Volatility: ATR and Bands

A 30-pip stop is generous on EUR/USD and absurd on GBP/JPY. Traders who use the same pip distance everywhere are not applying a rule; they are applying a habit to instruments that do not share a scale. Volatility measures give you the scale.

What ATR computes

Average True Range is the average distance price covers in one period, over the last N periods — 14 by default.

"True range" rather than plain range because it accounts for a gap: it is the largest of the current high-to-low, the distance from the previous close up to today's high, and the distance from the previous close down to today's low.

Two properties make it useful, and one limitation makes it safe to rely on.

• It has no direction. ATR rises in a crash and in a rally alike. It measures magnitude only, which is exactly what a stop distance needs.
• It is in the instrument's own units. ATR on EUR/USD comes out in pips of EUR/USD, so it needs no normalising.
• It is backward-looking, like everything else here. Today's ATR describes the last fourteen periods, and tomorrow can be quieter or wilder.

Stops sized from volatility

The logic is straightforward: a stop should sit outside ordinary movement, or ordinary movement will hit it.

1. Read ATR on your trading timeframe.
2. Multiply. One to two ATR is the usual band. Tighter than 1 ATR gets hit by noise; much wider than 2 is rarely paying for itself.
3. Place the stop that distance from entry, or that distance beyond the structural level, whichever is further.
4. Size the position from the result, using Track 5's formula. The size falls out; it is not chosen.

Worked, on a $5,000 account at 1%

The three rows have wildly different stop distances and nearly identical risk. That is the mechanism doing its job — and it is why the same method can be run on a quiet major and a volatile cross without changing anything except the input.

Bands and standard deviation

Bollinger Bands draw a moving average with a band two standard deviations above and below it, recalculated each bar. The bands widen when recent closes are dispersed and narrow when they cluster.

The textbook line is that about 95% of observations fall within two standard deviations, so a touch of the band is unusual. This is where the misreading starts, because that figure assumes a normal distribution and price returns are not normally distributed — the extremes happen far more often than the model says.

The squeeze — bands narrowing to an unusual degree — is the one genuinely useful signal, and even it is only half a signal: volatility compression is followed by expansion, and the bands say nothing whatever about the direction.

Volatility by session and by pair

Volatility is not a constant property of an instrument. It varies across the day and across instruments, and both variations are predictable enough to plan around.

By session

Track 2 covered the mechanism; here is the consequence. A pair typically covers the clear majority of its daily range during London and the London–New York overlap, and very little of it during the quiet Asian hours.

• A stop sized from daily ATR is far too wide for an Asian-session trade, which will never travel that far.
• A breakout rule calibrated on London will fire constantly and fail in Asia, where the same rule is measuring noise.
• ATR read on an intraday chart already reflects the session mix, which is another reason to read it on the timeframe you trade.

By pair

None of this makes one instrument better than another. A volatile pair with a wide stop and a small position carries the same money risk as a quiet pair with a tight stop and a large one — which is the entire argument for measuring volatility rather than guessing at pip distances.