In Chess.com my rating is between 1400-1550, in Chessmaster it is barely 1050 and Yahoo.com my rating is 1150.  Is the Chess.com rating inflated or are the other deflated?  Or maybe I struggle with games where there is time or a computer involved?

Ratings tend to be inflated here.  My Yahoo! rating stays between 1500 and 1700, which is fairly accurate.  Here I float in the high 1700s to the low/mid 1800s.

This question has appeared in Chess forums since about the beginning of online chess. The best answer I can offer to you is this: There is no meaningful correlation of ratings in different systems. This means that a FIDE rating is not comparable to a chess.com rating, which is not comparable to a yahoo rating, which is not comparable to an ICC rating, etc... Some people will see their ratings very close, others will see huge disparity like you. The reasons for this include the following: 1) different sites use different formulas, or similar formulas with different variables (often called k-factors); 2) a rating is only a rough measure of your performance against players in a particular pool, namely the place where the rating comes from; 3) you play completely different types of chess on different servers. This is an important point. On yahoo or pogo you will typically play mostly blitz or bullet games. On sites like playchess or ICC, you will play a lot of Blitz, but may play some more classical time controls if you prefer. On correspondence sites like Chess.com you will typically take much more time to analyze difficult positions, which will usually lead to a better standard of play. Also, in Correspondence Chess it is perfectly acceptable to make use of opening books and databases to help guide your games, has been forever in CC as far as I know. As far as the rating Chessmaster gives you...don't pay any attention at all to that. The CM tutorials are great (at least that's what I hear from almost everyone who has used them, I personally use Fritz8 for any post-game analysis that I do), but the ratings that it assigns you is completely arbitrary. I hope this helps answer some of the questions that you may have.

 86

Thank you agent 86.  That was a very comprehensive answer.  I guess the best way to really measure how you are doing is to look at the % of the group ie a person is in the top 20% for example.

 

1. it is 32 degree fahrenheit but only 0 degree celsius. Is fahreheit inflated? It is also more than 273 degree kelvin - kelvin scale *must* be hyperinflated :)

2. at least in (1) there  is a linear correlation (statistically: +1). In the chess case it would be 0.1-0.3 max. It is almost like comparing apples and footwear.

 

Perhaps inflated is the wrong word, but the different systems do correlate to one another.  Much like measures of temperature, the rating systems are all based on equations using the same input type and having a common resulting goal.  I assume that when one refers to an inflated rating, it is implied that he/she means as a direct comparison to a FIDE or national rating.  Using that as a standard, there are definite mathematical comparisons that can be made between each system, just as we are able to convert temperatures.  Since the variables are weighted differently in each system, a more complicated equation is necessary to gain an accurate conversion, but this does not mean that they are incomparable.

While I certainly understand and appreciate your reasoning here, I have to respectfully disagree. If you take a random sample of players who play across several sites, and compare their ratings on the different sites, I'm sure that you will see that any hope of finding a meaningful correlation is short lived. I've seen people argue that one sites ratings are always higher than another, only to have just as many people come around and argue the exact opposite. The variables that play into online ratings simply do not allow for any kind of meaningful comparisons. What you can do is see how well you match up against other people who play here, or there, or the other place... What you can't do is make a general statement from that information because what is true for you will be just as untrue for the next guy, and over a large sample there will simply not be enough consistency to show a statistical correlation. It's been argued, attempted, and repeatedly failed to the best of my knowledge. I am sorry that I cannot provide you with a link to any data on the matter, I'm just doing this from memory. I used to believe that a correlation must exist, no matter how difficult it would be to define...but based on everything I have seen on the subject I cannot support that position any more. Of course, just because I disagree with you doean't mean that I am right either. This is just my evaluation of the subject, and I have absolutely no issue with other points of view, but I will try to argue and support mine.

Excellent!  I'm aware that the ratings from two different sites are not necessarily enough for the comparison.  One would also need to be able to input the weights of the main varible weighting differences.  I just got caught up in thinking about the mathematical side of it.  There is, of course, with all data on hand, a way to calculate an equation that would give an accurate conversion, but it would be so impractical as to not be of use to the average player.  In short, you are absolutely correct that the ratings alone cannot be compared with accuracy.  One would need to compare the systems themselves.