In this paper, we will propose a practical method for improving the performance of a maximal predictability portfolio (MPP) model proposed by Lo and MacKinlay and later extended by the authors. We will employ an alternative version of MPP using absolute deviation instead of variance as a measure...