Im trying to construct bounding spheres for arbitrary polyhedrons, and so far, ive found 3 ways to do it: 1. Use the average of all the verticies as the center, and then loop through the verticies to find the longest radius. 2. Construct an AABB, and use its center as the center of the sphere, then find the longest radius as in #1. 3. Use Welzl's algo, which i dont fully understand yet (By the way, does anyone know of a place that has a clear explanation of it?). I found some code online that compares methods #1 and #3based on execution time and radius when ran on a set of random points, and modified the code to add in method #2 too. #2 turned out the fastest, followed by #1, with #3 being the slowest (taking about 3-4 times longer that #2). But #3 always gave a small radius, while #1 and #2 had roughly the same radii. So im wondering, how big of a radius-length advantage does Welvl's algo give over the other 2 methods? Is it worth the extra execution time? I plan on constructing the spheres on the fly every frame, since my objects change quite often.