The present paper introduces a probabilistic failure model and aims at establishing a general assessment system to predict a decrease in reliability of composite materials due to cumulative damage. The failure model used in this system requires a good actuality and utility, i. e. : 1) Actuality-the model should represent more acurately the actual failure process of composites ; 2) Utility-the constructed model should give a good approximation to actual strength properties of composites when the program is run with currently used specimen sizes. In the previous third report, a statistical analysis on the dynamic failure process simulation model was proposed and compared with the static failure process simulation (1 st report). In this report, a statistical analysis is made based on the dynamic failure model and compared with the static one in terms of the different strength parameters. It shows that both static and dynamic simulation give little difference as far as the average values of tensile strength and maximum strain are concerned, but that the dynamic simulation gives a better estimation in terms of the dispersion of strength data and the actual failure pattern. A macro model is also proposed in this report, which is a new cumulative failure model composed of elements of micro models based on the previous dynamic failure process simulation models. A macro model is successfully used in estimating the statistical nature of strength properties of composite materials of actual size. It shows that the strength of composites depends clearly on the length but not on the width of a specimen.