How good would your accuracy have to be to be able to compete with 1900‘s?

Accuracy is already curved at each rating level so that is an impossible question.

@OP

On top of everything else, every game is different. There will be sharp games where accuracy will naturally be lower because in some positions one or 2 moves will keep you in the game.

On the other hand there will be games where pieces are traded, there is an endgame and many moves are good enough.

In the first game you might get 75 accuracy and still win, while it is entirely possible to have 90 accuracy in the second game and for it to be a draw for instance.

One player might play sharper stuff and have lower accuracy as a result but as long as he gets the better of his opponent he might be successful.

The other might play in a more safe manner and not be as good, even though the average accuracy in his games is higher than the aforementioned person.

Along with that, it is easier to have a good accuracy against a weaker opponent. If an opponent blunders a piece without any compensation, it is easier to play the rest of the game, so the accuracy will be higher.

On the other hand, if you play against a better player he will put pressure on you, making your decisions more difficult which will lead to lower accuracy.

In short, the accuracy can't really be viewed in a simplistic way. Perhaps on average across many different players one can draw some conclusions across various ratings, but it will vary from case to case for sure.

https://support.chess.com/en/articles/8708970-how-is-accuracy-in-analysis-determined

CAPS and CAPS2
Chess.com’s Accuracy score is now powered by 'CAPS2,' an improved version of the original Chess.com 'CAPS' (Computer Accuracy Precision Score) algorithm.

Moves are still compared against the top engine recommendations, but the math on how these are calculated has changed. Why? Well, most chess players - even low-rated ones - make a lot of the best moves!

Previously, CAPS (v1) looked to create a 0-100 band within the normal human player range. So, scoring perhaps 40% 'Best' moves, which is very low, was equal to single digits on CAPS. And, scoring a very high number of 'Best' moves, but not a perfect game, was often rated 99.9 on CAPS, even though it wasn’t played perfectly.

This made some people feel bad (on the low end), and led to a lot of cheating allegations (on the high end). The new Accuracy scores, based on CAPS2, replicate the feeling of being graded on a test in school.

You will notice that the majority of scores now fall mostly between 50 and 95, which provides a more intuitive understanding of how accurately you played in your game.

When I win a game, my accuracy generally sticks around the 60-70% range. When I lose, it's not uncommon for it to be >80% (unless it was a short game). This is generally normal for a lot of games.

I've also had my fair share of games where I won despite having the lower accuracy. The most extreme example was when my opponent had 54% accuracy and I won with 42% accuracy because he blundered mate in 1 by taking a "free" rook 💀

There is a fundamental problem with this question, which is that your average accuracy score is always going to depend somewhat on your opponents moves and how hard they are to respond to as well. For example I, at 1800, will have consistently very high accuracy against a 1400, while consistently low accuracy against a 2200. This is not because I necessarily am playing worse or better, but because the 2200 will pose me with more difficult moves to respond to, and more critical positions to play, so the potential for accuracy loss is just greater.