In order to overcome the difficulty that the value of testing coefficient in traditional load test evaluation has been rough for a long time and cannot adapt to the current evaluation of bridge bearing capacity, and to improve the accuracy of theconstant value range of bridge testing coefficient, the measured value of testing coefficient of a large number of prestressed concrete small box girder is taken as the statistical sample.The testing coefficient of bridge bearing capacity evaluation was studied and optimized by using the probability statistical method. The selection principle of the statistical samples and the refining constant range of bridge testing coefficientwere put forward.The results show that the principle of sample selection determines the dispersion of test data. The range of the testing coefficient of different structural parts is different, and the value of the testing coefficient according to the structural parts can obviously improve the accuracy of the testing coefficient constant. The research can promote the development and perfection of the bridge load test.