The genetic breeding of Eucalyptus species from the subgenus Blakella, especially E. maculata and E. torelliana, is on the rise in Brazil. These species stand out for the quality of their wood and the possibility of taking advantage of the heterosis between them; however, they are still at the beginning of domestication. From the above, this study aims to: estimate genetic and phenotypic parameters; use this information to compare alternatives that can improve selection efficiency in progeny tests; and verify the viability of selecting the best individuals from each progeny in several locations. For this, DBH (cm) data were used from E. maculata progeny evaluation experiments conducted in two sites and E. torelliana in three sites, at six years of age, with 40 repetitions and of single-tree-plot. It was found that, for both species, the expected gain from selection was significant in all sites. In combined selection, the intensity of selection among the progenies was very mild (a large number of progenies being selected), even considering that the heritability on mean among the progenies was much higher than that obtained among individuals. The number of individuals selected per progeny was very different. This fact can reduce the effectiveness of recurrent selection in the medium term. Using clones as common controls is viable for selecting a single improved population based on the evaluation of progenies in more than one site.