Yes the engine is rendering an eval for a position many moves deep, but it still has to reach this position one move at a time, so I think my point still remains. Its search and eval function weren't tuned with that in mind.
As for filling in the sandwich, sure, I agree, but I'm not sure a 20 move opening book would be good. Since these engines are so strong I was thinking more along the lines of 5 moves of book, 10 at most, then let the engine makes decisions from there. If it's inadvertently getting a passive middlegame all the time, then I think that would skew the results.
As for engine-made opening books, that's more like cheating to me. Basically as if you're giving it more time to think.
Indeed. Well actually, it should combine all top engine games (hundreds of thousands of games, each involving substantial computing time. Centuries of computing time. So it's getting assistance with a huge computational task for move 1. But how is is better to use GM games? Because they are weaker?
As for the endgame, yeah an EGTB is like cheating, so I'd be willing to say no EGTB.
By the way, I've beaten SF at full strength, no opening book, when I make it give me knight odds and it plays into a French exchange type mid game. We just trade pieces and I have an easily winning endgame.
Maybe this experience makes me biased, but I think no opening book could really skew the results.