Over the last decade, multiparental populations have become a mainstay of genetics research in diploid species. Our goal was to extend this paradigm to autotetraploids by creating computational tools for the analysis of connected F1 populations derived from a set of shared parents. In a companion paper, software to reconstruct F1 progeny in terms of parental haplotypes was described. For this study, we developed software for quantitative trait locus (QTL) mapping via Bayesian regression of phenotypes on the parental genotype probabilities. Statistical properties of the QTL model were explored by analyzing simulated half-diallel diploid and tetraploid populations with different population sizes, genome sizes, and numbers of parents. As expected, the LOD threshold needed to control the false positive rate increased with genome size, ploidy, and parents. Across the different scenarios, the number of progeny per parental haplotype (pph) largely determined the statistical power for QTL detection and the accuracy of the estimated haplotype effects. A QTL with heritability 0.1 was detected with 90% probability at 60 pph, while only 40 pph were needed to estimate the haplotypes with 90% accuracy. Our methodology includes a comprehensive treatment of dominance for multi-allelic QTL, which was illustrated by analyzing potato tuber shape in a 3 × 3 half-diallel population. A well-known QTL on chromosome 10 was detected, and the best-fit model included both additive and dominance effects. In terms of practical impacts on breeding, the software is already being used to select offspring based on the effect and dosage of particular haplotypes.