I may have violated my position on avoiding misnomers in the last blog. The residuals there would more correctly called radial pair differences. With an estimated center for the circle one can compute a mean radius for the initial points and a better set of radial residuals as well as their rms error. The expected error in the mean radius would be εrms/√5=0.173. Half the radial errors would have a magnitude less than 0.512 pixels as might be expected when working with integers.