The new statistics for verification of global numerical models, distance, anomaly similarity and tendency similarity are introduced in functional way in this paper and takes the global characteristics into consideration. The effect of the three indexes for global numerical models is verified and the corresponding calculation method is given in the paper. The forecast results of global WRF model are taken as an example to evaluate its ability to predict weather system of medium-range. Results show as follows:the new statistics have explicit physical meanings, convenient for theoretical analysis and taking the characteristic of numerical model with spherical meridian grid into consideration. Thus, they have high accuracy and are universal. They can be easily used for verification of numerical models with output in spherical latitude and longitude grid. The calculation results of new statistics are compared with typical statistics (anomaly correlation coefficient and root-mean-square error, which are suggested by World Meteorological Organization). The calculations of new statistics are consistent with that of typical ones. The new statistics can also be used for the verification of vectors, like wind field. The calculation results of new statistics of global WRF show that the prediction of global geopotential height is usable in seven days and better in five days for the example in this article.