In this work, a new method of prediction of China summer precipitation was suggested. The new ensemble method is a combination of singular spectrum analysismaximum entropy method (SSA MEM) and optimal subset regression (OSR), in which the inner characteristics of the series and influence of outer forcing factors are both calculated. The hindcast of the summer precipitation during period of 1961-2000, and the prediction of 2001-2004 are made separately. Results show that the anomaly correlation between the hindcast and the observed of precipitation for 160 stations reaches 0.85, passing through significance level of 0.01.The accuracy of summer prediction is obviously over 69%, which is slightly better than climate prediction on average. It indicates that the performance of the method suggested in this paper is well on the prediction of summer precipitation and is especially well over Northeast and Southwest China, which stands for its application into prediction practice.