In Africa, average sweetpotato storage root yields are low and breeding is considered to be an important factor in increasing production. The objectives of this study were to obtain variance component estimations for sweetpotato in this region of the world and then use these to determine the efficiency of variety trials and breeding systems. From an incomplete series of variety trials in Kenya and Uganda (comprising 15 genotypes, three locations, two crop durations and three seasons) variance components were estimated -using the REML method -for the following traits: storage root yield, biomass production, harvest index, storage root dry matter and the Elston index which was used to aggregate all four traits. The storage root yield of clones across all 12 environments ranged from 15.2 to 33.0 t/ha. The variance components were significant for all traits, except the genotype by season interactions for storage root dry matter. The efficiency of selection systems was determined for total test capacities of 450 and 900 plots and using not more than five locations. Two-stage selection was 17-81% more efficient than one-stage selection after two seasons of testing. Two-stage selection was optimized under the restriction of using at least two locations at step 1. Trials with two locations and one replication at step 1 and five locations and two replications at step 2 were appropriate both for selecting for storage root yield and for simultaneously selecting for storage root yield, biomass production, harvest index and storage root dry matter content by using the Elston index. There are indications that such an allocation of resources for breeding sweetpotato is also appropriate for other regions in Sub-Saharan Africa.