Abstract
The development of tropical maize populations with high heterosis potential is essential for sustaining genetic progress in hybrid breeding programs, yet accurate selection remains challenging due to genotype-phenotype interactions and inbreeding depression. This study evaluated the efficiency of five selection strategies in recurrent selection programs using F(2) populations derived from commercial maize hybrids: Smith-Hazel Index (SHI), Base Index (BIA), Mulamba-Mock Index (MMI), REML/BLUP for grain yield (BLUP_GY), and REML/BLUP for inbreeding depression (BLUP_ID). Consistency among methods was assessed with a heatmap, and predicted genetic gains were compared with realized field performance. Predicted gains were highest with MMI and BIA for grain yield and ear weight, although realized results revealed discrepancies, particularly for BLUP-based approaches. Notably, BLUP_GY, which had the lowest predicted yield (4025 kg ha(-1)), achieved a realized yield of 5620 kg ha(-1), surpassing BIA and SHI. This indicates that additive potential was underestimated in predictions, likely due to dominance and environmental effects in early F(2) cycles. Overall, BLUP-based methods proved effective in identifying progenies with higher additive value, and their integration with phenotypic indices is recommended to combine short-term realized gains with sustained genetic improvement.