To reveal the deviation angle changing pattern in axial flow compressors, an off-design deviation angle prediction model for full blade span was developed through statistical analyses and analytical derivations. By studying the compressor test recordings collected from NASA open technical reports, a linear correlation between incidence and an implicit function of deviation angle was found from 5% to 95% span. An analytical method was introduced to predict the linear correlation through derivations of the implicit function. The derived results show good agreement with the regression values of the experimental data. The off-design deviation angle is predicted by decomposition of the derived linear correlation at full span. An experimental database with 21 different blades is used to validate the prediction model. Span range from 5% to 95% and incidence range from −25° to 22° are covered by the 1494 test points in the database. The prediction errors are smaller than 6° for all test points, 4° for 99% points, and 2° for 90% points in the database. The current method was also applied to the off-design deviation angle estimation of a low speed four-stage research compressor at full span in the numerical design phase. It shows the deviation angle cloud maps acquired from off-design 3D Reynolds-averaged Navier–Stokes simulations are consistent with the results calculated by the prediction model.