Volatility indicators

Keltner Channel: Formula, Settings, and Whether a Break Means Trend or Reversal

How Keltner channels set an ATR-wide frame around a 20-day EMA, how often prices close outside them compared with Bollinger Bands, and what closes outside the channels did next on Bitcoin, the S&P 500 ETF, gold and EUR/USD.

· Founder of PatternsWizard
Checked 10 min read
Keltner Channels(20, 10, 2) on Bitcoin, daily, August 30, 2026 to October 28, 2026. Illustration of how the indicator reads, not a backtest.
On this page
  1. How Keltner channels are calculated
  2. How to read the channels
  3. How often prices close outside the channels
  4. The channels on a real chart
  5. Trend or reversal: the full history
  6. Does it work?
  7. Frequently asked questions

Keltner channels wrap a 20-day exponential moving average in two lines set two average true ranges above and below it. The average gives the direction, the ATR the width. They look like Bollinger Bands, but because the width comes from the size of daily ranges rather than from the spread of closes, the lines are smoother and a close outside them means something slightly different. Below: the formula and its history, how often prices close outside compared with Bollinger Bands, a recent Bitcoin rally that rode the upper channel for nine days, and what both readings of a break, trend and reversal, did over the full history of four markets.

Keltner channels(20, 10, 2) on daily Bitcoin

Keltner Channels(20, 10, 2) on Bitcoin, daily, July 1, 2026 to October 28, 2026. Illustration of how the indicator reads, not a backtest.
  • Price, daily candles (hollow up, filled down)
  • Middle, EMA(20) of the close (solid line)
  • Channels, ±2 × ATR(10) (dashed line)
  • Shaded area between the bands
Keltner Channels(20, 10, 2) on Bitcoin, daily, July 1, 2026 to October 28, 2026. Illustration of how the indicator reads, not a backtest.

How Keltner channels are calculated

The version on most charts today, and the one StockCharts uses, has a 20-day exponential moving average of the close as the middle line, with channels two 10-day ATRs above and below. StockCharts traces the idea to Chester Keltner's 1960 book How to Make Money in Commodities, whose "Ten-Day Moving Average Trading Rule" used a 10-day simple average of the typical price, (high + low + close) / 3, with the 10-day average of the high minus low added and subtracted. Linda Bradford Raschke introduced the ATR-based version in the 1980s.

The average true range counts gaps and intraday ranges, so a market that gaps a lot gets wider channels even if its closes do not move much. StockCharts notes that Keltner channels are smoother than Bollinger Bands, because the standard deviation behind Bollinger Bands is more volatile than the ATR.

Bollinger Bands inside the Keltner Channel: the squeeze

Bollinger Bands inside the Keltner Channel: the squeeze: a dashed Keltner Channel of nearly constant width and Bollinger Bands around the same average. The Bollinger Bands narrow until both lines sit inside the channel, a shaded stretch marked squeeze on with red dots below it, then widen back outside the channel as the price breaks upward, marked fired with green dots.
Schematic, not market data.Because the bands react faster than the channels, the bands can narrow inside the channels in a quiet market: the squeeze in the table below.

How to read the channels

What you seeThe usual reading
Channels sloping up, price in the upper halfAn uptrend; pullbacks to the middle line are the classic entry
Close above the upper channelEither a breakout with strong momentum or a short-term extreme
Close below the lower channelEither a breakdown or an oversold extreme
Flat channels, price swinging between the linesA trading range: the lines act as overbought and oversold levels
Bollinger Bands inside the Keltner channelsThe squeeze used by the TTM Squeeze: unusually low volatility

StockCharts calls Keltner channels a trend following indicator used to identify reversals with channel breakouts and channel direction, and also shows them as overbought and oversold levels when the trend is flat. The two uses give opposite instructions at the same line, so the slope of the middle line is what decides which one applies.

Same close above the band, two readings

Same close above the band, two readings: left: a flat moving average with its upper and lower bands; price closes above the upper band once, then returns to the middle, so the overbought reading fits. Right: a rising average with its bands; price rides the upper band and keeps closing above it, which is the trend itself.
Schematic, not market data.The same close above the upper channel reads as an extreme when the middle line is flat, and as the trend itself when it rises.

How often prices close outside the channels

We counted daily closes outside the default Keltner channels, a narrower 1.5 ATR version and Bollinger Bands(20, 2) over the same days, from the 61st bar of each series to October 28, 2026 (October 23 for EUR/USD).

