I try to optimize one year and then use optimized paremeter for th next year (H4 chart) results are negative. So for me so far it's nothing more then curve fitting
Dislikeda shift found by fitting it to only 3 months H1 data worked also during 4 other months. This should also work forward.
I just started a new forward test on EUR/USD H1, magic_shift found by fitting it from March until now is 192, positive correlation (long when positive, short when negative).
I will post updates whenever it opens a new trade:
AttachmentIgnored