Many people use some percentage of an ATR-like measure for stops. This is maybe OK if you are trading on a daily timeframe, but on an intraday time frame, the swings can get stupid (for example, oil, nasdaq compared to ES/RTY). I am struggling to automate the selection of a stop that takes into account these intraday swings. I can eyeball it for each instrument, but that isn't scalable. For ES, you might use the range of the half hour candle, whereas for gold you might need to use the hourly candles. I'm just making this up. There must be some way to quantify the range of intraday swings that are dependent on not just time, but price action as well. Ideally I'd like to somehow capture the range of the instrument using swings like the screenshot.
One option: Determine moving average crossovers, call those swings (on average probably true) and use the high/low as the range for that swing.
The peaks and troughs of Rate of Change might be useful. The major problem being how you will define a swing.
Swings/turns come in all shapes, forms, sizes, deviations....trying to guess at any one moment in time is very tough...without reasonably knowing the underlying assumptions, expectations and rationale behind it all. Similar to all the people/sheep on social media who say.....Buy the Dip.....but it dips, and still dips, and still dips....while you lose money and your mind. Understanding with wisdom versus gambling....has profound differences and results.
I just want something idiot proof that approximates it correctly. Rate of change is a good suggestion.
This dumb approach shows good results for index futures and oil. Gold not so much, but that's expected since it's a bastard.
You are asking for the key to solve the day-trading problem. If you can make it, well that's it. You can get some ideas at these kind of papers: https://www.researchgate.net/public...olatility_forecasting_from_implied_volatility
What’s the purpose of the stop? Do you want it placed outside «noise»? One suggestion could be the 1/5-minute range value. Maybe plot an MA on that and decide on some kind of multiple of that.