Market, fromOutside Bollinger 20 / 2Outside Keltner 20 / 2 ATROutside Keltner 20 / 1.5 ATRBollinger inside Keltner
S&P 500 ETF, April 199310.7%18.2%34.8%28.5%
Gold ETF, February 200512.7%23.7%37.2%18.0%
EUR/USD, March 201212.1%12.4%25.1%39.5%
Bitcoin, October 201712.2%15.0%27.8%41.8%

Share of daily closes outside the lines

Share of daily closes outside the lines. S&P 500 ETF, from April 1993: Bollinger Bands(20, 2) 10.7%, Keltner(20, 10, 2) 18.2%; Gold ETF, from February 2005: Bollinger Bands(20, 2) 12.7%, Keltner(20, 10, 2) 23.7%; EUR/USD, from March 2012: Bollinger Bands(20, 2) 12.1%, Keltner(20, 10, 2) 12.4%; Bitcoin, from October 2017: Bollinger Bands(20, 2) 12.2%, Keltner(20, 10, 2) 15.0%. Our computation; the inputs and the method are in the text.
  • Bollinger Bands(20, 2)
  • Keltner(20, 10, 2)
Our computation; the inputs and the method are in the text.Daily closes from the 61st bar of each series to October 28, 2026 (October 23 for EUR/USD).

With the standard settings, closes outside the Keltner channels were more frequent than closes outside Bollinger Bands on all four markets, and on the S&P 500 ETF, gold and Bitcoin most of them were on the upside: 14.3% of days above against 3.9% below on the S&P 500 ETF. The channels are not always the narrower of the two: they were narrower than the bands on 71.2% of days on gold and 58.6% on the S&P 500 ETF, but on only 45.7% on Bitcoin and 48.6% on EUR/USD. The difference is in trends. The standard deviation of closes grows when price runs away from its average, so Bollinger Bands widen with a strong move, while the ATR only grows if the daily ranges do. The data agree: sorting the days of each series into quarters by the size of the 20-day change in the close, measured in 10-day ATRs, the channels were narrower than the bands on 87.9% (EUR/USD) to 98.2% (gold) of the days with the largest moves, and on only 9.8% (Bitcoin) to 45.8% (gold) of the days with the smallest. A close above the upper Keltner channel is therefore a more common event than a close above the upper Bollinger Band, and should not be read as equally rare. The Bollinger Bands were fully inside the Keltner channels on 18.0% to 41.8% of days. To measure how far apart the Bollinger Bands themselves are, or where the close sits between them, see BandWidth and Bollinger Bands %B.

The channels on a real chart

The full chart below the introduction shows Keltner channels(20, 10, 2) on daily Bitcoin from July 1 to October 28, 2026. There were 13 closes above the upper channel and none below it, in four runs:

First close aboveCloseUpper channelCloses above in a row10 bars laterBack at the middle line
August 19, 202669,334.7967,736.369+12.83%September 10, +10.43%
August 29, 202678,23078,140.111+0.29%September 10, −2.12%
September 3, 202681,270.3780,618.471−5.45%September 10, −5.79%
September 21, 202686,62084,041.792−2.01%October 7, −3.81%

Bitcoin, 2026: change from each breakout close to the first close back at the middle line

Bitcoin, 2026: change from each breakout close to the first close back at the middle line. Break of August 19: +10.43%; Break of August 29: −2.12%; Break of September 3: −5.79%; Break of September 21: −3.81%. Our computation; the inputs and the method are in the text.
Our computation; the inputs and the method are in the text.Keltner(20, 10, 2) on daily Bitcoin, July 1 to October 28, 2026, the window of the full chart below the introduction.

The first break, on August 19, was the start of the move: price closed above the channel nine days in a row, from 69,334.79 to 80,249.58 on August 27, and was still 10.43% above the breakout close when it first came back to the middle line on September 10. The three later breaks came near the top of the rally, and price was lower each time it came back to the middle line. Read as a trend signal the first break worked; read as an overbought signal it would have meant shorting the first day of a run that added another 15.74% by August 27. The later breaks rewarded the opposite reading, and the channel itself did not tell them apart at the time.

Trend or reversal: the full history

To test the two readings on more than one rally, we measured both on four daily series. The trend reading: the return over the next 20 days after each first close above or below the channels, against the average 20-day return of all days. The reversal reading: buy the first close below the lower channel or short the first close above the upper one, and exit at the first close back at the middle line (short result = entry minus exit, divided by entry), one trade at a time, no stops, no costs.

