In this paper, we propose an ordered selection technique to improve the performance of the decision-directed RLS adaptive channel tracking algorithm in 2x2 spatialmultiplexing MIMO-OFDM systems under high mobility. The proposed technique uses the decision of high reliability order during MIMO signal detection and then selectively activates the RLS adaptation only when less interference and error will be generated if a wrong decision is still made. Simulation results show that the channel tracking accuracy of the proposed approach outperforms the 2D-RLS algorithm and the conventional linear interpolation algorithm. Moreover, we also show that both ordering and selection are effective and necessary to greatly suppress the error propagation and coefficient misadjustment. With the proposed technique, the system performance has obvious SNR improvements.