Abstract: The azimuthal anisotropy of pre-stack P-wave is the most widely used method to detect fracture. This method mainly uses the long-axis and minor axis information of the ellipse fitted by the AVAZ seismic data to evaluate the fracture. When applying it in practice, it is considered that the variation of AVAZ response characteristics in the fractured reservoirs is just determined by the anisotropic parameters, and the influence of non-anisotropic parameters is not taken into account, which leads to the multi-solution of fracture evaluation. In order to improve the accuracy of this fracture evaluation technology, we started with the model analysis, the anisotropic parameters and lithological parameters were set to normal distribution functions, then the Monte Carlo stochastic method was used to carry out the pre-stack AVAZ forward modeling. Firstly, the influence of the anisotropic parameters on the ellipse fitting was simply analyzed. It is concluded that the anisotropic parameter γ has the greatest influence on the elliptic flatness B/A and the anisotropic factor B, the second is the parameter δ, and the parameter ε has the smallest influence. Then, the B and B/A response characteristics of the velocity and density models with different standard deviations were further analyzed. It is concluded that the change of velocity and density should be limited in a small range, in which the fracture evaluation technology is reliable. In addition, the variation of primary wave and shear wave velocity(especially the primary wave velocity) has greater influence on the pre-stack AVAZ response, and the density has almost no effect. Therefore, when applying this method to evaluate the fractured reservoirs, dramatic changes in the lateral lithology will increase the results' uncertainty.