In this work two complementary methods for detection of divergent selection between pairs of populations connected by migration are introduced. The new statistics do not require knowledge on the ancestral or derived allelic state and are robust to false positives. Additionally, it is not necessary to perform neutral simulations to obtain critical cut-off values for the identification of candidates. The utility of the methods is demonstrated in simulations. First method, called nvdFST, combines information from the haplotype patterns with inter-population differences in allelic frequency. Remarkably, this is not a FST outlier test because highest FSTs are not a priori selected to identify loci. On the contrary, candidate loci are chosen based upon haplotype allelic class metric and then the FST for these loci are estimated and compared to the overall FST. Evidence of divergent selection is concluded only when both the haplotype pattern and the FST value indicate that outcome. It is shown that powers ranging from 70-90% are achieved in many of the scenarios assayed while the false positive rate is controlled below the desired nominal level (γ = 0.05). Additionally, the method is also robust to demographic scenarios involving population bottleneck and expansion. The second method is developed for cases with independently segregating markers, where the information contained in the haplotypes vanishes. In this case, the power to detect selection is attained by developing a new FST extreme-outlier test based on a k-means clustering algorithm. Both kinds of strategies have been implemented in the program HacDivSel