Generalised predictive control (GPC) has many attributes which make it an effective practical tool for the control of scalar systems. The method is easy to understand/implement and has won popularity among many practicising engineers. Despite its success it does not have a general stability and robust stability theory. The paper proposes a new algorithm which deploys most of the ideas of GPC but yields control configurations with guaranteed stability and provides a systematic means of optimising robust stability margins. It achieves this through the use of a particular stabilising feedback controller which replaces the relevant transfer function operators by simple polynomial operators. This simplification also reduces the computational burden of a computer implementation of the method. The superiority of the results of the paper are illustrated by means of numerical examples.