Dam deformation is characterized with periodicity, tendency, randomicity. So dam deformation can be regarded as sum of periodic deformation, tendency deformation, and random deformation, of which periodic and tendency deformation, gained by deterministic method, are researched sufficiently, but study on random deformation is inadequacy. Because dam random deformation at time t is mainly influenced by dam state at time t-1, little influenced at time t-2 or earlier, dam random deformation can be idealized as stochastic process described by Markov chain. In this paper, a new dam safety monitoring model is established using Markov chain, based on statistic model. First, regression analysis was applied to establish dam deformation statistic model, by which dam deterministic deformation can be obtained. Then dam random deformation component can be gained by subtracting deterministic deformation from total dam deformation. Consequently, by analysis of time series of random deformation, transition probability matrix of random deformation can be obtained and used to analyze dam deformation. Case study showed that this new model is superior to traditional statistic model, and can be applied to establish dam safety monitoring model.