Stockfish vs alpha zero: no opening book?

I keep seeing all over the internet that Stockfish didn't have access to its opening book during the match with AlphaZero. What is the source of this information? It doesn't say this in the paper. I think it was just a nasty rumor started by Nakamura misunderstanding things. 

Since the paper had the threads and hash size, it stands to reason that they would have included the opening book in the paper if it was used, since that is an important pieces. Also, the 1200 extra games played, 100 in each of 12 set openings, AZ actually lost some of the games, so it stands to reason that the lack of an opening book may have played a part in the initial set of games.

Haha that's very circumstantial evidence. There is evidence that points the other way; namely, Stockfish made different moves from the same position in different games. It would only do this if using a book; otherwise I would imagine the move choice would be deterministic.

That is not true. The engine will often evaluate a number of lines very closely to each other and depending on the exact clock times, the one minute time per move control, a different move very well could have been evaluated higher than a move played in a previous game in the exact same position. In fact, if a book was being used, it is very likely the engine would have made the exact same move each time the position was reached, since it won't evaluate until after leaving book.

no, there must be randomness in how the engine chooses moves from the book; otherwise it'd be easy for an opponent to know what it's going to do.

There may be some randomness if the opening book being used contains multiple choices in the same openings (not sure how common that actually is in a well tuned opening book). The fact that it chose different lines doesn't show it was using a book; it is much more likely that search depth and time resulted in different moves being chosen when the same position came up (if that is indeed the case, since only a few games were even released).

I don't think they would give stockfish an opening that's been refuted. Doesn't matter if everyone knows the opening stockfish is going to play it would still get a good position (unless it was given a bad opening but never a losing one). If someone went to Alpha zero and said "Stockfish is going to play the queens gambit" it wouldn't matter. Because the queens gambit is a sound opening the opponent could do thousands oh hours of research but white will always get a good position regardless because the engine knows what to play against any move the opponent will play in the opening. 

Each engine had 1 minute per move. It is very possible that the engine was evaluating the position and the move selected at the time was different due to CPU variances (the cycles won't be the same every single time). The opening is a place where standard engines don't evaluate very well to begin with (which is why they need opening books) and a minor variance one way or the other can influence move choice very easily.

https://arxiv.org/pdf/1712.01815

Thank you, and like the old saying... one paper is worth a thousand posts