How do you decide when to intervene manually?

Mine is: I may close, I may never open. Cutting something is a risk decision. Adding one is me deciding I'm smarter than the plan I made when I was calm.

Add a delay to yourself. If I still want to intervene after writing down why, I do it. Usually the writing is enough.

Both of these are better than what I've been doing, which is "vibes and regret."

Grandpa's rule is the one I stole years ago and it still holds!! Close yes, open never. The urge to add is ALWAYS the loud part of my brain talking!

I kept a log of every override for about six months. I was right on the individual trade maybe 60% of the time and still ended the period behind where the system would have been, because the ones I let ride after overriding turned into a different animal entirely. Turns out being right about the trade and being right about the intervention are two separate questions and I only ever checked the first one.

funding rates at 3am have made me press buttons no system of mine would ever press lol. my hack is the override has to be typed on a laptop, not the phone. sounds stupid but the friction of getting up filters out like 80% of it

I would ask what you are actually observing when you feel certain. In my experience the intervention impulse fires most often at the exact points where the strategy's edge is thinnest, which means my sample of "obviously about to do something dumb" is biased toward high-variance moments rather than genuinely bad ones. Have you tracked whether your overrides cluster in particular conditions?