Realised volatility versus ATR on very short windows
Posted: Sat Oct 31, 2026 7:48 pm
ATR is widely used, but it has quirks that matter when working in seconds and minutes.
It's based on the high-low range of each bar, which depends on the bar boundaries. A burst of movement that straddles two bars may look smaller than one inside a single bar. That makes ATR noisy on very small timeframes.
Realised volatility, calculated from the standard deviation of tick or second returns, captures the actual path of price. It responds faster and doesn't depend on candle boundaries.
For scalping, that responsiveness helps. A rolling realised volatility over the last 60 seconds can flag the start of a move earlier than a one-minute ATR.
Calculating it needs tick or second data and a bit of code. Many platforms allow custom indicators for it.
The downside is that tick data contains the bid-ask bounce noise, which inflates the figure. Using mid prices or sampling at fixed intervals, such as every five seconds, reduces the distortion.
Which volatility measure do you trust for short-term decisions?
It's based on the high-low range of each bar, which depends on the bar boundaries. A burst of movement that straddles two bars may look smaller than one inside a single bar. That makes ATR noisy on very small timeframes.
Realised volatility, calculated from the standard deviation of tick or second returns, captures the actual path of price. It responds faster and doesn't depend on candle boundaries.
For scalping, that responsiveness helps. A rolling realised volatility over the last 60 seconds can flag the start of a move earlier than a one-minute ATR.
Calculating it needs tick or second data and a bit of code. Many platforms allow custom indicators for it.
The downside is that tick data contains the bid-ask bounce noise, which inflates the figure. Using mid prices or sampling at fixed intervals, such as every five seconds, reduces the distortion.
Which volatility measure do you trust for short-term decisions?