Large-scale renewable energy suppliers and electric vehicles (EVs) are expected to become dominated participants in future electricity market. In this paper, a competitive bidding strategy is formulated for wind power plants (WPPs) and EV aggregators in a pool-based day-ahead electricity market. A bilevel multi-agent based model is proposed to study their bidding behaviors, with market clearing completion in the lower level and revenue maximization in the upper level. A stochastic framework is developed to incorporate the uncertainties in maximal power production of WPPs and EV aggregators and bid prices of other participants. The process of bidding decision is formulated as a stochastic game with incomplete information, in which electricity suppliers including WPPs and EV aggregators are considered as players of the game, their lack of information in this stochastic market environment is counterbalanced by a multi-agent reinforcement learning (MARL) algorithm named win or learn fast policy hill climbing (WoLF-PHC) with maximizing their own profits by self-game. The feasibility and effectiveness of the proposed model and the WoLF-PHC solution approach are successfully illustrated using a modified IEEE 6-bus system and a modified 118-bus system with different numbers of market players.