[albntomat0]

This sounds like a frequentist vs Bayesian statistics discussion, which involves (this is a simplification by me, a non-expert in the area) different definitions of probability. The frequentist view is along the lines of rolling a 10 side die hundreds of times, recording the results, and determining that each side comes up equally. The Bayesian view is that the probability measures our certainty about some event. For example, take the hypothetical point in time where all ballots have been cast, but have not been counted. One could use polling data, etc to model the odds that a particular candidate has won. However, the frequentist approach doesn’t really make sense here, as the ground truth already exists (all ballots cast), so rerunning the the event doesn’t make sense.

Once again, I’m not an expert, so I recommend looking for additional explanations, if you’re interested.