This paper discusses various conflicting methods of estimating magnetic disc seeks when their files are processed randomly, pseudo-randomly and serially. Several common methods are proven to be significantly inaccurate and alternative methods are developed analytically.