The Keltner breakout rules: why the slope filter is not optional
A band touch is not a signal. What makes this system work is the condition attached to it, and most published versions leave that condition out.
Key takeaway
- Entry requires two things: a close beyond the band and a middle line already sloping that way
- The multiplier decides everything about trade frequency — 2 is conventional, 1.5 doubles the signals and halves their quality
- The middle-line stop and the fixed ATR stop produce different systems; test the one you will run
Building the channel
- Middle line: an exponential moving average of the close. Twenty periods is the standard; it is fast enough to turn inside a swing and slow enough not to chatter.
- Band distance: average true range over 10 periods, multiplied by 2.
- Upper band: EMA + 2×ATR. Lower band: EMA − 2×ATR.
The two lookbacks do different jobs and there is no reason for them to match. The EMA defines the trend you are trading with; the ATR defines what counts as an unusual move today. A slow EMA with a fast ATR is a deliberate and common combination.
The entry has two conditions, not one
Close above the upper band and a middle line that is already rising. Either alone is a worse system than both together, and it is the second condition that most write-ups omit.
A close above the upper band while the EMA is flat or falling is the signature of a counter-trend spike in a range — the single most reliable way to lose money with this indicator. The slope condition costs you a handful of genuine reversals at the very start of a new trend and removes a much larger number of these.
Band alone
- Fires on every spike, in any regime
- Range-bound markets generate a signal every few days
- No information about direction beyond the touch itself
Band plus slope
- Only fires when the trend and the trigger agree
- Silent through most of a range, which is correct
- The EMA supplies the context the band cannot
How you measure the slope is a choice worth making explicitly: the EMA being higher than it was N bars ago is the simplest test, and it needs its own N. Five to ten bars is a reasonable range, and as with every other parameter here, what you want is a plateau rather than one lucky value.
The multiplier is the real parameter
Changing the EMA length adjusts the system at the margin. Changing the ATR multiplier changes what the system is.
| Multiplier | What it does | The cost |
|---|---|---|
| 1.5 | Many signals, entries early in a move | A large share are noise; costs accumulate fast |
| 2.0 | The conventional setting; a real departure from normal range | Enters later than the true start of a move |
| 3.0 | Only extreme moves qualify; very few trades | Misses most trends entirely; long idle periods |
Two stops, two different systems
The middle-line stop puts your exit at the EMA, which trails upward through a trend and tightens automatically as volatility falls. It gives the trade room early and takes profits mechanically late.
The fixed ATR stop — two ATR from entry, say — sits at a known distance you can size against before entering. It does not adapt, and in a strongly trending market it will be far behind price for a long time.
- Middle line: better trend capture, harder to size against, and it can be a long way from entry when the channel is wide.
- Fixed ATR: sized precisely, exits some trends early, and is the version that composes cleanly with volatility-based position sizing.
- Close back inside the channel is the exit both versions share, and on its own it is a reasonable complete exit rule.
Common questions
- Close beyond the band, or intraday touch?
- Close. With ATR bands an intraday touch happens constantly — a single wide bar can pierce the band and close in the middle of the channel, and that bar is not a breakout by any reasonable reading. This is one place where waiting for the close is not a stylistic preference.
- Can I use Keltner channels for mean reversion instead?
- People do — fading a band touch when the middle line is flat is a coherent range strategy, and it is the exact inverse of this system's slope condition. It is a different system with different statistics; do not run both on the same instrument and expect the results to add up.
- Why ATR(10) with an EMA(20)?
- Convention, not derivation. The faster ATR makes the bands respond to a volatility change before the EMA turns, which is usually what you want — the channel widens as the move begins rather than after it. Matching them at 20 and 20 is also defensible and slightly slower.
These are documented methods described for study. Nothing here is investment advice, a recommendation, or a claim about future returns — every system on this page has losing periods, and the pages say where.
Reading about a system is not having one.
Plutux is where you write your own rules down, test them against real data, and keep the record your memory would otherwise rewrite. Join the waitlist for early access.