Abstract: The similarity prediction method has been adopted extensively in short-rang climate prediction for many years, the best similarities of Chinese summer preci pitation have been analyzed and discovered that the anomaly correlation coeffici ent (ACC) of the best similarity of recent 23 years can achieve at 0.382, which is far higher than other methods, and can be regarded as the upper limit of simi larity prediction. Its prediction accuracy is about 70%, lower than the theoret ically limitation of seasonal precipitation forecast.The traditional similarity method has been extended to the extensive similar ity one, or called pan similarity method. The extensive one contains the both of plus and negative similarity. When the anomaly correlation coefficients of the former fields such as temperature, height of 500 hPa are negative, the predictio n should also be negative. The various results have been got through the various fields, and the last one is the mean of them.The experiments of recent 23 years indicate that the forecasting score (ACC) reaches 0.130,which is higher than the multiyear mean one. But the good effect cannot be got in every year, maybe only one similar criterion used. Soother criterions such as distance function can be used as another one to improve the pr ediction level. It is obvious that the method has good application value in active operatio n. As present forecast level is not high, the statistic forecast research should not be abandoned while developing the numerical model in the short-rang climate prediction.