Stockfish would likely go 100-0 against Magnus, let alone any other human, in these early midgames that you believe machines "cannot accurately analyze." I mean relative to perfect chess your point there is accurate, relative to humans - it is not. This is how computer preparation has so radically changed the game. If one player walks into the prep of another player, they stand an extremely strong chance of getting a much worse position. Carlsen is the strongest player in the world yet he was completely lost before Fabiano had even left his opening prep in the game 8. The only reason he was able to survive is because Fabiano did not play particularly well (relative to the class of these players) once he stopped playing memorized moves.
This, then, sets the stage for what highest level chess is turning into. And that is becoming the avoidance of risk and the avoidance of the unknown, because if you enter into anything remotely risky or unknown your opponent may well have already looked at it with Stockfish and you're just going to get slaughtered. This is why things like the Berlin have all but entirely replaced things like the Najdorf. At the highest level, you can no longer afford to risk playing these interesting, dynamic, and risky type openings. Bc1/Bb2 in game 9 is a great example. It's not a particularly amazing idea but it entails minimal risk for white and, more importantly, was probably not analyzed by Caruana. Which it wasn't, yet the position was riskless enough that Fabiano to neutralize it without much trouble.
So I agree the Roshambo of who's memorized what is a somewhat interesting metagame on its own. But I think this metagame is starting to dominate the game itself. I'd rather see the players just play what they think is best without this nagging concern about whether the response they see is going to be human, or the start of a lengthy period of memorized computer analyses. And that's precisely what Fischer Random achieves.
This, then, sets the stage for what highest level chess is turning into. And that is becoming the avoidance of risk and the avoidance of the unknown, because if you enter into anything remotely risky or unknown your opponent may well have already looked at it with Stockfish and you're just going to get slaughtered. This is why things like the Berlin have all but entirely replaced things like the Najdorf. At the highest level, you can no longer afford to risk playing these interesting, dynamic, and risky type openings. Bc1/Bb2 in game 9 is a great example. It's not a particularly amazing idea but it entails minimal risk for white and, more importantly, was probably not analyzed by Caruana. Which it wasn't, yet the position was riskless enough that Fabiano to neutralize it without much trouble.
So I agree the Roshambo of who's memorized what is a somewhat interesting metagame on its own. But I think this metagame is starting to dominate the game itself. I'd rather see the players just play what they think is best without this nagging concern about whether the response they see is going to be human, or the start of a lengthy period of memorized computer analyses. And that's precisely what Fischer Random achieves.