Rating

oooooh, that is why i see so many closes cause of abuse, its not actual abuse, its just cheating and unfairness

Having multiple unauthorized accounts is abuse, using them to manipulate your rating is cheating. I've seen cases where the main account is closed for FPV and the others for abuse. It doesn't really matter in the end.

If you let the model take accuracy as an input in addition to the unknown base strength parameter, and track players' typical accuracy scores in games, you could still get a model that both solves backwards to give ratings updates (which would now depend on accuracy) and makes predictions of game outcomes when given ratings and typical accuracy scores.

However, doing this in a naive way would probably make the strength/ rating metric less useful, since you would probably mostly end up using accuracy to calculate the expected game outcomes, and rating would mostly become a metric of whatever accuracy didn't already account for- something like how bad your blunders were, perhaps (so someone who played perfectly 99.8% of the time but always lost a queen when blundering would get a low rating compared to a player who makes many small inaccuracies). Trying to fix that by messing with how accuracy is measured and giving something like "percentile of accuracy the player got compared to the distribution they normally get" as an input rather than accuracy directly could solve some of that, but then your score might go up less when you won especially accurate games, since the high accuracy made your result less surprising.

There are a dozen other things you could try past that point to keep the rating meaning intact, but it would be finicky, potentially prone to manipulation, and wouldn't ultimately add that much.

What the original poster meant was probably something more like "give a 'game performance' rating guess-the-Elo style with a tool like Maia, and update my rating towards that number as well as with the standard Elo/ Elo-alternative formula after a game." I actually don't see a problem with that general idea. It might cause ratings to converge faster as long as the Elo-guesser model was good, say if a player left the online scene (or just their account) for a time and gained a couple hundred rating points between chesscom games. It would make cheaters' ratings climb even more quickly, but chesscom already has systems in place for refunding points and such, so presumably any newly added potential for abuse from this change is something that they could handle.

Then again, there's the question of what the benefit is. Players whose ratings increased because of it might like it, but I doubt they would like it tanking their hard-earned rating after a particularly sloppy game.