Can quantitative genetics predict long term evolution? Can we ignore genetic details in order to quantify the response to selection after many generations?  Nick Barton and Harold P. de Vladar have formalized an analogy between statistical mechanics in physics and population genetics, providing a formal method to forecast the evolution of polygenic traits. The results are applied to study the evolution of genetic variability - a fundamental and debated issue in evolutionary genetics - without the need to know the frequency of alleles, but without disregarding their evolutionary effects, and allowing for arbitrary genetic, dominance, and epistatic effects. Harold de Vladar is extending it further to understand the rate of change of the G-matrix with pleiotropic effects.