The based on probability theory of modern technology through computer's functions could be generated virtual data. Simulation research could catch on the possible results before conducting empirical research. But how much data should each variable produce when the test implementation of simulation research. And there is no a certain criterion. The purpose of this study was to explore the simulation research of experimental design, and the decision the suitability of operation times in the variable. The subject of the study was an appropriate set distribution theory for the scoring scale of sport skill tests. The generalizability theory (G-theory) procedure of research method was used to analyze the percentage of variables effect for normal distribution, logistic distribution, binomial distribution, and uniform distribution. The analysis model was (P: S) × I. First of all, generalizability study (G-study) estimated variance components, and then decision study (D-study) decided the suitable number of operations in the variables of simulation research. The results were: 1. G-study was obtained the average variance components of random error (pi:s), which was accounted for 98.44 %. It showed that the simulation research fitted random sampling theory. Secondly, the order of the sample size nested within sampling frequency (p:s) was accounted for 0.94 %, and the sampling frequency crossed with test length (si) was accounted for 0.39%, and test length (i) was accounted for 0.23%, and the sampling frequency (s) was approaching zero; 2. D-study: The three sample size were set for large sample (n=1,000), medium sample (n=300) and small sample (n=30). The two test length were set for short test (i=10) and long test (i=50). The sampling frequency was set for the five times of random sample. Conclusion: Using G-theory could decide the suitable operating frequency of variables in the experimental design of simulation research.