I notice in the collision detection routine that I'm using, constants such as epsilon and pi, as well as variables such as "hover above plane" and "push ball up", and the number of iterations to use for certain steps, appear to greatly effect the accuracy of the system (duh). The way I'm looking at it, there must be some grand ratio that all of these variables/constants should be at, in order to get the most accuracy out of the coldet routine, depending on major things such as unit scale, API, and OS, as well as the maximum velocity of particles, etc. Anyone have any pointers? [edited by - mattbbangin on February 27, 2004 9:28:34 AM]