It is currently possible to genotype cattle for markers for at least 50,000 single nucleotide polymorphisms (SNP). Estimates of SNP effects from genome-wide association studies can be used to estimate the breeding values of windows of adjacent markers and the variance of each window across animals can be used for QTL mapping. This study compared Bayesian methods for QTL mapping 10 traits in 1,048 Hereford beef cattle genotyped using the Illumina BovineSNP50 chip. Deregressed EBV were used as observations in a weighted analysis to estimate marker effects using BayesB and BayesC methods with Π (fraction of markers which assumes to have zero effects) equal to 0, 0.95 or 0.99, or BayesCΠ (concurrently estimating Π). The proportion of genetic variance and map position of windows that are QTL candidates were compared among these methods. The proportion of genetic variances for QTL windows were highest in BayesB due to less shrinkage of SNPs with large effects. The estimated proportion of genetic variance explained by any QTL increases with Π due to fewer markers being fitted. Map positions of candidate windows were similar for different methods except BayesC with Π=0 (GBLUP) which produced almost uniform genetic variances across windows. Among all methods, BayesB with Π=0.99 showed better performance to detect QTL windows (specially those explaining moderate genetic variance) and is recommended for QTL mapping. Results show 1 Mb windows are suitable to capture candidate QTL in beef cattle.