Here is a game between two masters, White (2300) and Black (2600)
The computer scoring gives these masters -
White: 4 inaccuracies, 2 mistakes, 1 blunder, 37 average centipawn loss
Black: 1 inaccuracy, 0 mistakes, 1 blunders, 22 average centipawn loss
[...]
I'd call this a master-level game. What about you?
14.e4 was really a bummer. Instead 14.e3 would have secured a nice solid += advantage. If the computer thinks that the level of play in the compared games was similar, then I think it's a safe assumption that its judgement in this regard is not very reliable. I very much respect the old masters and their pioneers' work, but if someone showed me that Staunton game and said it were played by two modern masters, my guess would have been something between 3-minutes blitz or bullet.
The Wookie vs. droids scene! In that context, "upset" was translated as "soliviantar". Doesn't it sound lovely?