One of the difficulties to incorporate the probabilistic methods for glass design is the differences between cumulative distribution function of the glass fracture stress. Typically, three different testing procedures are used: four-point bending, and coaxial double-ring tests, with small and large test surface areas. The direct comparison of the probability distribution function presents discrepancies related to the stress state - uni or biaxil- and, the behaviour of the plate - large or small displacements. The results of the tests (load - pressure - fracture probability) are adjusted with a finite element model, applying an interactive procedure to estimate the effective area for every specimen. The final parameters for fracture stress improve the estimation of the fracture probability of every sample with errors lower than 10%.