my rule: the bot gets one bad night, not two

been running perp strategies on automation for a while now and the thing that finally stopped me blowing up wasn't a better entry signal, it was a boring admin rule. process as it stands: - every strategy gets a fixed notional bucket. it does not get to borrow from another bucket because it's "on a run". i tried that once. it was not on a run, it was on variance. - daily loss cap per strategy, hard. hits the cap, it's flat and off until i look at it manually the next day. this is the "one bad night" thing. crypto trades while i'm asleep and drawdown compounds fastest at 4am when nobody's watching. - i journal the bot, not just the trades. every time i touch a parameter i write down what i changed and *why i believed it*, in one sentence. reading those back a month later is genuinely humbling. half of them are just "felt slow" which is not a reason. - i size assuming the funding side goes against me the whole hold. if the trade only works when funding is friendly, it's not a trade, it's a subsidy. where i'm still sloppy: i restart things too fast. cap gets hit, i stare at the logs for twenty minutes, convince myself it was a one-off liquidity thing, and turn it back on same day. that's me overriding my own rule with vibes. so now the rule is the restart has to happen the following day minimum, no exceptions, and if i can't wait that long the real problem is my position size, not the strategy. curious how other people handle the restart decision. do you have an actual checklist for turning something back on, or is it gut feel like mine used to be? and does anyone run a cooldown that's longer than a day, like a full week off after a bad stretch? tempted but worried i'd just never turn it on again.