In evolutionary game theory, cooperation is beneficial to increasing collective revenues. Presently, although the cooperation level has been raised in abundant studies, the correlation between theory and practice is ignored, while the update time of the strategy is restricted to one point of earnings. Therefore, we introduce an observation mechanism to investigate the average return of the agent during the observation period, including four observation models: complete synchronization, relative synchronization, Gaussian asynchronization and exponential asynchronization. Subsequently, we employ the Monte Carlo method for the experimental simulation and analyze its effect on the cooperation level of prisoner's dilemma game. The results suggest that there is a time-observing threshold in every model and the cooperation level is optimum at the moment; the highest level of cooperation is realized in the exponential asynchronous model; the cooperation level in the Gaussian asynchronous model is positively correlated with the square deviation. Meanwhile, the heterogeneity between different agents also produces a positive influence on the cooperation. This work is conducive to eradicating the limitations of benefits in traditional research and promoting the improvement of the cooperation level. Some theoretical basis are provided for supporting social group cooperation.