Indicator library · Position management
Chandelier Exit, An ATR Trailing Stop
A stop that hangs a multiple of the average true range below the highest high since entry, and rises as that high rises. The only measure in this library concerned with leaving a position rather than finding one.
The calculation
For a long position: exit = highest high − (multiplier × ATR). For a short: exit = lowest low + (multiplier × ATR). Conventionally three times a 22-period average true range, with the high taken since entry or over the same lookback.
Two properties follow and both are deliberate. The level ratchets: it rises with the highest high and never falls, so a gain is progressively protected without anyone having to guess where the move ends. And the distance is denominated in the instrument’s own volatility rather than in a percentage, so the same rule is neither tight on a quiet stock nor loose on a turbulent one.
What the multiplier is actually choosing
The multiplier is the only real decision in this indicator, and it is a decision about how much ordinary movement you are willing to sit through, not about how much you are willing to lose. Those two are often confused, and the difference is the reason ATR is the input.
| Multiplier | Behaviour |
|---|---|
| 1.5 ATR | Reached by ordinary pullbacks. Exits early and often; a shakeout like the one in the figure would close the trade. |
| 3 ATR | LeBeau's convention. Wide enough to absorb a normal pullback, so the give-back at a genuine turn is correspondingly larger. |
| 5 ATR and beyond | Rarely reached by noise and rarely reached in time. At some width it stops being a stop and becomes a formality. |
There is no setting that exits at the high and survives every shakeout, for the same reason no moving average is both early and reliable: the information needed to distinguish a shakeout from a reversal arrives after the fact. What the multiplier buys is a choice about which of the two errors you would rather make, and how much of the move you are prepared to hand back for it.
Why a stop is a decision made in advance
The one property that makes this measure worth anything is that it is computed before the session it applies to. The level for tomorrow is known tonight (a highest high and an average true range, both from data already in hand), so there is nothing left to decide when price reaches it.
That is the whole discipline, and it is why a trailing stop belongs in a reference about measurement rather than in one about prediction. A rule that is re-examined when it triggers is not a rule: it becomes a judgement made at the worst possible moment, under the pressure of an open position and a falling price. Writing the multiplier down in advance, and letting the arithmetic place the level, is what converts an unknown loss into a known one.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Used in a range | The high stops rising, so the level sits below the range and a normal swing reaches it. Trailing stops are trend tools. |
| Expecting an exit near the high | The give-back is the multiplier times the ATR, by construction. Anything better requires knowing the future. |
| Volatility collapse | A falling ATR pulls the stop closer on its own, so a position can be closed by the market going quiet rather than by it turning. |
| Trigger convention unstated | An intraday touch is materially tighter than a close below. The same multiplier means two different stops. |
| Gaps | A gap through the level exits at whatever the market opens at, not at the level. The stop is a decision rule, not a guaranteed price. |
| Level allowed to fall | A variant that lets the stop drop back with the high is not this indicator, and gives up the ratchet that made it worth using. |
What volume adds
The exit level is computed from highs, lows and closes, so it cannot distinguish the two cases that matter most when it is approached. A pullback into the level on drying volume is a move that has run out of sellers; the same pullback on the heaviest volume in weeks is a market changing its mind, and the second is far more likely to keep going.
Nothing about that argues for overriding the stop, a rule that gets overridden on judgement is not a rule. Where it does belong is in the decision made beforehand: an instrument whose recent advances have come on thinning participation deserves a tighter multiplier from the outset, chosen in advance and written down, rather than a wider one defended in the moment.
Frequently asked questions
How is the chandelier exit calculated?
For a long position: take the highest high since entry — or over a lookback, depending on the variant — and subtract a multiple of the average true range. Conventionally three times a 22-period ATR. For a short position, take the lowest low and add the same multiple. The name is Chuck LeBeau’s: the stop hangs from the high the way a chandelier hangs from a ceiling, and it rises as the ceiling does.
Why measure from the high rather than from the entry?
Because a stop that never moves gives back everything a trade gained, and a stop that trails the price protects the gain without asking you to predict the top. Anchoring to the highest high means the exit rises with the position and never falls. The ratchet is the whole point, and a variant that lets the level drop back is not this indicator.
What does the multiplier decide?
How much ordinary movement the position is allowed to absorb before the trade is closed. At three ATRs the stop sits far enough away that a normal pullback in that instrument will not reach it; at one and a half it will be reached frequently, including by noise. The multiplier is a statement about how much adverse excursion you are willing to sit through, denominated in the instrument’s own volatility rather than in a fixed percentage.
Why use the ATR rather than a percentage?
Because a fixed percentage means something different on every instrument. Eight per cent is a routine week on a small-cap and an extraordinary month on a utility, so a percentage stop is tight on one and loose on the other without your intending it. The ATR expresses the distance in units of what that instrument actually does, which is the same argument that makes it the right input for position sizing.
Should the trigger be the close or an intraday touch?
The close, if you want the level to mean what the calculation implies. An intraday touch is reached by any spike, including a single unrepresentative print in a thin book, whereas a close below the level says the session settled there. Using the close makes the stop meaningfully wider than the number suggests, and using the touch makes it meaningfully tighter, either is defensible, and the choice belongs in the record beside the multiplier.
Does it work in a range?
It does what it is built to do and the result is unhelpful. In a sideways market the highest high stops rising, so the level stops ratcheting and simply sits below the range until a normal swing reaches it. Trailing stops are trend-following tools; in a range they close positions at the bottom of the range with dull regularity. A regime measure such as the efficiency ratio or ADX is the honest companion.
How much does it give back at a reversal?
By construction, roughly the multiplier times the ATR, that is the whole design, not a flaw. A trailing stop cannot exit at the high, because identifying the high requires knowing the future; what it does instead is convert an unknown loss into a known one. The figure on this page gives back a defined amount at the turn, and the alternative on offer is not a better exit but a prediction.
Can the same idea be used to enter?
It is occasionally used that way, a close above the short-side level treated as a long entry, and it inherits every weakness of a breakout system while adding a volatility-scaled delay. The measure was designed for exits, and exits are where it is defensible. Nothing about the calculation contains information about whether a new move will continue.