Does the rating estimator factor in your actual rating?

Just wondering if the rating estimator in game reports somehow adjusts to use your actual rating as a "base".

Asking because I feel there are games I play horribly yet are rated as decent byt he rating estimator, and close to my level. Case in point: the following game. I looked at the game report and it says I played like an 1850....

Lol oh to clarify, I mean the chess.com rating estimator that appears in all game reports. As below:

Yes, it uses your actual rating as a base to determine your estimated rating.

I see, thanks!

Yeah, curious if anyone has input on this.

I don't know if it does, but it would be a pretty useless feature if it does. For eampe if an 800 rated player plas a brilliant game that, coincidentally, happened to also be played by a 2800 rated GM (I'm talking about the exact same moves), and analyses their game they could get an extremely high accuracy but the computer might say they only played like a 1200 rated player.

A high rated player also might play what they think is a terrible game, but what they think is a terrible game could be a missing tactic or miscalculating an endgame, but a bad game for a low rated player might be geting checkmate in 3 moves. This could be why high rated players are surprised when the game review says they played like a 1400 or 1500 when they think they should have played like a 600 or 700 rated player.

what happens for me is when me or my opponent make a few dumb mistakes with time trouble in the endgame the rating estimator subtracts about 500-700 points from our actual elo

The site hasn't released details, other than it does take the player ratings and accuracies into account.

It will give ratings on unrated players if you enter unrated games into the computer. For example i entered in Bobby Fischer game and without telling it was Fischer or Kasparov and it got some very high ratings and it didnt know it was Bobby Fischer or Kasparov and no rating was given to the computer of the players strength inputted in any way. The big question to me is ive gotten some high ratings in bullet games but is this the same formula is that same as the slow control ratings or if that is factored into the bullet rating and does the computer know to differeniate the ratings on the differnt time controls. In my test i ran bullet game at 2350 vs 2150 but without knowing the time controls it was 2250 vs 2000 so some adjustment is made in faster chess.

Interesting, thanks for sharing

That's a good point, but I'm not sure if that is how it works in practice....I've somehow gotten a 2400 evaluation vs 2050 for my opponent (for a ~1600 level rapid game, dunno how that happened) so I don't know that there is a huge upper cap on the rating evaluations even if it takes the user's rating into account. Interestingly enough, I used the pgn to create an "unrated" evaluation and the result was very similar -- 2500 against 2000.

Very interesting observation. Yes, it seems that it takes the time control into account. I tried rating estimation of one of my awful bullet blunderfests, and with the 1 minute time it gave an evaluation of 1200 vs 1150, whereas without the 1 minute time it gave an evaluation of 950 vs 950 LOL. I did include our ratings (1700+ on both sides) in both evaluations, so it seems that the estimator has no qualms in calling a horrific game what it is.

I used a spreadsheet with around 100 of my games analyzed based on accuracy, rating, and estimated rating. This is how it works based on what I see: First off, for a bit of clarity, the way it works has changed slightly since being released (I've noticed, since re-analyzing the same few games over a couple weeks). The time-control does have an effect on the estimated rating. For instance, consistently, when adding that it was a 1 minute game, the estimated rating changed from 2100 vs 2600 to 1950 vs 2550 (idk why it would go down, but it did, lol). Also, the player's ratings are also including in the algorithm. For instance, that same game (as a 1-minute game), should've shown as 1950 vs 2550, but showed instead as 1750 vs 2300 after realizing that we were ~1600-1650 rating. One interesting thing is the accuracy. The accuracy *shouldn't* take into account the time control, but somehow it does (slightly)? 2100 vs 2600 (without stating the time control) showed as 80.8 vs 92.7 accuracy, whereas the 1950 vs 2550 (after stating the time control) showed as 80.3 vs 92.7 (mine was the same, but the opponent's accuracy changed slightly).

Here's the key information: Each game is split up into the opening, middle game, and endgame, as shown right beneath the estimated rating. Each seems to hold the same weight. So, if the opening is really short, but you mess it up, then 5 moves will have the same weight as the next 20 (the middle game). Second: time control matters, but not a whole lot, whereas players' ratings are a relatively strong change to the estimated rating. Third: Every game has certain "key moments". These are moments where maybe there's a tactic, or a way to positionally capitalize on an inaccuracy by your opponent, etc. Each of these situations are stated as a certain rating difficulty, similar to puzzles. Naturally, higher rated players will have closer games and more advanced tactics/"puzzles" that they're solving within the game. If a game is very straightforward and pretty much every move is winning because your opponent blundered their queen, two bishops and their first-born child in the first 4 moves, then it won't give you a high estimated rating, even though you played perfectly, since the "positional puzzle ratings" throughout the game weren't ever that high.

That's all the information I've gathered thus far. I hope the information was helpful, and that it made sense

EDIT: with the "level of difficulty of the game" that I tried explaining, it then associated your accuracy based on those positions to see what percent of that "level-of-play" you actually achieved, for lack of better words.

Wow, thanks for sharing your detailed analysis!
Yes, now I understand why some of my games that my opponent just fell apart have low ratings, whereas I guess that more complex games can have higher ratings.

After Google Translating, I see that the above comment is this:
Take a game of Carlsen vs Hikaru. Change the Elo to 400 in the PGN. Analyze!
Challenge accepted!

Okay, it was recommended to change the names also (I actually learned some German, so I understood the comment above.)

Okay, I managed to analyze this game four times. All analyses changed the usernames to MC and HN.
With the original ratings, the estimation was 2500 vs 2650.
With white as 400 and black's original rating, the estimation was 2050 vs 2100.
With black as 400 and white's original rating, the estimation was 2100 vs 2250.
With both players as 400, the estimation was 1000 vs 1100. With the following text blurb:
"That game was pretty competitive. Both players had an amazing opening. It was an incredible middlegame by both players. Although White played a great endgame, Black did even better."
So....lol. Anyone want to interpret this?

yea i too played a 98.8 accuracy game and got estimation of only 1950 rating (original was 1200). My friend is 2150, he played a 80% accuracy game and got around 2100 as an estimation.

why this incorporates the player rating:

maybe to prevent c word allegations

 or because accuracy is also affected by player rating, so having this purely based on accuracy would be wrong

or maybe a really good game is rating + (something) where something might be based on the rating (to make the increases smaller for better players?)

and maybe it takes in an assumed rating for the unrated calculation??