ATR and volatility: sizing the market's normal noise

You get stopped out, and then price recovers and runs to the target you originally had in mind, without you. It happens often enough that people start believing their broker is hunting them. Usually the explanation is duller: the stop was placed inside the range the asset covers on an ordinary day, so an ordinary day removed it.

Volatility is the missing variable in most retail stop placement. A 1% stop is generous on one instrument and absurd on another, and nothing about the number 1% tells you which is which.

What ATR measures

Average True Range takes the true range of each bar - the largest of high minus low, high minus previous close, and previous close minus low - and averages it over N bars, usually 14.

The gap-aware part matters. A bar that opens well above yesterday's close and then trades quietly still represents a large real move, and true range captures that where a simple high-minus-low would not.

The output is in the instrument's own units: dollars, points, pips. ATR of 1,400 on an asset trading at 60,000 means an average daily travel of roughly 2.3%. ATR of 0.6 on an asset trading at 40 means about 1.5%. Comparing raw ATR across assets is meaningless. Comparing ATR relative to price, or comparing an asset with its own recent ATR, is not.

Noise is not the same as being wrong

Here is the mental shift that makes ATR useful. A market moving against you by less than its normal daily range has not told you anything. It has done what it does every day.

A stop is a statement that your idea is invalid, and an idea is not invalidated by routine movement. If your reasoning is based on a daily structure but your stop sits at 0.4 ATR, then your stop is measuring something much smaller than the thing you were reasoning about. The mismatch guarantees a low win rate that has nothing to do with the quality of your analysis.

This is why stop distance has to be derived from the market rather than from your comfort level. The two most common ways people set stops - a round percentage, or the dollar amount they are willing to lose - both ignore the instrument entirely.

Wider stops do not mean bigger risk

The objection is immediate: a wider stop loses more money. That is only true if position size is fixed, and position size should never be fixed.

Risk on a trade is stop distance multiplied by position size. Fix the risk first as a percentage of equity, then let stop distance determine size:

• Equity 20,000, risk 0.5% = 100 currency units at risk.
• Stop distance 2.5 ATR, and ATR is 1.60, so the stop sits 4.00 away.
• Position size = 100 / 4.00 = 25 units.

Double the ATR and the position halves, while the money at risk stays at 100. Volatile assets get a wider stop and a correspondingly smaller position. That single rule removes most of the accidental variation in how much any given trade can cost you.

Indikora works this way by default: risk per trade of 0.5% to 0.75% of equity depending on style, with size derived from the stop distance rather than chosen first.

The chandelier stop

A fixed stop is a line in the sand from entry. A chandelier stop moves.

It anchors to the highest price reached since entry and sits a multiple of ATR below that high. Indikora uses 3 x ATR. As price advances, the high advances, and the stop follows. It never moves down.

Two properties make this behave well. It adapts to volatility, so a market that becomes wilder gets more room automatically instead of stopping you out for expanding. And it converts an open profit into a floor without requiring a decision, which removes the moment where you would otherwise talk yourself into moving a stop.

The cost is real and worth stating: you will always give back some of the peak. A 3 x ATR trail hands back roughly three average days of movement from the high on every exit. That is the price of not exiting early on every pause. There is no trailing rule that keeps the peak.

Choosing a multiple honestly

People spend a long time searching for the best ATR multiple. The search is mostly wasted, for the same reason as tuning moving average lengths.

Small multiples raise your win rate on the losing side and cut your winners short. Large multiples give back more but stay in longer. Somewhere around 2 to 3.5 ATR is where most trend-following work lands, and the exact number inside that band matters far less than applying it consistently.

If a backtest only works at 2.7 and falls apart at 2.5 and 3.0, that is not precision. That is a curve fit, and the next lesson on backtesting is about how to catch it.

Volatility is not the enemy of a position. It is the medium the position moves through. Once your stop is measured in ATR rather than in dollars you feel comfortable losing, most of the "stopped out then it reversed" experience quietly stops happening.