the angle/window scaling thing from that crypto momentum writeup

Read the crypto momentum entry piece in the library last night, the one with the liquidity floor on DayAmount, and I've been chewing on it all morning. Mostly because it broke an assumption I'd been carrying around for months. The part that got me: ChangeAngle(n) being an actual angle in degrees, so the underlying move it demands scales with the window length. I had been sweeping window and angle as if they were independent knobs. They are not. Which means half my "parameter surface" was probably just regions where the condition was mathematically almost unreachable, and I was reading those flat dead zones as "robust". So, numbered, because that's how my brain works: 1. If angle and window are coupled, the honest thing to sweep is the implied change-rate threshold, not the angle. Has anyone actually done it that way? I'm thinking of picking a target rise in the change rate first, then solving backwards for the angle at each window I test. Feels cleaner but I've never seen anyone describe it. 2. The 60s/115 strength combo in the example — that's tuned for something. I don't want to lift it. I want to understand what regime it assumes. Fast window plus a strength filter reads to me as "only take the ones already extended", which historically is where I get chopped. 3. The liquidity floor. I like it conceptually a lot more than I like it as a fixed number, because a fixed dollar volume floor is a different filter depending on what the overall market has been doing. Anyone using something relative instead, floor as a percentile of the universe, or scaled to the asset's own trailing median? 4. The hold cap argument I have no quarrel with. "No session close means nothing ends a bad trade for you" is the single best sentence in the writeup. Equities gave me a free exit every day and I never appreciated it until I started testing 24h stuff and watched positions just sit there through the night doing whatever they liked. 5. My worry with the time exit is that it becomes the parameter everyone secretly overfits. It's so easy to sweep hold time, find the value that rescues your equity curve, and tell yourself it was risk management. I'd rather set it on "how long am I willing to be exposed while asleep" grounds and never touch it again. Mainly asking about points 1 and 3. Anyone got a defensible way to sweep coupled parameters without kidding themselves?