ATR bands fail, and how do you check for it?

ATR bands lag turns and tighten in quiet markets, which produces clusters of false breaks, so run four checks on your own strategy in a fixed order with pass conditions written before you look. The lag comes from the average: ATR(14) needs several bars of large ranges before the band widens, so the first closes of a volatile stretch land far outside it. The tightening comes from the same average: in a sideways stretch the ranges shrink, ATR falls, the bands pull in, and small moves that mean nothing close outside them. Each of those closes is a breakout entry that reverses a few bars later.

The four checks, in order, each with its pass condition written down first:

1. Time in market. Compare the reported share with the share you expected when you wrote the rule. A breakout rule you meant to be selective that sits in the market 80 percent of the time is riding drift, not breakouts.
2. Drawdown relative to profit. Set the margin in advance, for example a maximum drawdown no larger than half of net profit. Read the two numbers from the report and compare.
3. Trending and sideways stretches. Split the test period into stretches you would call trending and stretches you would call sideways, and require the rule to be positive in more than one kind of market. A breakout rule that made everything in two trends and lost in every range has not been tested in the market where it fails most.
4. Rerun with slippage doubled. Change only the slippage setting in backtesting, run again, and require the result to stay above zero. A rule that survives only at one slippage value depends on cheap fills. For the ES run above, a second $25 per trade is small next to the $713.64 average trade, so this check mostly tests the trade list; for an intraday rule with a $30 average trade it is the check that decides.

The next step is a small, honest grid. Write out a breakout rule set and a fade rule set for the same ATR bands, choose lengths 10, 14, 20 and 50 and multipliers from 1.0 to 3.0 in steps of 0.5, charge your real commission and at least one tick of slippage per side, and run all of it on one market and one bar size. Then read time in market and the drawdown next to net profit for every cell before you look at anything else, and keep only a setting that sits in the middle of a plateau.

Frequently asked questions

What ATR multiplier should I use for ATR bands?

There is no single correct value; the multiplier sets how often price reaches the bands. At 1 ATR closes outside the bands are frequent and many reverse; at 3 ATR they are rare and come late. Test a grid from 1.0 to 3.0 in steps of 0.5 together with several ATR lengths, and choose a value from the middle of a region where neighbouring settings all behave alike.

Are ATR bands better for breakouts or for reversals?

The band itself has no direction, so the answer depends on the market and the period. A breakout rule enters in the direction of a close outside the band and does well in trends; a fade rule enters against a touch of the band and does well in ranges. Write both rule sets, run each once with the same costs and sessions, and compare their time in market and their drawdown relative to profit.

What is the difference between ATR bands and a Keltner channel?

A Keltner channel is one specific form of ATR bands: an exponential moving average in the middle, often 20 bars, with bands a multiplier times ATR above and below, often 1.5 to 2. Plain ATR bands can use the close or any moving average as the middle line. Bollinger Bands are different in kind, because their width follows the standard deviation of the close rather than the true range.

Can ATR bands be used as a trailing stop?

Yes. A lower band placed a multiplier times ATR below the close or a moving average is a trailing stop that gives a position more room when volatility is high and less when it is low. The multiplier decides how often the stop is hit; test it on your own trade list with costs included, and check that the stop distance is many times larger than your slippage per trade.

References

• Wilder, J. W. (1978). New Concepts in Technical Trading Systems. Trend Research.
• Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). John Wiley & Sons.
• Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.