Realized volatility from ticks beats ATR for sizing scalps
ATR calculated from one-minute bars is a staple for sizing stops and targets. It has drawbacks: it uses only high, low, and close, so it misses the path inside the bar, and it responds slowly to rapid changes in volatility.
Realized volatility computed from tick or second-level returns uses far more information. It reacts quickly when activity picks up and can be measured over short windows, say the last 60 seconds. That fits the time frame of scalping better.
A practical method: sample mid prices every second, compute log returns, and take the standard deviation over a rolling window. Scale it to the holding period to estimate the typical movement during a trade.
Using that estimate, stops and targets adjust automatically. In quiet moments, they tighten. In active moments, they widen, keeping the probability of being stopped by noise roughly constant.
Be wary of microstructure noise at very high sampling rates, where bid-ask bounce inflates measured volatility. A sampling interval of 1 to 5 seconds usually balances detail and noise.
Which volatility measure does the system use to size its stops, and how often does it update?
Realized volatility from ticks beats ATR for sizing scalps
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: Realized volatility from ticks beats ATR for sizing scalps
Fourteen-period ATR on the one-minute chart was too slow for the way I scalp. On several London openings it lagged the pickup in activity by six to eight minutes, so my stops were still sized for the quiet half hour that had just ended. I switched to realized volatility from two-second mid returns over a ninety-second window, updated every fifteen seconds, and I keep a floor so the EURUSD stop cannot tighten below three point four pips.
Over the next forty-one trading days the share of trades stopped out by noise dropped from thirty-one percent to nineteen percent on the same signals. Targets scale with the same number, which stopped me from holding a three-pip goal into a dead tape without noticing. Sampling every tick inflated the reading with bid-ask bounce, which is why I settled on two seconds.
I still glance at ATR as a background check, but the live size and stop now come from the short realized window. It feels closer to what the next minute is likely to do than a measure built from completed one-minute bars.
Over the next forty-one trading days the share of trades stopped out by noise dropped from thirty-one percent to nineteen percent on the same signals. Targets scale with the same number, which stopped me from holding a three-pip goal into a dead tape without noticing. Sampling every tick inflated the reading with bid-ask bounce, which is why I settled on two seconds.
I still glance at ATR as a background check, but the live size and stop now come from the short realized window. It feels closer to what the next minute is likely to do than a measure built from completed one-minute bars.