Field trials to evaluate the performance of new varieties are an essential component of potato breeding. Besides the genetic differences, environmental factors can lead to variation in a trial. In variety trials, the observed differences amongst varieties should reflect genetic differences, without a large impact of the random or systematic variation in the field. One way to reduce within-field variation is to adjust the plot size and its shape in a trial. Two years of field trials in which individual plants in 90-plant plots of both diploid hybrid and tetraploid varieties were measured provided data to derive relationships between LSD% and plot size and shape. We provide a method to estimate the equations to calculate the expected variation when using different plot dimensions in a relatively homogeneous trial field for tuber yield, tuber volume, tuber count, tuber shape and the standard deviations of tuber volume and shape. Compared with the yield traits, the variation for tuber shape was relatively small. The effect of plot shape was minor. With these equations, breeders can determine what plot dimensions are needed to reach the desired precision for each trait.

Diploid hybrid potato breeding is emerging as an alternative to breeding tetraploid potato clones. The development of diploid breeding varieties involves recent, shallow pedigrees with a limited number of founders. Within this context, alternative QTL detection methodologies should be considered to enable identification of relevant QTLs and characterize the founders of the pedigree. To that end, we are using a dataset of multiple diploid potato $$\hbox {F}_3$$ F 3 families under selection derived by a cross between an inbred Solanum chacoense and an outbred diploid Solanum tuberosum, and identify QTLs for tuber fresh weight. We used three methods for QTL detection: (1) a Genome Wide Association Study model, (2) a linkage approach tailored to the population under study and (3) a more general approach for modelling multiallelic QTLs in complex pedigrees using identity-by-descent (IBD) probabilities. We show that all three approaches enable detection of QTLs in the population under study, but the method that makes better use of IBD information has a more direct and detailed interpretation by linking QTL alleles to the founders.