While I recognize the truth in the article, like you I found it incredibly counter-intuitive. I thought for sure that only choosing values of n within the range “i” hasn’t been assigned to yet would yield some sort of bias- Being that some cards wouldn’t be moved as many times as others (since whatever got moved to cards[i] in each iteration would stay there), I thought for sure that there would be some way to predict, based on the order of the starting deck, at least the probability of each card being closer to the front or the back of the deck.
Thanks for the explanation! I like when being wrong is so fascinating.