Based on the grain yield data of regional trials with 233 winter- and spring wheat cultivars (lines) in rainfed farmlands at 82 locations in four subregions of China in 2003-2009, this paper studied the experiment precision (EP), variety comparison precision (VCP), and testing-site discrimination ability and representativeness of national regional trials, and comprehensively evaluated the trial environment. The results showed that in one-location-one-year experiments, the average coefficient of variation (CV) and the relative least significant difference (RLSD) were 6.1% and 10.5%, respectively, and in multi-location-one-year experiments, the CV was all within 8.2%, and the CV and RLSD were mostly well controlled, indicating that the trials had a high precision. The testing-site discrimination ability was the highest in the northwest spring wheat subregion, but showed less difference in the other subregions. The testing-site representativeness was the best in the northeast spring wheat subregion, and the worst in the northwest spring wheat subregion. On the basis of the comprehensive consideration of the testing-site discrimination ability and representativeness, and by using the parameters of the environmental comprehensive assessment (r(g)h) of GGE model, it was shown that the proportion of the ideal trial locations for wheat in our rainfed farmlands was only 32.4%. Among the wheat production regions, the proportions of the ideal trial locations were in the order of northwest spring wheat subregion (40.9%) > northeast spring wheat subregion (33.3%) > Huang-Huai winter wheat subregion (30.4%) > north winter wheat subregion