In two-tier femtocell networks, the frequency reuse among cells leads to cross-tier interference (CTI), the primary constraint to system performance improvement. In order to alleviate the impact of CTI, we propose a user-selected power control algorithm based on the non-cooperative game theory in consideration of user diversity gain. In the proposed scheme, the macro-user with high channel gain and the femto-user with low channel gain at the macro base station are scheduled to be “active” in one slot but to be “opposite” in next slot in order to achieve an optimal overall performance. Numerical results show that the proposed scheme is capable of outperforming the traditional stochastic schedule strategy by more than 30% improvement in terms of average macrocell capacity. Furthermore, an obvious performance gain in the percentage femtocells beyond its signal-to-interference-and-noise ratio (SINR) target can also be achieved by executing the macrocell link-quality-protection strategy.