Pre-stack seismic inversion can offer more geophysical information, but its accuracy can be affected by variable factors. For this reason, firstly, geological models of various velocity characteristics are designed, then, the pre-stack inversion is carried out for the seismic data extracted based on the geological model, and then, based on comparison of the inversion error to the theoretical model, influential factors of pre-stack seismic inversion such as initial constrained model, signal-noise ratio, inversion wavelet and reflectivity computational formula are analyzed. The study indicates that: the more accurate the initial constrained model, the better the inversion results; the higher the signal-noise ratio of the seismic data, the better the inversion results; the closer the wavelet matching with the seismic dominant frequency, the more accurate the inversion results; and Zeoppritz equation can be used as reflectivity computational formula because of its relatively reasonable applicability. The geological model and actual data inversion shows that inversion accuracy can be increased by establishing initial inversion model with elastic properties close to those of the actual model, boosting signal-noise ratio through filtering and modifying seismic data, extracting wavelet matching with the seismic dominant frequency from well logging and seismic data, and using applicable reflectivity computational formula.