Abstract: In order to make the reliability research of the fatigue fracture behavior in a very high cycle region, the very high cycle fatigue S-N test data of GCr15 bearing steel are estimated accurately. Its S-N curve relation is described by the Bansqun model and its probabilistic S-N curve model is created. The method relating statistics of extreme values (SEV) and Murakami equation are adopted to estimate the size of maximum inclusion and the corresponding fatigue strength of GCr15 bearing steel, respectively. Test results show that the very high cycle fatigue S-N test data can follow three-parameter Weibull distribution better. Their shape of probabilistic S-N curves all exhibit a continual descending trend. Based on the size of maximum inclusion which can follow the Gumbel distribution better, the size of maximum inclusion is predicted by the SEV exhibits linear ascending trend along with the augment of the volume of the steel. The result predicted by the Murakami equation explains the reason of probabilistic S-N curves exhibiting a continual descending trend.