Speed with filters and without filters

 If I let the software run without any filters  it will generate thousands of strategies in a matter of a couple hours.  I can then go through manual retesting and filtering like we did under SQ3 to try and arrive at the result I want and maybe have 100 or 200 strategies left over.  Both steps would take for example four or five hours  


 To the contrary, if I set many filters to achieve the same desired results as the first scenario, after four or five hours I may only have 10 strategies. 


 Why is this and can it be improved? 


 In addition, based on how this fitness ranking influences the logic, I am not sure that when generation results become so slow that the work is really efficient.