Neat implementation!
I am confused.
First, I think there must be a typo, since we want to use the CDF when dealing with flat data, right?
Given that and your code, I am reading that we should use \mathcal{L}(o_{n+1} | o_n) = F_A(o_{n}) instead of \mathcal{L}(o_{n+1} | o_n) = F_A(o_{n+1}) when o_n = o_{n+1}.
But if o_n = o_{n+1} then F_A(o_{n+1}) = F_A(o_{n}), so I must be misunderstanding something here 