Many outstanding library filtering tools are currently available for computational compound library filtering, however, only a few systematic, effect-specific screening methods exist. We have further improved the concept of Ligand Efficiency (LE, potency/size) by introducing completely size-independent statistical LEs produced from binding affinities obtained by molecular dockings of 1,090 FDA approved drugs to a diverse protein set. The statistical LEs incorporated the minimum and the standard deviation of the calculated binding free energy values, and their relationship to molecular weight was characterized by hyperbolic fittings. Our results indicate that drug molecules possessing the same effects tend to have similar statistical LE values and that this advantageous feature can be exploited to perform effect-specific library filtering. We have constructed filtering criteria for 165 effect categories which can primarily be used for drug repositioning, but can be further expanded to the search for new drug candidate molecules. Since the determination of statistical LEs needs a large set of docking processes per each drug molecule, linear models were created in order to estimate statistical LEs with easily determinable physicochemical and topological descriptors of the drugs calculated by the Calculator plugins of the ChemAxon JChem Base software v5.3.7. These models facilitate the generation of statistical LE values and represent the basis of a suitable tool for rapid effect-specific compound library filtering.