Accurate acquisition of dynamic load allowance (DLA) based on measurement data is essential to the safety assessment of a bridge. When static load tests cannot be achieved, and filtering fails, the estimated DLAs from the experimental method vary widely due to the choice of a left or right band. In this paper, the proposed weighted average method (WAM) is used to possibly solve the above problem in continuous gird bridges. Two-span and three-span precast concrete box-gird bridges were selected to optimize intercepted segments of WAM for the first time with the assistance of standard deviation and coefficient of variation in statistics. Then, a DLA measurement case of the negative bending moment was utilized to verify the validity of the WAM. The results show that the intercepted segments of 10L/16 to L were suitable for the WAM to calculate the DLA of the negative bending moment due to small offset moments and stable variation coefficients. The WAM had a strong anti-interference ability of outliers filtering in “bad data,” which differed significantly from the experimental method. In three measurements of a field bridge, DLAs obtained by the WAM had less dispersion than the experimental and low-pass filtering methods.