Sorry - I wandered into the conversation, didn't I? (I realise you were asking Spectre and not me!). It's just that whenever I've seen backtesting of different ways of doing things (not only entries and/or targets and/or stop-losses), for intraday trading purposes anyway, an ATR-multiple has always performed much better than an SD - both in my own backtesting (all the way from the days back in history when I was a retail spot forex trader) to the "institutional research" I've seen re automated futures-trading systems. So my impression - though I can't call it more than that - is that for most everyday trading purposes, ATR is probably, in principle, a more reliable approximation of "current volatility" (in its relevant sense, anyway) than SD's are. Apologies for a probably-unhelpful answer.
Calculating SD using price is like chasing a feather in the wind. Price time series are non-stationary data. A stationary time series is one whose statistical properties such as mean, variance, autocorrelation, are all constant over time. One common method to check if data is stationary is called an Augmented Dickey-Fuller test. Returns are generally more stationary than price, which is why return space is used so often in econometrics.