Hello, I'm new here, I mean though I've been immersing myself in Quantopian for a few months now I feel like I've still just barely dipped my toes in.
I've been thinking -- It seems like with some algorithms there will be ideal thresholds and window lengths and such that will be contingent on factors such as avg. traded volume, market cap, volatility, beta, etc. of the individual stocks you're running through it. For example, a larger cap stock will move slower, and thus may need a longer window and tighter thresholds to effectively trigger something like a moving crossover signal. I figure there must be some systematic way to calculate these values rather than just throwing random numbers in and seeing how it pans out. Can anybody suggest what steps I need to take to run that kind of analysis? Or is this just a rabbit hole? I guess the next question is whether this kind of analysis would provide reliable data reflecting repeatable trends or whether it'd simply produce the ultimate overfit.
Just some thoughts -- but I don't know where to begin.
