Ratings due to game scores

Having played some people with extremely low ratings and at the end of the game the computer scored them as playing at a very high rating in our game, why not average that out as well for a rating system of a different type? The only caveat I'd throw in is that they must be completed games with a winner and loser or a draw, time outs should not be included.

If I cared about ratings the fun would end because I would stop making the "I wonder what might happen" moves!!!

Its not a matter of caring, I'm going to play anyone who wants to, but I think that using the single game ratings system would be something to see, not even USCF could do that.

This exact question has been asked before, and it’s just because the computer estimated elo stat is very inaccurate, always guesses high, and isn’t nearly reliable enough for something like a rating system. It’s pretty much just a gimmick to make you feel good when it’s high, not very useful. Using it as the basis for elo would make it the rating system extremely chaotic, with people thousands of points above or below their true strength depending on whether they play simple games (which makes the estimated elo go ridiculously high) or complex games (which makes it go low). Also because of the simple fact that the current rating system works very well, if it ain’t broke don’t fix it, especially with a system that doesn’t really work at all.

As long as the system treats every game the same way I don't care, we would know going into it that it's an AI opinion no matter what. Consistently would be the only thing I care about if nothing is weighted to skew the rating it is an AI opinion nothing else. Another thing is if it does learn as it goes, it will improve as time goes on with every game.

It’s not ai, it’s just a computer. And it doesn’t treat every game the same, the computer can’t tell whether a game is complex or simple and therefore gives complex games much lower ratings. Also even if this were fixed, everyone’s ratings would be constantly rising, and extremely unstable. And again, the rating system right now works really well and chess.com isn’t going to spend tons of money and work making a new system that doesn’t work

You can either have an ai opinion or an actual objective rating, your objective rating will give you your actual score and pair you against people your level much better than a computers opinion

We have the actual objective, I'm not arguing to replace it, I'm saying something new could be offered.

I think it’s just too much effort for chess.com to make something that isn’t that useful

The 'estimated elo' thing is completely bogus to begin with. During speedruns from master players they start a game at 700 elo and play a perfect game and the 'estimated elo' comes to 1500. Well they are rated well over 2000 for real.

So the conclusion I would make is that the system just adds or subtracts numbers to the elo you started the game with based on whether you played well or not. It does not actually gauge what your elo is or should be.

Nor could it. Maybe it's good for young kids to get excited about when it tells them they played double what their elo actually is?

I don't see anything wrong with the current system. If you were really double your rating in skill you'll get there.

If the game is judged every number should be on that one game alone, then averaged out. If every game is regarded the same way each game's rating should be a stand-alone game, not this is how good you are. That would or should come from the established system provided, that everyone is playing are not cheating. It would be no different from the puzzle rating or some other method.