The lack of a fundamental theory of hysteresis is a major barrier to successful design of structures against deterioration Development of a practical method for identification and prediction of degradation is an important task. This paper has a two-fold objective. First, a robust identification algorithm will be devised to generate models of degradation of a structure from its experimental load-displacement traces. This algorithm will be based upon the generalized differential model of hysteresis and the theory of genetic evolution, streamlined through sensitivity analysis. Second, it will be validated by experimentation that a model of degradation obtained by identification can be used to predict the future performance of a structure. Through brute-force identification of hysteretic evolution or degradation, it becomes possible to assess, for the first time in analysis, the performance of a real-life structure that has previously been damaged.