“Reference corpus can be a mixture of different genres, and the bigger the better ......” (Scott,2016:112). With BaREnloB as the reference corpus, this paper makes a comparison of the grade trends of the epistemic modality sequences used in the argumentative essays in Written English Corpus of Chinese Learners 1.0(WECCL 1.0). Research findings reveal:1) whether through variance analysis or robustness test, all BIS,LOG_L and LOG_R basically show significant grade differences except for the analogue epistemic sequences. 2) Means plots of variance analysis suggest that, as to epistemic sequence, both BIS and LOG_L rise with grade and reach the peak in the third grade, and fall to the lowest in the fourth grade, while LOG_R goes up slowly with grade, reaching its peak in the fourth grade; As to the dynamic sequence, BIS and LOG_L are gradually decreasing with the rise of grade and there are reversed and high values in the second grade while LOG_R goes down in riptide and slack tide; Regarding the analogue epistemic sequences, there exists no significant grade difference although BIS, the LOG_L and LOG_R all remain basically flat and consistent as the grade rises, and sometimes fluctuates slightly, only to attain the best in the fourth grade. 3) The results of parametric and non-parametric test are fundamentally different, and some conclusions are completely opposite to each other.