This study was designed to assess the reliability of grapevine leaf bioassays for predicting disease resistance on fruit in the field. The efficacy of various grapevine quantitative trait loci (QTLs) for conferring resistance to downy and powdery mildew was evaluated in bioassays and in a 2-year field experiment for downy mildew. The resistance genes studied were inherited from Muscadinia rotundifolia (Rpv1 and Run1) and from American Vitis species through cv. Regent (QTLRgP and QTLRgD). In bioassays, genotypes carrying Run1 blocked powdery mildew development at early stages. Genotypes combining Run1 with QTLRgP displayed no greater level of resistance. For downy mildew, genotypes carrying Rpv1 and/or QTLRgD were more resistant than the susceptible cv. Merlot, and showed a high level of leaf resistance in the field (<10% severity). Disease levels on bunches were much higher than those on leaves, with a high variability between Rpv1 genotypes (1–48%). A Bayesian decision theory framework predicted that an OIV-452 threshold of 5 in leaf bioassays allowed accurate selection of grapevine genotypes (P = 0·83) with satisfactory disease severity on bunches. Therefore, this study validates that the use of early bioassays on leaves, as currently performed by grapevine breeders, ensures a satisfactory level of resistance to downy mildew of bunches in the field.