Oil yield is a key trait for oil palm breeding in order to achieve two major objectives, meet the growing global vegetable oil demand and ensure the environmental sustainability of this farming. Oil yield throughout the life span of the plantation is a highly camp/ex trait affected by many phenological, physiological or morphological components that interact together. In the reciprocal recurrent selection (RRS) used by CiradPalmElit and its partners, diverse genetic populations are improved by selecting the best genitors within populations on one or few yield component traits. The objective of this study was to identify quantitative trait loci (QTL) for yield traits segregating in the breeding populations and use them to optimise the RRS process through a simulation/prediction approach. A Bayesian pedigree-based QTL mapping analysis was performed to analyse a multiparental population involving nine major founders of the breeding program. QTL were identified for ail yield traits, with co-localisation and linkage reflecting the relationships among the m. The identifies of QTL aile/es were predicted at the founder level and allowed to trace their segregation among the parents of a genetic trial of second generation planted ten years later and evaluated for the same yield traits. Using the QTL ejfect estimates from the first population and identifies of QTL allele of 89 parents tested in 113 progenies, the yield traits were predicted with low to strong accuracies depending on the trait considered Final/y, virtual genotypes were obtained by meiose simulations and their yield traits were predicted Thereby bath retrospective and exploratory analyses of the breeding program were conducted in order to estimate the gap with the potential selection gain for the first cycle, and to provide valuable information for orientingfuture oil palm breeding strategy.