Most of the optimization algorithms to solve the slope critical sliding surface have the disadvantages of complex structure, difficult to determine the parameter value, and poor optimization effect. This study introduced the estimation of the distribution algorithm based on the Gaussian distribution model, and combined with the sliding surface calculation and analysis model using the simplified Bishop method, to establish a new critical sliding surface search method with simple biological collaboration and competition ideas; secondly, a local search method for the 3-degrees of freedom was designed to compensate for the poor local search performance of the estimation of distribution algorithm. The standard and improvement methods were applied to the three calculation examples of increasing slope section complexity, respectively. The orthogonal experimental results from the standard method were validated by range analysis and multivariate analysis of variance, and the comparative analysis of the calculation of the standard algorithm and the improved algorithm was conducted. The results show that the standard estimation of distribution algorithm can be used to calculate the critical sliding surface of slopes. When the calculated case is simple, the control factors have limited influence on the calculated results; when it is complex, the population size has a significant influence. Compared to the standard algorithm, the improvement algorithm has better calculation and faster speed, and can effectively reduce the impact of the population size on the calculation. The preliminary verification shows that the model is more robust and has a broad application prospect. This study provides a new insight to explore the application of the distribution estimation algorithm in the slope critical sliding surface.