Also I'm quite confused by the premise of that statistical nonsense above, is he basically saying that one should expect two candidates from the same party to recieve the exact same number of votes in two elections four years apart?
Not exactly. He's saying the exact 2016 result represents the average the candidate should receive in 2020. By average, I mean pretend you could run the 2020 election over and over again, average all of the results. Then he calculated the probability that the 2020 results could have happened if the average tends to the 2016 result.
Here's an analogy. Let's say you're 5'10. You postulate because of this that your siblings should have an average height of 5'10. But you have a sibling (only 1) who's 6'5 and conclude this is extremely unlikely because the average height of your siblings is supposed to be 5'10. So they must be adopted or something.