Bike-sharing systems (BSSs) offer convenient transportation services with environmental and social benefits. However, they also bring operational complexity, with rebalancing bikes being a very challenging one. The importance and difficulty of building the reverse logistics for BSSs are evident from the data obtained from Mobike, one of the largest dockless bike-sharing platforms in China. This paper proposes an evolutionary game theoretic approach to identify the best bike-sharing sourcing strategies, including self-operation, joint venture, and outsourcing. We show that the self-operation mode is an evolutionary stable strategy to rebalance Mobike. However, if bike-sharing firms cannot achieve economies of scale and significantly reduce the variable costs of the self-operation mode, the outsourcing mode will become the optimal choice. From a long-term perspective, the joint venture mode is never an attractive strategy for the bike-sharing firms under study. Our model can select the optimal sourcing strategy to rebalance bikes in BSS to maximize profit when competition and cooperation are jointly considered.