Aspheric nonnull testing methods can achieve more flexible measurement of aspheres than the null testing ones. However, due to the violation of the null condition, retrace error exists and data reduction is difficult for this reason. In order to solve this problem, a reverse optimization data reduction approach based on theoretical simulation of the testing system is proposed. Simulation system is set according to the experimental apparatus. Surface figure errors are variables during the optimization procedure while the experimental wavefront is set to be the objective function. Aspheric surface figure error can be then reconstructed by fitting the solutions obtained from the optimization process. Simulation verification, experiment and error analysis are considered for this method. An experiment for testing a concave paraboloidal mirror with 101.0 mm clear aperture is taken. The result is very similar to that of null test (autocollimation method) with the peak-to-valley value and root-mean-square value errors less than λ/20 and λ/50, respectively.