A class of estimators for the selective advantage, s, in a Wright-Fisher model with two alleles, variable population size, and genic selection is derived via martingale theory. Explicit expressions are given for these estimators which only involve simple computation. The optimal estimate among this class of estimators is obtained. Asymptotic results are readily established by an application of a martingale central limit theorem. The performance of this optimal estimator is compared to known estimators by means of a simulation study.