Why size a position off volatility instead of a flat percentage?

A 5% stop means two entirely different things on two different stocks. On a sleepy utility, 5% is a month of movement. On a volatile small cap, 5% is Tuesday. Sizing every trade off the same percentage quietly gives you wildly different odds of being stopped out by noise alone.

Volatility-based sizing fixes that by measuring how much the instrument normally moves, and setting both the stop distance and the size in those units.

ATR: the normal-movement ruler

Average True Range (ATR) is the average, over the last N bars (14 is the common default), of the true range — the greater of: today's high minus today's low, |today's high − yesterday's close|, and |yesterday's close − today's low|. The extra terms exist so that gaps count as movement.

ATR is not a direction signal and not a prediction. It's a units converter: it tells you what "one normal bar" is worth in dollars on this specific instrument. The extra terms are why the two sessions below — same high, same low — have true ranges that differ by a factor of two.

The sizing formula

Shares = (Equity × risk fraction) ÷ (k × ATR)

where k is how many normal bars of breathing room you're giving the trade. Set k = 2 and the stop sits two average bars away.

Worked example

A hypothetical $50,000 account, 1% risk = $500, k = 2. Two stocks trading at the same $40:

Both positions risk $500, or a whisker under it: 500 ÷ 1.60 is 312.5 shares, and rounding down to 312 leaves $499.20 at risk rather than $500.00. Round down, always — the budget is a ceiling. Both give the trade two average bars before the exit triggers. The dollar exposures differ by more than 3× — and that is the intended result, not a bug. The calm stock earns a larger position precisely because a given dollar move there is a bigger event.

A flat "5% stop" rule would have given both stocks a $2.00 stop and identical share counts: 0.8 ATR of room on Stock A (stopped out constantly) and 2.5 ATR on Stock B (fine). Same rule, two different games.

What ATR sizing does not do

• It's backward-looking. ATR describes the last 14 bars. Volatility clusters and jumps — the bar before an earnings release tells you very little about the bar after it.
• It sizes for ordinary movement, not for a gap. True range already counts yesterday's gaps, so they are inside the average — but no average of past ranges bounds tomorrow's, and a position sized to survive 2 ATR of noise is not sized to survive a 6 ATR overnight gap. Unit 2 covers what a stop can and cannot promise.
• It doesn't make the trade good. Volatility normalisation makes positions comparable. Comparable bad trades are still bad trades.

And the thing it measures arrives in clumps, which is why the ruler is always slightly the wrong length: quiet follows quiet until it doesn't.

In the data

The ruler has a length, and the length changes the reading. Below is the S&P 500 fund with its ATR drawn at three windows, 5, 14 and 50 sessions, on one scale:

The short one jumps at every burst of movement and the long one barely notices, and they tend to disagree most just after a quiet spell ends. A stop sized on one of them is a different stop from one sized on another, so quote the window beside every ATR you use. And it is in dollars, so divide by the price before comparing it with any other instrument's.

Try it now

1. Start with the ruler. Compute the true range of both marked sessions on the first schematic above and confirm which of the three candidate terms wins in each. Then read the real thing for the two instruments below, clearly different in character and over the same year: the latest 14-day ATR of each is in the first two tables, the latest price of each in the third, and the charts under them draw the ATR across the year.

2. For each, compute the stop distance at k = 2 and then the share count for a hypothetical $25,000 account at 1% risk. Round down.
3. Compute the dollar exposure of each position and write one sentence explaining why the larger exposure is the lower-volatility name. If that sentence feels backwards, re-read the worked table — it is the whole point of the lesson.

Unit done. Next: the exit that all of this arithmetic depends on.