Genome-wide association studies (GWAS) have been widely used to discover the genetic basis of complex phenotypes. As such, GWAS provide starting points and potential opportunities for researchers to develop methods to predict complex traits. As a natural extension to GWAS, the polygenic risk score (PRS) is one of most popular methods for complex traits prediction in the genetics community. At the same time, other PRS extensions and alternatives have also been developed in the context of greater precision in prediction. To explore the advantages and disadvantages among various prediction methods under different genetic architectures, we leverage genotype data and phenotype data collected from surveys from a subset of over one million AncestryDNA customers who have consented to research. Our study shows that no specific method dominates across all traits, as each method has its merits in terms of model complexity and prediction accuracy under certain genetic architectures. While more in-depth research is still required, this work provides a sound foundation in the comparison of various statistical techniques for prediction.