In plant breeding, Multi-Environment Trials (METs) evaluate candidate genotypes across various conditions, which is financially costly due to extensive field testing. Sparse testing addresses this challenge by evaluating some genotypes in selected environments, allowing for a broader range of environments without significantly increasing costs. This approach integrates genomic information to adjust phenotypic data, leading to more accurate genetic effect estimations. Various sparse testing methods have been explored to optimize resource use. This study employed Incomplete Block Design (IBD) to allocate lines to environments, ensuring not all lines were tested in every environment. We compared IBD to Random line allocation, maintaining a consistent number of environments per line across both methods. The primary objective was to estimate grain yield performance of lines using Genomic Estimated Breeding Values (GEBVs) computed through six Genomic Best Linear Unbiased Predictor (GBLUP) methods. In the first five methods, missing values were predicted before cross-environment adjustment; in the sixth, adjustment was performed directly. Using the Bayesian GBLUP model, we analyzed genotype performance under both IBD and random allocation. Results indicate that computing GEBVs for a target population of environments (TPE) using available phenotype and marker data is effective for selection. The IBD method showed superior performance with less variability compared to random allocation. These findings suggest that using IBD designs can enhance selection accuracy and efficiency, and that pre-adjustment prediction of missing lines may not necessarily improve selection outcomes.