For multiple-input/multiple-output (MIMO) wireless communications systems employing spatial multiplexing at the transmitter, we have recently been studying the advantages and disadvantages of genetic algorithm (GA)-based detection vs. the maximum-likelihood (ML) detection and linear detection, for various channel fading assumptions, e.g., Rayleigh and Rician fading, fixed and random azimuth spread (AS) and Rician K-factor, and various ranks of the channel matrix mean. In this paper, we step away from comparing their performance and complexity and focus instead on the selection of GA parameters, such as population size, P, generation number G, and mutation probability, pm, for the GA in MIMO detection. Thus, we employ a meta (or outer) GA to optimize P, G, and pm values for the inner GA employed for MIMO detection. The empirical distributions of the selected parameter values are then compared for various channel fading assumptions. It is found that the optimum GA parameter values for MIMO detection are directly affected by fading type, AS and K distribution, and especially by the rank of the channel matrix mean. The meta-GA approach helps reveal that the parameters of the inner GA should be tuned in order to achieve maximum performance for the lowest numerical complexity. Future work will seek efficient methods.