Paris law constants, C and m, are two vital parameters of the Paris law for fatigue crack propagation of steel structures. Traditionally, the constants are determined through a linear regression of experimental crack growth curve under tension-to-tension cyclic loading. However, the direct linear regression would provide the value of C with larger error under negative stress ratios (tension-to-compression cyclic loading) probably caused by prestress. This study firstly proposed a reverse-reasoning methodology to obtain and calibrate Paris law constants based on the notched specimens with non– and prestressed CFRP (Carbon Fiber-reinforced Polymer) reinforcement under a wider range of stress ratios. Finite element (FE) modeling was then performed using a combination of the programs FRANC3D and ANSYS to compute the stress intensity factors (SIF) as well as crack tip shape evolution of the respective specimens. A sensitivity analysis was also made to investigate the variety of the constants affected by crucial factors. The comparison of experimental and theoretical results confirms excellent accuracy of the proposed method. The results indicated that the constant m is stable around 3.15 and prestress effect is significant not only on reduction of effective SIF range but also on change of the constant C. Moreover, the suggested value of m and C are 3.15 and 1.08E-12 for prestressed CFRP reinforced steel structures under negative stress ratio (R ≤ 0), and are 3.15 and 2.17E-13 for the positive stress ratio (R > 0) in the unit of mm and MPa, respectively.