ATR measures how much a stock moves in a typical session. Wilder introduced it in 1978 to size positions and place stops in commodity futures, where contracts gapped overnight and a plain high-minus-low understated the real movement. The call is @ATR(high, low, close, 14).
How it is computed
True range for a single bar is the largest of three distances:
TR = max(high - low, abs(high - prev_close), abs(low - prev_close))
The second and third terms are what make it a true range. If HDFCBANK closes at ₹1,700 and opens the next day at ₹1,640 after results, the intraday high-minus-low might be small while the actual move was ₹60. ATR is then Wilder’s smoothed average of true range over N bars.
The output is in rupees. A stock with an ATR of 18 has moved about ₹18 per session on average. That figure is not comparable between a ₹90 stock and a ₹3,900 stock, so most useful rules normalise it by price:
@ATR(high, low, close, 14) / close > 0.03
which selects securities whose typical daily range is more than 3% of price.
Typical parameters
| Expression | Window | Use |
|---|---|---|
@ATR(high, low, close, 14) | Wilder’s original | general volatility measure |
@ATR(high, low, close, 20) | one trading month | smoother, slower to react |
@ATR(high, low, close, 5) | one week | reacts fast, noisy |
Fourteen is Wilder’s number, carried over from the same book that gave us 14-period RSI. It is a convention, not an optimum.
How to read the output
ATR is a magnitude with no direction. A rising ATR means the stock is swinging more, whether up or down. It is most often used for sizing and for stop distance, where the logic is that a stop placed two ATRs away is loose enough to survive ordinary noise. Two is another convention.
Risks and caveats
ATR is backward-looking. It reports the volatility that has already happened, and volatility clusters but does not persist forever. A quiet ATR before an earnings date or an RBI policy announcement understates the risk you are about to take.
The gap component makes ATR sensitive to corporate actions. An unadjusted split or bonus prints as an enormous one-day gap and inflates ATR for the whole smoothing window afterwards, which silently distorts any position size or stop derived from it.
On thinly traded small-caps the daily high and low can be set by a handful of trades, so the range reflects illiquidity rather than genuine volatility. Circuit limits compound this: a stock locked at its upper band has a tiny recorded range on a day when the real move was capped by the exchange, not by supply and demand.
Sizing positions by an ATR window you optimised on history is a subtle form of overfitting, because the window choice feeds directly into leverage.