Community › Automated Trading

the angle scaling thing in that crypto momentum writeup

GGapHunterMike· Jun 2026 ago· 👍 6· 💬 6
read the crypto momentum entry piece in the strategy library. the one with the liquidity floor on it. the part that stuck with me is the angle scaling. i never thought about it that way. degrees over a window, so the wider the window the more actual move you're demanding. that explains something. i had a version of this running with a 300s window and the same threshold i used on a 60s window. it never fired. not once. i figured my filter stack was too tight somewhere. it wasn't. the condition was just unreachable. so two things i'm chewing on. one. if i want a 5m window instead of 60s, what angle are people actually landing on. the piece gives the math for why you have to drop it but not a number. i can back into it but i'd rather hear what's held up live. i don't want to overshoot and end up with a filter that lets everything through. two. the hold cap. 3600s. i get the logic, nothing else ends the trade when you're asleep, no session close, agreed. but it's labelled swing in the header. an hour cap on a swing horizon feels like it's cutting the runners off at the knees. or is the point that unattended isn't swing, it's just momentum with a timer bolted on. also curious if anyone moves the volume floor around by pair or leaves it fixed. i'd assume fixed is safer but i've never tested it. not asking for a config. asking what broke for you.
G
GapHunterMike
17 posts · 0 followers
+ Follow

Related reading

Automated Tradingthe angle window math from the sleep-safe crypto writeup — am i reading it right?Automated Tradingwidening the angle window without killing the entry - sanity checkStrategies & Backtestingthe angle scaling bit in that crypto momentum writeup - questionsStrategies & Backtestingthe angle/window scaling thing from that crypto momentum writeup

6 replies

MLcurious· Jun 2026 ago
ok the unreachable condition thing just explained a month of my life, i had a model spitting signals into a rule that literally could not evaluate true and i was staring at the model. dumb question but does ChangeAngle care about the size of the move at all or only the slope? like a slow grind and a sharp pop could give the same angle if the window is right? still confused on that part honestly.
N
NightOwl_Yuki· Jun 2026 ago
on the hold cap thing — i trade nights so this is basically my whole life lol. my take is the cap isn't there to optimize the exit, it's there so you don't wake up to something that's been bleeding for six hours with no one watching. it's insurance, not strategy. i keep mine tight for that reason and i'm fine giving up the odd runner.
P
PremarketPete· Jun 2026 ago
- had the same fixed vs per-pair question - ended up fixed, one floor, applied to everything - reason: per-pair meant i was tuning a number per pair and i couldn't tell tuning from curve fitting anymore - fixed floor is boring but i know what it's doing - re: window, whatever you pick, log the fire count for a while before you judge the win rate. zero fires reads the same as a bad filter until you check.
M
MomoQueen· Jun 2026 ago
don't back into the angle from math alone!! i tried that, scaled it by the window ratio and it still barely fired, because price doesn't move in a straight line for 5m the way it does for 60s. noise cancels itself. i had to go lower than the pure ratio said. log your would-be triggers for a week before you let it trade, that's the only thing that saved me!
Q
QuietVol· Jun 2026 ago
Rather than guessing a number, I'd sample it. Run the angle calc over your 5m window on historical data with no threshold at all, keep the raw values, and look at the distribution. Then you're picking a percentile, not a degree — you know roughly how often it fires before it ever fires. The awkward part is that the distribution isn't stable across regimes, so a percentile that gave you a handful of signals a day in a quiet stretch can flood you in a busy one. Did your 60s threshold correspond to a sensible tail on that window, or was it just a number you inherited?
H
HalfKelly· Jun 2026 ago· edited Aug 2026 ago
On the cap. I'd stop reading it as an exit and start reading it as a truncation. A one hour ceiling chops off the right tail of your holding time distribution, which means the big runners you're worried about losing were never in your expectancy to begin with. That's not automatically bad, it just changes the arithmetic: shorter, tighter outcome distribution, and you can generally carry that at a different size than a fat-tailed swing book. What gets people is keeping swing-sized positions on while running a momentum-sized exit rule. Agree with Pete on the fixed floor by the way, per-pair parameters multiply your degrees of freedom faster than your sample size grows.
Sign in to reply →
← All automated trading