This paper designs and evaluates a vehicle-to-grid (V2G) electricity trader capable of selecting an appropriate subset out of a large number of electric vehicles (EVs) which want to sell their energy to a microgrid. A genetic algorithm, tailored for this trade coordination, reduces the amount of unmet demand forecasted one day advance in the microgrid. Each subset is encoded to an integer r vector whose element has either 1 or 0 according to whether the associated EV is included in the subset or not. The evaluation function estimates the fitness of a feasible solution, employing a fast heuristicbased unit scheduler. Its lightweight-ness allows the genetic algorithm to calculate the fitness of the massive number of feasible subsets, each of which has a fixed number of EVs. This admission test gives a chance for EVs to contact with other microgrids when they are not accepted to the final trade schedule. The performance measurement result obtained from a prototype implementation reveals that the proposed scheme achieves up to 20.8 % performance improvement over the random selection scheme in terms of unmet demand. Moreover, the proposed scheme can efficiently cope with overload condition, that is, many EVs are concentrated in a single microgrid, judging from its stable performance curve.