Genomic models that incorporate dense marker information have been widely used for the prediction of genomic values since it was first introduced (Meuwissen et al. 2001) , in order to select animals and plants in breeding programs (Dekkers 2007; Heffner et al. 2009; Jannink et al. 2010; de Roos et al. 2011) , or to predict the risk of diseases in humans (Wray et al. 2007; de los Campos et al. 2010) . Genomic prediction relies strongly on the existence of a relationship between individuals in the reference and validation populations. A way to validate a genomic prediction model is by performing a cross-validation, in which the phenotypes of the validation population are masked, and predicted genomic values are then correlated to the true phenotypes to obtain the accuracy. As linear mixed models are used for genomic prediction, the accuracy of prediction depends on the proportion of phenotypic variance explained by the SNP-genotypes. However, if the assumptions of the model are not met by the data provided, the prediction accuracy will be lower than the expected. We propose a statistical equation to predict genomic prediction accuracy . Our approach proposes a measure that correlates the discovery and validation populations based on their SNP-genotypes available and the model assumed for genomic prediction, that is then used to make inferences of the genomic prediction accuracy.