I would be careful with the relationship between Fst and Hz. Fst explains the genetic diversity due to population sub-division and that can be framed in terms of Hz, but it is not, strictly speaking, taking Hz into account. See Allendorf and Seeb, 2000 for a good example. There are also a few examples of selection detection methods that add Hz to the mix (avoiding a self-cite here ;) )
More than saying anything, I strongly recommend reading Genetics in geographically structured populations: defining, estimating and interpreting FST Holsinger & Wier Nat Rev Gen
It is a complex subject, with lots of literature. The above paper is probably the anyone best hope to make sense of a complex topic...
But you are on to something: A recent paper used allelic differences to find a gene (EPAS1) under selection in high-altitude (Tibetan) populations: 2 caveats to think about: it did a pair-wise comparison of populations (i.e. only two populations) and used SNP data (i.e. bi-allelic). Have a think about how to expand simple allelic differences to many populations and many alleles/locus...
1) Allele frequency differences does not take heterozygosity into account.
2) It is a common measure where zero = panmixis; 1=fixation.
3) Fst weights the subpopulations.
If you think that using differences in allele frequency distribution is a better measure have a look at this paper:
They use and compare several methods for detecting selection. They explain the equations if you want to implement them yourself. I have been very happy with directional Fst and log10(theta_fu/theta_fu).