To investigate the accuracy of genomic breeding values, different scenarios were defined by accounting for polygenic effects, a different number of quantitative trait loci (30, 90, 150), and three levels of heritability (0.15, 0.25, and 0.4). The Bayes B method was used to estimate marker effects. A historical population was simulated stochastically, which consisted of 100 animals at first 100 generations, then the population size gradually increased to 1000 animals during the next 100 generations. The animals in generation 201 with known genotypic and phenotypic records were assigned as the reference population, and animals of generation 202 were considered as the validation population. The genome was comprised of one chromosome with 100 cM length and 500 markers that were distributed through the genome randomly. Picking up the information that was not captured by linkage disequilibrium (LD), including polygenic effects in the predictions increased the accuracy of genomic evaluations. As the trait heritability went from 0.15 to 0.40, the average genomic accuracy increased from 0.48 to 0.64. An increment in the number of quantitative trait loci (NQTL) declined the accuracy of the Bayes B method. This study suggests that the highest accuracy (0.74) was achieved when additive genotypic effects were coded by a few quantitative trait loci and a lot of small effects included in the prediction of genomic breeding values.