Sometimes we tend to forget that:
Whatever the specific approach, competent, scalable genetic algorithm performance can only be obtained when building block linkage information is already in the problem coding or is identiﬁed before or during the genetic algorithm’s search. The scalability of genetic algorithms depends critically on the representation. Once the appropriate schemata are identiﬁed, the problem becomes “mixing-easy”, and fast, reliable, scalable evolutionary computation can be achieved.
Dirk Thierens, Scalability of Simple Genetic Algorithms, Evolutionary Computation, 331-352, 1999
And we shouldn’t…