In the real world is the success set not known. One has to 'approach' this unknown set by 'searching' it. If we assume the genotype as some n-dimensional coordinates in a n-dimensional space then are the values of the real evaluation function some additional space which includes the world as a parameter. Because the real world can change through time we have to assume that the 'ideal' success set is not necessarily a 'fixed' set; the ideal success set can change too! For some finiteduration one can assume that the success set is stable and the evaluation function is then an empirical estimate of the evaluation function .