Early elimination of poor crosses based on an objective criterion allows increased allocation of resources only to a few promising crosses for identifying superior recombinant inbred lines (RILs) for use as pure-line cultivars in self-pollinated crops. Early generation (F2:3) prediction of frequency of superior RILs that could be derived from advanced generations of crosses is one such criterion. We predicted the frequency of transgressive RILs from two horse gram crosses (namely HPKM 320 × CRIDA18-R and IC 361290 × Palem 1) for primary branches per plant, pods per plant, pod weight per plant and grain weight per plant based on mid parental value, additive genetic effects and additive genetic variance estimated from trait means of parents, and their F2 and F2:3 generations. The predicted frequency of RILs that transgressed better parent/two checks varied with the cross and the trait within a cross. The frequencies of transgressive RILs predicted from IC 361290 × Palem 1 were higher than those predicted from HPKM 320 × CRIDA 18-R for three of the four traits. As expected, the minimum population size required to recover the transgressive RILs predicted from IC 361290 × Palem 1 was relatively smaller than that from IC 361290 × Palem 1. Increased allocation of resources for handling segregating populations derived from IC 361290 × Palem 1 is expected to result in superior RILs for use as cultivars. We believe that the objective criterion used in our study is handy in identifying superior RILs in early segregating populations derived from a few promising crosses.