Network coding has been widely used in wireless networks to improve the system performance. In this direction, a retransmission mechanism with probabilistic network coding was proposed in previous work, where the certain value of probability is unknown. To overcome this drawback, we formulate an optimization model to detect the optimal probability for network coding. A Stackelberg game theoretic framework is employed to maximize the system performance, where the node asking for help is modeled as a buyer and the relay is modeled as a seller. According to the proposed approach, both sellers and buyers would be optimally motivated, while retransmission with network coding could be performed efficiently. Both simulations and theoretical analysis confirm the feasibility and effectiveness of the proposed approach.