According to the data of radar cross section (RCS) under different polarization modes difficult to accurately measure aircraft stealth performance in each angle domain, an evaluation method based on the moving-polynomial smoother algorithm and detection probability model is proposed. Firstly, the circumferential RCS data of HH polarization and VV polarization is treated smoothly to decline relative error of RCS polarization by setting smooth window. Then, radar detection probability and detection probability error of smooth data are calculated based on the single-hit detection and the incoherent integration detection. Combined with two kinds of error conditions and according to the principle that the best moving-polynomial smoother is that relative error of RCS polarization is a minimum with the undistorted raw data information, and the optimal incoherent integration detection effect is that the error information of detection probability in each angle domain is reflected as much as possible. The optimal combination solution is made up of the smooth window and incoherent integration pulse is defined. Finally, the optimal simulation results of the smooth RCS data and detection probability are obtained by using traverse comparison, and then aircraft stealth performance is analyzed quantitatively and evaluated effectively in each angle domain. Simulation results show that incoherent integration detection probability of the smooth RCS data could evaluate aircraft stealth performance in key angle domain, and the assessment effect is the best when the combination solution is (10, 10).