I think we will find that certain Strategies will fit / work better with the short term re-optimise approach and some won’t.
Yes and the only way to test this is to do it manually. We need to basically pick multiple random segments of data (all of the same duration) and then optimise each data segment and see how it performed immediately afterwards with the same values. We then need to choose a different duration and do the same on every possible combination of data segments. Only then can we truly know whether large data segments and not re-optimising very often is better than small ones and often re-optimising. After all this work we may decide on a duration and so how often to re-optimise but this might just end up being a re-optimising curve fit on history!
It’s more manual work than I can stand and I’m not sure that we would learn much from it.
