In line-of-sight (LOS) environments, the signal detection of multiple-input multiple-output (MIMO) systems becomes difficult because the correlation among channels becomes higher when the phase differences of direct paths are close to each other. In this paper, we propose a MIMO system with precoding to maximize the minimum distance (denoted hereafter as Max-dmin precoding) to improve the performance of MIMO systems in LOS environments. Max-dmin precoding can improve the detection performance of maximum likelihood detection (MLD) by maximizing the minimum Euclidean distance between symbol points at the receiver side, but the design becomes complicated as the number of substreams increases. The design of our proposed Max-dmin precoding is easy by using the property that the first eigenvalue of channel matrix becomes dominant in LOS environments even when the number of substreams is large. We use long-term average channel state information (CSI) instead of instantaneous CSI for precoding. Moreover, we propose a reduced-complexity MLD that uses the ordered reliability of symbols that results from Max-dmin precoding. From the results of computer simulation, we show that our proposed scheme with reduced-complexity MLD can achieve low complexity and good bit error rate (BER) performance in LOS environments with Rician factor K = 5 dB.