In general I'm just throwing around "x=slider*10"s and "y=slider*.0001"s for many global constants to see what happens. When starting from zero, I typically look at a sample of 64 randomly initialized behaviors at a time with ~10k particles per behavior. I'm trying to get something like this:
