Multiple input multiple output (MIMO) systems are shown to offer significant gains in error rate performance and capacity over single input single output (SISO) systems. To alleviate the resource requirement and complexity issues involved in the implementation of MIMO systems, receive switch and stay diversity combining is considered at the cost of modest performance degradation compared to full complexity systems. To the best of our knowledge, for the first time in the literature, the design and performance of MIMO systems employing switch and stay combining at the receiver with imperfect channel is considered. A minimal complexity training scheme suitable for both switching and channel estimation is described. From the pairwise error probability (PEP) upper bound derived, the optimal switching threshold as a function of SNR is obtained. It is shown that by using the optimal switching threshold full diversity is achievable. Power distribution between training and data space-time (ST) codewords is optimized to minimize the loss suffered due to channel estimation. Simulation results have also been included to corroborate our results.