I am working on an analysis problem that I was hoping folks could provide feedback on.

I have a paired timeseries dataset (groups A and B) from which I can manually calculate how frequently group B exceeds group A. That is, if B > A, then we get a 1. If less, then 0 etc.

I was wondering how best to determine the frequency at which group B exceeds group a if the data were not paired (hypothetically). I tried plotting cumulative relative frequency of both groups to see where the two curves crossed. However, the percent time that group B is greater than group A does not match the manual calculation from paired analysis.

Any thoughts?