This article presents a stochastic approach to predict fatigue crack propagation (FCP) diagrams of continuous crack growth (CCG) and discontinuous crack growth (DCG) in polycarbonate (PC) under cyclic loading. First, it is assumed that the macroscopic fatigue crack propagates stochastically. The transition probability is then expressed in conjunction with the craze fibril breakdown model for CCG. Second, the stochastic process is applied to DCG assuming that DCG occurs because of an unstable crack growth in the craze zone. A fracture criterion using a stress intensity factor is introduced for the unstable crack growth. As a result, we obtain an FCP diagram where the rate in CCG is lower than that in DCG. The stress intensity factor range for the DCG-CCG transition can be theoretically determined. Finally, to verify the present approach, the experimental data of DCG and CCG of PC are fitted to the Paris equation. In addition, the relationship between the DCG band size and the number of cycles required for DCG is predicted in order to compare it with the experiment data. POLYM. ENG. SCI., 2013. (c) 2013 Society of Plastics Engineers