Information relative to haploblocks, i.e., haplotypes constructed based on linkage disequilibrium, can be used to potentially increase the accuracy of genomic evaluations. Jónás et al. (2016) developed an approach to optimally select SNP in order to reduce the frequency of under- or over- represented haplotype alleles. When applied to haploblocks in the Montbéliarde breed, this strategy led to substantially better validation results. In this study, we applied this strategy to 36 traits in the French Holstein breed, with a large reference population. Several scenarios for constructing haploblocks were compared with GBLUP and BayesCpi In contrast with previous results with smaller reference populations, the overall average correlations between predictions and phenotypes were within ±0.5 percentage point of the one obtained with GBLUP, whatever the scenarios (including the BayesC case), with very little variation across traits. Among the various haploblock construction strategies, the one in which haploblocks endpoints were defined based on a linkage disequilibrium measure on the HD chip gave the best results. Our implementation of a genomic evaluation based on haploblocks was too simplistic: it assumed that all haploblocks contributed to each trait and were all explaining the same fraction of the total genetic variance. Despite these naïve assumptions, it led to results comparable to the current most popular approaches. Any approach giving a higher variance to haploblocks actually carrying QTL should hopefully give better results. The idea of haploblocks remains appealing because it fits better the concepts of chromosome segment transmission and of maintenance of LD between SNP haplotypes and QTL. Keywords: genomic evaluation, dairy cattle, methods, haplotype, haploblock