In this paper, we present a new time-varying sliding surface to achieve fast and robust tracking of higher-order uncertain systems. The surface passes through an initial error, and afterwards, it moves towards a predetermined target surface by means of a variable named by sliding surface gain and its intercept. Specifically, the sliding surface gain is determined so that its initial value can minimize a shifting distance of the surface and that the system roots in sliding mode can follow certain stable trajectories. The designed sliding mode control forces the system errors to stay always on the proposed surface from the beginning. By this means, the system remains insensitive to system uncertainties and disturbances for the whole time. To illustrate the effectiveness of the proposed method, the comparative study with conventional time-invariant sliding mode control is performed.