I use Ninja Trader and apparently the walk forward optimization on Ninja Trader is bugged and it is not isolating the day, considering 2 days like one day and even 3 days like one day. So I cant really answer one question making walk forward optimizations... Well, the doubt is...
In the timeframe of 15 minutes which series of time is the best for optimization? For example, we can use 10 days to optimize each 1 day of the year (out of sample) and do that for the 355 days of the year. Or we can use 5 days (more randomness but more recent patterns) to optimize 1 day or we can even use like 60 days to optimze (less randomness).
The problem on Ninja is on the picture attached. You see, the day is duplicated, so 10 days optimization is best than 11, and 9 is best than 10 and so on. The day is duplicated so less days improves it wrongly. So the question cannot be answered.