The aim of this thesis was to optimise the breeding program of Nile tilapia for a smallholder production system, thereby contributing to closing the yield gap. From Chapter 2 to 4, we generated information that can be used to optimize the current GIFT breeding program for Nile tilapia. Identifying the major factors contributing to the current yield gap in smallholder Nile tilapia farms is crucial and the first step to any intervention. In Chapter 2, based on meta-analysis of literature, dissolved oxygen was identified as one of the major environmental factors contributing to reduced growth and FCR. In the presence of substantial dissolved oxygen differences between the environments, genotype by environment (GxE) interaction can be expected. The output from Chapter 2 was used to design the GxE interaction experiment (Chapter 3). In Chapter 3, heritabilities for harvest weight (HW 0.18 and 0.23), thermal growth coefficient (TGC, 0.21 and 0.26), survival days (0.03 and 0.04) and size traits (0.08 to 0.45) were estimated, in an attempt to indicate that these traits can be improved by selective breeding. We estimated the genetic correlation for these traits between aerated and non-aerated ponds. The genetic correlations (rg) between the aerated and non-aerated ponds for harvest weight (0.80±0.30) and growth (0.81±0.27) indicate some GxE interactions. The less than unity genetic correlation implies that the genetic improvements gained in aerated ponds will not be fully expressed in non-aerated ponds. Therefore, breeding programs should consider GxE interaction. In Chapter 4, genetic parameters for log-transformed variance of deviations (LnVar), a good indicator of resilience, were estimated. Heritable genetic variation was found for LnVar and it was noted that LnVar is more expressed in the non-aerated pond compared to the aerated pond. Finally, in Chapter 5, we estimated genetic parameters for critical swimming speed (Ucrit) and the genetic correlation between Ucrit and harvest weight of fish raised in a non-aerated pond. Substantial heritable variation was found for Ucrit. The estimated genetic correlations between Ucrit early in life and HW after grow-out in non-aerated pond, and Ucrit and growth were negative, implying that Nile tilapia with higher Ucrit early in life had lower HW and growth later in life. In this Chapter 6, the findings from the previous chapters combined and discussed theirimplications for breeding programs