How to implement new optimisation method for strategy testing?

 

Hi,


I see it is possible to run genetical based algorithm for selecting the "best" parameters, however as with all optimisations, it is a possibility that it will settle into a local extremum, instead of the global one.

Still to avoid running a complete multi dimensional scan, I learned a technique from side channel analysis, in which you scan along a multi dimensional grid, with low resolution, and at the points where you see promising combination, you will focus on that neighbourhood with a higher resolution in the next iteration.


How and where to implement this optimisation method so it can be selected from the list? (or not selected, but used anyway)