the 40 min time stop in that sleep-safe crypto setup — too blunt?

CryptoKarl· Jun 2026 ago· 👍 6· 💬 8

read the unattended crypto piece in the strategy library last night (the one with three separate exits) and I can't stop chewing on one part. the bit I actually agree with straight away: writing the three exits as separate branches instead of one big or-condition. I've done the clever one-liner thing before and then stared at it half asleep on a phone screen trying to work out which clause fired. so yeah, boring code as a safety feature, sold. what I'm stuck on is HoldTime >= 2400. forty minutes and out, no questions asked. on the entry side it's asking for real momentum — big turnover, strength gate, plus an angle over a 2 min window. so you're deliberately buying things that are moving. then the time exit treats a runner and a dead sideways bag exactly the same. feels like the one exit that can cut the trade that was actually paying you. things I'm considering instead of just raising the number: - leave 2400 but only apply it if profit is under some small threshold, so flat-and-boring gets flushed and a winner is left alone - keep it strict, accept the clipped tails, on the grounds that unattended means unattended the other thing that got me was the angle scaling note, same degrees over a longer window demands way more actual movement. I wanted to widen my window to filter noise, but if I widen and keep the threshold I'm quietly making my gate much harsher without noticing. so what's the sane way to pick the new threshold when you change the window? do people re-derive it or just eyeball how many entries per day they still get? curious what the unattended crowd here does. I'm running perps so my downside branch matters more than most, and I do NOT want to be the guy who softened a stop at 3am.

CryptoKarl

30 posts · 0 followers

+ Follow

8 replies

QuietVol· Jun 2026 ago· edited Aug 2026 ago

The time stop isn't really an exit rule, it's a statement about your holding-period distribution, if the entry signal's edge decays by ~40 min then it's doing its job, and if it doesn't, you're truncating the right tail for no reason. The profit-conditional version you describe is basically saying "only flush the ones that didn't work", which sounds sensible but does change the shape of what you keep: the survivors are then exited by the stop or the target only, so your loss tail gets a bit longer too. Have you actually looked at where your exits currently cluster — i.e. what fraction of closes are the time branch firing, and what the P&L distribution of those looks like versus the other two?

MomoQueen· Jun 2026 ago· edited Aug 2026 ago

The clipped runner thing bugs me too but honestly in momentum most of what's still open at 40 min is dead weight! I'd rather flush ten of those than babysit one. If you do go conditional keep the threshold tiny, like barely above break even — otherwise you're just holding losers longer and calling it patience.

DrawdownDave· Jun 2026 ago

I once turned a time stop into a "time stop unless it's green" and then quietly widened what counted as green because a trade kept just missing it. Two weeks later I had no time stop at all and a bag I was calling a swing position. Not saying don't do it, just... write down the number somewhere you can't edit at 3am.

DataDrivenDee· Jul 2026 ago

Genuine question on the profit-conditional idea — how would you test it without kidding yourself? If you rerun history with the softer rule you'll see the winners that got to keep running, but you also need to count the ones that were green at 40 min and gave it all back. Would you be able to log, for every time-stop exit you've already had, what the position was worth 20 or 30 min later? That's the cheap version, no rewrite needed.

SpreadWatcher· Jul 2026 ago

Nobody's mentioned that on perps you're paying to hold. A time stop is partly a funding rule, not just a signal-decay rule, and if you extend the clock on the winners you're also extending your exposure to whatever fee schedule you're on. Cheap thing to check before you tune anything: what does 40 min actually cost you round trip vs the average size of the moves you're catching.

NightOwl_Yuki· Jul 2026 ago

for the angle window question i just eyeball entry counts, not gonna pretend otherwise. widen the window, watch how many fires per night, tune the threshold til it's roughly what it was before, then leave it alone for a couple weeks before judging. and yeah re the 40 min — my overnight stuff is mostly asleep-market hours anyway, moves are slower, so i run a longer clock there than i would in the busy hours. might be worth splitting it by time of day rather than by profit?

CoffeeAndCharts· Jul 2026 ago

morning check-in habit here so the 40 min thing lands different for me — anything still open when I sit down with coffee I already treat as suspect. DrawdownDave's slippery-green story is the one I'd pin above the desk though, that's exactly how mine went too.

HalfKelly· Jul 2026 ago· edited Aug 2026 ago

On the angle window: you can just re-derive it instead of eyeballing. If you're measuring degrees, the angle comes from arctan of (price change / window length) in whatever units your chart axes use, so doubling the window while keeping the same degrees means you're demanding roughly double the price move. To keep the gate equally strict, take tan(old threshold), divide by the ratio you widened the window by, and take the arctan of that. Then sanity check against Yuki's entry-count method — if the maths and the count disagree wildly, your axis scaling isn't what you think it is. Separate note: if you do end up letting winners run past the clock, your average holding period goes up, which quietly raises your open exposure per unit of time even at the same position size.

← All strategies & backtesting

the 40 min time stop in that sleep-safe crypto setup — too blunt?

Related reading

8 replies