The Keltner Channel, Drawn Out
A Keltner channel looks almost identical to a Bollinger band. There's a line down the middle and a band either side.
What differs is the ruler used for the width. Bollinger uses the standard deviation of closes; Keltner uses ATR. Same question, different instrument — and overlaying them shows how much that actually matters.
1. An EMA with ATR bands
The middle line is an EMA rather than Bollinger's simple average, so it leans slightly more on recent prices.
The bands sit a multiple of ATR either side of it. If you've read the ATR lesson this shape is familiar — the same raw material as the trailing stop, placed symmetrically on both sides instead of trailing on one.
A 20-period EMA for the middle, and bands 2 ATR either side using a 10-period ATR.
- Price (close)
- EMA 20 basis
- Bands — ±2 ATR
09/05 — the bands track ATR, so they widen with the size of recent bars rather than with how scattered the closes were.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
2. What ATR and standard deviation do differently
Overlay the two and the difference is immediate. Through quiet stretches they sit close together; when volatility arrives, Bollinger flares much harder.
Standard deviation squares its deviations, so one large move moves it a great deal. ATR is an average of bar sizes and simply doesn't spike the same way.
Same shape, different ruler: ATR measures bar size, standard deviation measures scatter around the mean.
- Price (close)
- Keltner
- Bollinger
Across this window Keltner's width ranged 1.5% to 9.0%, Bollinger's 1.0% to 26.6%. Standard deviation reacts far more violently to a burst of volatility than ATR does — which is the whole practical difference between them.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
3. When Bollinger sits inside Keltner
This is the best-known reason to run both at once. When the Bollinger band fits entirely inside the Keltner channel, standard deviation has compressed below what average bar size alone would suggest.
A widely-watched setup: standard deviation has compressed below what average bar size would suggest.
- Price (close)
- Keltner
- Bollinger
Bollinger sits fully inside Keltner for 95 bars here, across 6 separate stretches — the longest running 08/06 to 08/10. It's a description of compressed volatility, not a forecast; the Bollinger lesson's squeeze section makes the same caution at length.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
Plenty of write-ups present this as a signal that a big move is coming. That's an expectation, not an observation.
The Bollinger lesson makes the same caution at length in its squeeze section: narrow is a statement about the past, and it guarantees neither a direction nor that a move arrives at all.
4. Settings
The same two dials Bollinger has, doing the same two jobs.
- Price (close)
- Bands — ±2 ATR
At period 20, ×2.0: 30 of 181 closes finished outside the channel (16.6%).
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
5. All four side by side
The Academy covers four channels. Running each over identical data turns the difference between them into numbers rather than adjectives.
The same price data under each of the four channels in these lessons.
- Price (close)
- Keltner
Keltner: 16.6% of closes outside, average width 4.3%, ranging 1.5% to 9.0%. Four answers to one question — how far from the middle is far — and they disagree because they measure different things.
BTC/USDT · 4h · Binance · 2026-08-03 → 2026-09-05
6. In the Simulator
There's a Keltner breakout template: it enters when a close clears the upper band and exits when price falls back below the EMA basis.
To compare period and multiple combinations, the Sweep tool backtests every combination across the ranges you pick and ranks them.
- Keltner Breakout — enters on a close above the upper band, exits when price falls back below the EMA basis.