Haplotype blocks have recently been introduced formally as a simple population genomic pattern, together with an efficient combinatorial algorithm to detect maximal blocks in a set of phased genome sequences. By approximating the underlying haplotype structure at loci of interest without requiring model-based population genetic inference, they represent a quick and easy way to start understanding genetic variation. Here, I'll show how haplotype blocks can serve as the basis for fast, genome-wide detection of positive selection by adapting a previously published population genetic model. Besides its speed, the advantage of our approach is that it directly estimates a meaningful parameter, the selection coefficient. To illustrate the application and usefulness of detecting selection in general and our approach in particular, I'll discuss two usage modes: a candidate-based example, focussing on the European Lactase Persistence allele, and hypothesis-free genome-wide selection scans aiming at the discovery of new medically relevant variation. Finally, I'll mention potential extensions of our approach, most relevantly the problem of how to increase sensitivity to soft selective sweeps.