Soil liquefaction is a common natural disaster during earthquakes. In recent years, the evaluation of soil liquefaction potential based on shear wave velocity has received increasing attention. In this paper, the maximum likelihood method is used to calibrate the worldwide used deterministic liquefaction evaluation model based on shear wave velocity proposed by Andrus and Stokoe. On this basis, a probabilistic evaluation model of soil liquefaction potential based on shear wave velocity is proposed. The results show that the liquefaction evaluation curves obtained from lognormal distribution, normal distribution, minimum Gumbel distribution and maximum Gumbel distribution have some differences when the liquefaction probability is 5% and 15%, and little difference when the liquefaction probability is 35%. The model correction coefficient calibrated by the minimum Gumbel model is mostly supported by the liquefaction case database according to the Bayesian information criterion. The mean value of the model correction coefficient is 0.879, indicating that the factor of safety obtained by the model proposed by Andrus and Stokoe is smaller than the true factor of safety in the average sense. The coefficient of variation is 0.387, indicating that the model error involves some uncertainties. The liquefaction evaluation curve of the probability model established in this paper is consistent with that obtained by Bayesian mapping function in the literature, and is quite different from that obtained based on logistic regression method in the high cyclic stress ratio range in the literature.