A good point to start is Training a strategy in the manual.
Its not genetic for sure, and I have read something somewhere that basically zorro look for the best hill within the objective function results.
One recommendation dig into the manual. There are a lot of info even its not as easy find them sometimes.

Ciao