Abstract: Variety regional trial is an important stage in the process of crop variety registration, breeding and extension. The result of the yield is an important basis of variety registration. But the yield level of one or few sites was often obviously higher or lower than that of other sites in the national millet regional trials. In addition, the yield data of these sites, which accounts for a too high share in data summary, might decide the whole testing result. Therefore, how to reflect the actual yield level of tested varieties more objectively and accurately has restricted the work of regional trial and variety registration. In order to solve this problem, actual yield of every sites were replaced by theoretical yield in the sites by the author who engaged in national millet regional trials for many years. Firstly, the yields of the control of all sites were averaged. Secondly, according to the proportion of yield increase or decrease compared with the control in tested varieties of each site, the author counted the theoretical yield in the corresponding sites based on the data in first step. Thirdly, the author adjusted the yield level of all sites to the same level, then eliminated the influencing factors, such as land capacity and so on. It was verified that the proportion of yield increase or decrease compared with the control counted by this method was basically identical with the average proportion of yield increase or decrease in tested varieties of each site. This method could basically reflect the actual yield level of tested varieties. Great effect had been achieved and the method had great significance to the summary of the regional trials data.