Are you talking about optimizing the new time frame only, or optimizing all the data with the new time frame included?
If it's the new time frame only... well... then... opitimize it.
If it's only including the new time frame, you're insane to expect that the past week is going to provide breakthrough information when you started with 5, 10, or 20 years of data... I'd do it absolutely no more often than monthly... up to yearly, depending on how much effort you put into making sure that the optimization is rigorous (maximal statistical significance).