My general impression of the play at Karlsbad was quite poor, but the main flaws did not show up in the areas I expected. It is often said that the great growth of opening theory makes it hard to compare the chess of other ages with that of today, but I did not find this factor very important. It is true that there was no Sicilian Najdorf theory in 1911, but this is irrelevant as nobody played the Sicilian Najdorf. The range of openings played at Karlsbad was very narrow by today's standards. [...] The whole of ECO E was represented by just two games, nor was there a single game in the range B80-B99. The openings which were played had been developed theoretically, not to the same extent as today, of course, but enough so the players were not at a total loss.
He also concluded that the time control was not a factor. If not the openings and not the clock, what then? The words are Nunn's, the brackets are Watson's, the italics are mine:-
On the whole, the main deficiencies revealed at Karlsbad fell into three categories. The first was a tendency to make serious oversights. It is quite clear that the Karlsbad players were far more prone to severe errors than contemporary players. Even the leading players made fairly frequent blunders. Rubinstein, for example, who was then at virtually the peak of his career (1912 was his best year) failed to win with a clear extra rook against Tartakower. He also allowed a knight fork of king and rook in an ending against Kostic. [...] The second problem area was an inclination to adopt totally the wrong plan [examples follow]. The third main problem area was that of endgame play [horrendous examples of elementary blown endgames follow].
How exactly would chess960 return the royal game to the level of 1911? Taking Nunn's third point first, the endgame play in chess960 is exactly the same as in traditional chess. While there are some positional features that can occur in chess960 but never occur from the traditional setup -- a Bishop on a corner square blocked by an adjacent Pawn on the diagonal -- it is unlikely that they will endure into the endgame. Even if they do, they can be handled by the same general techniques that apply to all other endgames.
As for 'the wrong plan', Nunn's two examples are at move 20 and move 17, the point in a chess960 game where the position is looking very much like a game of traditional chess. In both examples, Nunn uses specific aspects of the position to determine a general course of play, a course contrary to the move selected in the actual game. This application of chess logic is no less valid in chess960 and any player capable of reasoning this way has a good chance of finding the right plan.
As for serious oversights, aka blunders, Nunn again gives two examples. The first leaves a piece en prise, while the second overlooks a two move tactical sequence. I can see modern masters making such mistakes in a blitz game, but not under standard time controls. Did the old timers calculate variations less effectively than modern players? So it would appear, but why?
So if it's not the opening, not the time control, not carelessness, not planning, and not endgame play, what is inherent to chess960 that puts chess back 100 years? The games from 1911 were played before the hypermoderns presented their case, before Nimzovich codified positional play, and before the Soviets adopted scientific methods of tackling a chess game. All of those evolutions apply just as equally to chess960 as to traditional chess. Chess960 doesn't invalidate them. It's not taking us into the past, it's taking us into the future.