that dip-fading writeup — does the asymmetry even survive on perps?

read the mean reversion piece in the strategy library, the one about buying stuff that's dropped way under its own average and why it mostly ends badly. honestly the most useful part for me was the bit where they point out the shape: small target, wider stop, nice win rate, ugly tail. that's basically every fade I've ever taken on perps, just with more leverage and less self respect. what I'm chewing on is whether the logic transfers at all. couple of things bug me: 1. the setup leans on a daily volume filter and a 300 period sma. stocks have a session, an open, a close, a vwap everyone stares at. perps don't close. my 300 bar average at 4am is a completely different animal than at US open, same number, totally different meaning. do you gate it by session/hour or just let it run 24/7 and accept the noise? 2. the time exit (~20 min) makes sense when there's a close coming. on crypto I've noticed the flushes that actually snap back do it in like 2-4 minutes, and if it hasn't bounced by then it's usually because someone is still unloading. so I'd be tempted to cut the hold time hard, but then you're eating a lot of tiny scratches. 3. the streak condition (20 falling bars or whatever) is the part I like most conceptually — it's not just "cheap", it's "cheap AND still going", which is exactly the thing that kills you. is that meant as a confirmation or is it actually a trap in disguise? feels like it makes you buy later, which is better, but also buys you into stronger momentum against. I'm not asking anyone to hand me numbers. more curious how people here think about the stop side. article is honest that -2% is the real risk of the whole category, but on a perp with any size that stop is not really a stop, it's a suggestion when things gap. tempted to just size way down and treat the wide stop as the actual position, instead of pretending I'll be able to exit at the level. anyone actually run a fade with a hard time-based bailout instead of a price stop? curious if the distribution looks less horrible or just differently horrible.