Market20 days after a close aboveAll daysFade: buy belowFade: short above
S&P 500 ETF+0.84% (344 cases)+0.92%96 trades, 79.2% winners, +1.51% average128 trades, 53.9% winners, −0.54% average
Gold ETF+1.29% (231)+0.92%66, 66.7%, +0.85%93, 57.0%, −0.76%
EUR/USD−0.11% (75)−0.07%42, 69.0%, +0.21%43, 69.8%, +0.26%
Bitcoin+9.09% (94)+2.98%25, 60.0%, −0.22%39, 41.0%, −8.68%

Average return over the 20 days after a first close above the upper channel

Average return over the 20 days after a first close above the upper channel. S&P 500 ETF: After a close above +0.84%, All days +0.92%; Gold ETF: After a close above +1.29%, All days +0.92%; EUR/USD: After a close above −0.11%, All days −0.07%; Bitcoin: After a close above +9.09%, All days +2.98%. Our computation; the inputs and the method are in the text.
  • After a close above
  • All days
Our computation; the inputs and the method are in the text.Keltner(20, 10, 2), daily closes, same periods as the table; overlapping 20-day windows, not independent cases.

Two fades were still open on the last bar and are not counted in the table: the buy on the gold ETF after its close below the lower channel on September 28, 2026, at 377.91, was 1.16% down at the October 28 close (373.51), and the buy on EUR/USD after its close of September 16, 2026, at 1.14645, was 2.28% down at the October 23 close (1.12031), the last bar of that series.

On Bitcoin, closes above the upper channel behaved like breakouts: the next 20 days averaged +9.09% and were up in 69.1% of cases, against +2.98% and 53.3% for all days, and shorting them lost 8.68% per trade on average, with a worst trade of −58.26%. On the S&P 500 ETF a close above the channel said almost nothing about the next month, while buying closes below the lower channel won 79.2% of the time. On EUR/USD neither reading did much. The 20-day windows overlap when breaks cluster, so the counts overstate how many independent cases there are.

Does it work?

As shown above, the answer depends on the market and on which of the two opposite readings you apply; neither held on all four series. The channels are useful as a volatility-scaled frame around a trend, and as half of the squeeze setup. Our bench measures chart patterns detected by rule against random entries, not indicators (see our methodology); the channel pattern page has the statistics for drawn channels.

For related tools, see Bollinger Bands, the STARC Bands, the Donchian channel and the envelope indicator, or the full list of volatility indicators.

Every pattern's results on one sheet

The only pattern cheat sheet with tested stats: 152 patterns, what the textbook says and what really happened on 6 markets. Free PDF, sent to your inbox.

The PDF by email, plus our newsletter: new tests and results that change. Unsubscribe anytime. Privacy.

Frequently asked questions

What are the best Keltner channel settings?

The common default is a 20-day EMA with channels two 10-day ATRs away, the StockCharts setting. A 1.5 ATR multiplier gives narrower channels: on our four markets, 25.1% to 37.2% of daily closes fell outside them.

What is the difference between Keltner channels and Bollinger Bands?

Keltner channels use the average true range around an exponential average; Bollinger Bands use the standard deviation of closes around a simple average. Keltner channels are smoother, and they widen less during a strong trend, so closes outside them are more frequent: on our four markets, 12.4% to 23.7% of days against 10.7% to 12.7% for Bollinger Bands.

Is a close above the Keltner channel bullish?

It depends on the market. On Bitcoin since 2017, the 20 days after a first close above the channel averaged +9.09% against +2.98% for all days; on the S&P 500 ETF since 1993, +0.84% against +0.92%.

Who invented the Keltner channel?

Chester Keltner described the original version in his 1960 book How to Make Money in Commodities. The ATR version used on most charts today was introduced by Linda Bradford Raschke in the 1980s.

What is a Keltner channel squeeze?

When the Bollinger Bands contract inside the Keltner channels, volatility is unusually low. The TTM Squeeze is built on that condition.

Sources

  1. StockCharts ChartSchool, Keltner Channels
  2. StockCharts ChartSchool, Average True Range (ATR)
, Founder of PatternsWizard

Passionate about finance, quantitative trading and data analysis. Amaury built the PatternsWizard test bench, which detects each pattern on full price history and measures what happened next against random entries.

PatternsWizard tests chart and candlestick patterns on full market history, against random entries. How we test patterns.

179 patterns testedRandom-walk self-test passed

Related guides

All volatility indicators articles