For earthquake hazard mitigation and performance-based earthquake engineering (PBEE), it is essential to predict probability of soil liquefaction of a site in a future seismic event. In this paper, a fully probabilistic method for liquefaction evaluation is proposed, in which both the uncertainties associated with the earthquake loading and the cyclic strength of soils are considered. To calculate the probability of liquefaction for a specific soil layer under a given earthquake, a reliability-based model is introduced, in which both the cyclic shear stress ratio and the cyclic resistance ratio of the soil are calculated based on the results of statistical regression analyses. In addition, the technique of Monte Carlo simulation is adopted to take into account the probable distribution of earthquake occurrence. This model is further extended to evaluating the probability of exceedance of a certain level of liquefaction severity (in terms of liquefaction potential index or post-liquefaction settlement).