Otherwise, you need to tweak it when using contender_duration so that the baseline series aren't incomparable to the control and experimental ones.