Testing methodology for the Chinese quantum satellite link?

Does anyone who understands quantum networking better than I do have an opinion on the testing methodology that the Chinese team used to confirm entanglement? I guess, more specifically, my question is: when they say that they got 911 positive results out of “millions” of attempts, does this significantly exceed any expected false-positive rate for the confirmation methodology? If so, by what margin? Obviously, if you were just flipping coins, and measured the results once, you’d get 50% positive correlation, twice and you’d get 25% correlation, ten times and you’d get 0.1% correlation, and you’d be at 911 out of a million. So, how much better than that are we talking about? -Bill