Peer-to-peer （P2P） trading provides an important way for promoting the consumption of renewable energy by prosumers and the reform of power markets. The information transmission in P2P transactions relies on the cyber-physical system （CPS）, however, issues such as data bias, transmission delays, and prosumers' own information protection make it difficult for trading participants to accurately obtain the state information of other prosumers, thereby creating an incomplete information environment. In addition, the bounded rationality of prosumers intensifies the uncertainty of trading behavior. To address these issues, a stochastic game-based optimization method for prosumer P2P trading strategy considering incomplete information is proposed. Firstly, Harsanyi transformation is used to convert incomplete information into complete but imperfect information. The resulting transformation is then refined using prospect theory to better reflect prosumers' actual psychological preferences. Secondly, a stochastic game decision-making model for prosumer P2P trading is constructed. By incorporating the Markov decision process （MDP） framework into the stochastic game, behavioral uncertainty in trading is effectively reduced, enhancing the stability and effectiveness of trading strategies. Then，to mitigate the “curse of dimensionality” caused by the exponential growth of state spaces in the stochastic game, an adaptive state-tree pruning technique is proposed to significantly reduce computational complexity. Simulation results show that the proposed method effectively alleviates the impacts of incomplete information, reduces behavioral uncertainty, optimizes trading strategies, and improves overall economic performance.