what does that do-nothing baseline look like for 24/7 crypto?

so i finally read the baseline thing in the strategy library, the one where you run a deliberately dumb entry (rising streak, mid-day window, decent turnover) with a sane exit and whatever it spits out is the bar your "real" idea has to clear. love the idea honestly. it's the kind of test that would have saved me like four months of screenshotting equity curves and feeling clever. problem is basically everything i trade is coins, at 2am, and the whole thing is written for US stocks. so i'm stuck on how to translate it without accidentally building a *good* baseline by mistake. two bits i can't figure out: 1. the time filter. the article's point is that it deliberately sits in the boring part of the session so the entry has no information in it. crypto doesn't have a boring part of the session, or rather it has like six of them depending on where you live. do i just drop the window entirely and let it fire round the clock? my worry is that removes the "uninformative" property in a different direction, since now the baseline is quietly harvesting whatever the US open does to majors, and then it's not a null test anymore, it's a session strategy i'd have to beat. 2. the exits. +2 / -1 / 20 min hold is asymmetric on purpose i think, and on a liquid large cap that's a real distance. on some of the stuff i touch that's noise, it'd get hit both ways in a minute. if i scale those numbers up so they're comparable in ATR-ish terms, am i still running the same experiment? or does rescaling the exit quietly change the baseline into something else, given the article's own conclusion is that most of the edge in the thing lives in the exit rules anyway. related third thing: for stocks the honest comparison is also "did you beat just holding the index". for crypto do people bother comparing against just holding the coin? feels like it'd wreck most of my ideas over long enough samples but maybe that's the point. anyone actually ported this over? curious what you kept and what you threw out.
N
NightOwl_Yuki
27 posts · 0 followers
+ Follow

Related reading

4 replies

P
PremarketPete· Jul 2026 ago
- keep the window, just pick one and hardcode it (some fixed UTC block, same one every day). the point isn't that it's boring, it's that you didn't choose it based on results - if you drop it entirely you can't tell whether the baseline is timing or entry - run both if you're curious, log them separately - rescale exits in ATR, not in percent. same experiment, different units
CryptoKarl· Jul 2026 ago· edited Aug 2026 ago
the exit thing you're worried about is real but not in the direction you think. rescale to ATR and you're fine, the baseline is still "no information in the entry" which is the whole point. what actually breaks the port is fees and funding, on stocks a 20min baseline costs you basically nothing, on perps you're paying to exist. my dumb baselines used to look flat until i added funding and then they were a slow bleed machine. and yeah, hold-the-coin as benchmark is brutal, that's why nobody posts it lol. i did it once for a whole year sample and quietly closed the spreadsheet.
N
NightOwl_Yuki· Jul 2026 ago
yeah the fixed-UTC-block-you-didn't-pick-from-results thing is exactly the bit i was missing, thanks. and karl, funding is a good catch, i'd have run the whole thing spot-style and then wondered why live didn't match. do you subtract it per bar or just as a flat per-hour drag? my hold is short enough that most trades wouldn't even cross a funding stamp, which feels like it'd make the baseline look better than any strategy i'd actually run for hours.
SlowSwing_Sam· Jul 2026 ago· edited Aug 2026 ago
the hold-the-coin comparison wrecking your ideas *is* the finding, mate. that's not a bug in the test. also gently noting you're doing all this at 2am to see if a 20 minute hold beats sitting still. The answer might be sitting there in the question.
Sign in to reply →
← All strategies & backtesting