Up to 80% of the automotive software can be generated from models. MATLAB Simulink is a common tool for creation of complex combinations of block diagrams and state machines, automated generation of executable code, and its deployment on a target ECU. The automotive safety standards require extensive testing of the developed models. Regression testing should be undertaken every time a model is updated to ensure that the modifications do not introduce new faults into the previously validated model. A common, time-consuming way is to rerun an entire test suite after even minor changes. This paper introduces a new method for automatic prioritization of test cases. The method is based on two principles: (i) A test case should stimulate an error in an updated block and (ii) the stimulated error should propagate to the place where it can be detected. The proposed method includes the evaluation of input vectors that are provided to updated blocks by each test case and a Markov-based stochastic error propagation analysis of the model. The application of the method is demonstrated with a Simulink model of a gearbox and a test-suite, automatically generated with the Reactis Tester.