Hi. I only agree with the part about ratings getting more accurate the higher the percentile a player gets. I believe there's a simple mathematical reason for that: If a site like this starts people at 1200 & they play, let's say 4 games, & their rating based on 4 games is 1420 for instance it won't be that accurate yet anyway because they started off with a 1200 rating & they haven't played enough games to even get an accurate average. In real OTB tournaments there is no artificial, just made up, rating to start people at. It's based purely on what your win % is & the average ratings of all you're opponents. Also, & this is more important, the higher ratings were obtained by playing more games so it has to be more accurate. Example, if a player is 2300 strength he's not going to have his rating reflect his true playing level with only a few games played because he was started of at an artificially created 1200 or whatever made up rating the rating system starts him at. But.... I could be wrong!
Chess.com ratings are inflated
Maybe its different in the crazy world of the USA but over here you start not at "101" but at unrated and are then assessed based on actual results. On chess.com there were not immediately a pool of people with ratings starting out as chess.com is only a small number of years old, so they needed to pick a figure and 1200 is not unreasonable. With over the board chess baselines exist and you can start new players as unrated as there are plenty of people to earn rating points from and the system is basically stable. Also note that the rating adjustment from their arbitrary 1200 baseline is far greater for losses and wins than it would be from someone starting over the board. If memory serves me right, the first win/loss was worth about 133 either way on chess.com so patzers go down fast.
I agree with Estragon & Here_Is_Plenty, they said it better than I did. I had started drinking already..... sigh..... Same idea but more to the point. Good input.
Chess.com ratings are higher than OTB ratings in general. That doesn't mean they are inflated.
I find it funny how the people who complain the most about the online ratings being inflated are usually people with high blitz ratings but mediocre online ratings.
In any case, you can't compare ratings across different player pools (e.g. online chess.com and USCF or FIDE OTB). Ratings only have meaning within their own player pools. How many times do people have to say this?
kenpo,
I'm sure your feeling about "something fishy" going on is fully justified. Still, it seems to me a lot of people with high blitz ratings but mediocre online ratings immediately jump to the conclusion that someone with a high (but not stratospherically high) online rating, but mediocre blitz rating, must be cheating, or there must be something flawed in the rating system. It never occurs to them to consider the possibility that correspondence is just a different game than otb or blitz, and maybe they're just not very good at it.
Another overdiscussed topic.
Again, as pointed by several people here and before, there is no need of correlation between your OTB rating, your online turn-based rating, your online live chess rating and your snail mail correspondence chess rating.
AcvilizedGentleman said: "Try to name a single player with really high blitz but really low online (who is also unable to climb further)"
Just look up the OP.
"I find it funny how the people that try to call the ratings accurate are people who cant even break 1200 in live."
Ah, but the difference is, I don't go around making excuses for my crappy blitz play. I don't go around saying, "Blitz ratings are inflated". I accept the fact that blitz is a different game than correspondence, that I'm a crappy blitz player, and that my blitz rating reflects this fact. I don't whine about how my crappy blitz rating is a result of cheating or flawed rating systems.
AcivilizedGentleman said: "Honestly though, there is no way a true 1800er..."
There's no such thing as a "true 1800er". The ratings have no intrinsic meaning, only the difference between ratings.
There is such a thing as ratings inflation and deflation, but that is not what most people are complaining about. True ratings inflation means the average rating of all players increases over time. It's a comparison of ratings in the same player pool measured at different points in time.
Most people seem to really be complaining about one of two things:
1. The average rating of online ratings is too high. But since only the difference between ratings matters, this complaint is unfounded.
2. The online ratings distribution is too skewed, presumably because many at the very top (> 2200, say) are using illegal computer assistance. This may or may not be true, (I think it probably is true), but it still would not have any affect on the difference in ratings between any two particular players who are not using illegal computer assistance.
The only way an individual's rating can be "inflated" is if they are either cheating or have a provisional rating.
2. The online ratings distribution is too skewed, presumably because many at the very top (> 2200, say) are using illegal computer assistance. This may or may not be true, (I think it probably is true), but it still would not have any affect on the difference in ratings between any two particular players who are not using illegal computer assistance.
Actually, it would have an effect. Ratings are a Bayesian estimator based on an assumed prior distribution of player performance. This prior distribution is used in determining your expected score based on the difference in ratings. If the actual distribution of player performance varies from the prior distribution then the expected score will be incorrect, and that will make the ratings adjustment based on the actual score incorrect. I doubt this makes a practical difference on Chess.com, but you'd have to look at the actual distribution to determine that.
Also, if there is no practical difference between the distributions, then it won't have an effect on any two particular players whether or not they're using computers. It's not that the computer screws up the ratings, it's that somebody is claiming that the computer's (correctly calculated) rating is their own.
It's true that normally chess.com rating is 100-300 higher than the FIDE rating based on few real experiences. Though some players also claimed that their OTB rating is higher than their chessdotcom rating.
My conclusion is, "Who cares?"
ironic_begar said: "Actually, it would have an effect. Ratings are a Bayesian estimator based on an assumed prior distribution of player performance. This prior distribution is used in determining your expected score based on the difference in ratings. If the actual distribution of player performance varies from the prior distribution then the expected score will be incorrect, and that will make the ratings adjustment based on the actual score incorrect"
I believe the prior distribution used is a distribution of an individual player's potential strengths, not a distribution of the ratings across all players. (Glickman compares it to each player having a box with slips of numbers representing potential strengths, and then a game is two players pulling slips of paper at random.) So in fact the relative skewness of the distribution of ratings across all players would not affect the computation of updated ratings.
"It's not that the computer screws up the ratings, it's that somebody is claiming that the computer's (correctly calculated) rating is their own."
Exactly. And this is why I get so upset about this. When people complain about "ratings inflation", a lot of the time it's simply a thinly veiled accusation of cheating.
I guess that settles that.
I think they are for some people, but not for others...I don't care to expound on this...use your imagination...I am sure certain things will be evident...
Please explain this to me.
I am matched against one player, whose rating as shown in the total games list is 800.
But when the actual board is opened, it shows as 2000!
I toured his account and found that he has not completed a single daily chess match and started playing just 5 days back. All matches are in progress with 4 to 5 moves completed.
His rapid rating some 1240.
How can anyone has 2000 rating in daily chess without completing a single match.
Is there some foul play?
@sanpatkan:
Apparently chess.com is allowing new players to choose their own starting rating all the way up to 2000! Someone in another thread got up to the mid 2300's in only 7 games including a loss. Eventually it all straightens itself out, but in the meantime there can be some strangely inaccurate ratings...
It's true; at OTB you start at 101 rating but online you start at 1200. And lots of people here with OTB ratings have online ratings much higher. The range is probably about 100-200 higher than OTB. The range of difference probably gets smaller the higher the rating.