I used it to pick the constants that minimized error in an approximation of inverse trigonometric functions (-π to +π). Found that it provided a good balance be

I used it to pick the constants that minimized error in an approximation of inverse trigonometric functions (-π to +π). Found that it provided a good balance between minimizing either the relative or absolute error.