The next-generation cellular networks are expected to be enabled by heterogeneous networks (HetNets). In HetNets, macro cell base stations (MBSs) and small cell base stations (SBSs) coexist to boost the capacity and improve the energy efficiency. However, the dense deployment of base stations (BSs) can significantly increase the energy consumption of networks. In this regard, BS's ON/OFF switching (alternatively termed as sleep mode) approaches are considered as a pioneering technique to save the energy of the networks. In this paper, we formulate the ON/OFF switching problem as a noncooperative game in satisfaction form to minimize the energy consumption while maintaining the quality of service. To solve the game, each BS utilizes an exploration approach, in which the BS selects its transmission strategy based on its strategy selection frequency in a distributed manner. The probability assigned to each strategy corresponds to the inverse of the times the BS has chosen its strategy. In this approach, if the BS is satisfied (i.e. its observed utility no less than a certain threshold), it has no incentive to change its strategy, otherwise it selects its strategy according to its probability distribution. Furthermore, the proposed approach is a low complexity algorithm, in which it needs to update only one element in the vector of strategy selection frequency according to the selected strategy. Simulation results show that the proposed scheme yields significant performance gains up to about 37% and 52% in terms of average energy consumption and average utility, respectively, compared to the benchmark mechanisms.