That data set is a bit small.
The best fit I've found so far(let U=uscf rating and C=chess.com rating) is
C=-7.3*10-8(U)^4+.000481(U)^3-1.1649(U)^2+1232.21(U)-478642
Yeah, it's nasty, not easy to write in a text editor, and R^2=.7966 so it's probably the best bet I'm going to find without getting out some serious computing power.
That data set is a bit small.
The best fit I've found so far(let U=uscf rating and C=chess.com rating) is
C=-7.3*10-8(U)^4+.000481(U)^3-1.1649(U)^2+1232.21(U)-478642
Yeah, it's nasty, not easy to write in a text editor, and R^2=.7966 so it's probably the best bet I'm going to find without getting out some serious computing power.
It seems fairly consistent among most people though, those were just two examples.
I was actually referring to likesforest's post, two examples isn't even worth running regressions on.
I have looked at a couple of dozen cc games here. Not all that many, to be sure, but in each case I see some quite bad play that does not correspond with what I know of USCF otb-ers with similar ratings. The cc games are much worse quality than what I see otb, with similar ratings. I can only conclude that the ratings are very inflated as they express chess skill/knowledge.
Live chess. Haven't looked at one game so I have no opinion.
This is a common misconception about Chess ratings. They don't measure chess skill or knowledge. They measure the likelihood that one player will beat another within a given population.
Ratings reflect relative likelihood of a long term result, not in personal terms but in statistical ones. Since chess games are won by and large with superior chess skill (or less crappy skill in many cases), ratings can fairly be held to reflect a level of chess skill. It is fair to state that a group of players who is beating another group 3-1 is the group with greater chess skill. A higher rating reflects more chess skill.
Define "chess skill." Is how much sleep I got last night part of chess skill? Because I think it will definitely affect my chess performance, which is what the rating system uses to make it's estimates. Also, chess skill is different within different populations, and ratings are population dependent. Furthermore, since ratings are population dependent, it doesn't make sense to compare ratings from different populations. My Chess.com rating is not in any way "inflated." It is an unbiased a represenation of my performance within the Chess.com population. No, it is not a good representation of my performance in the USCF population. But it was never intended to be that, and it never claimed to be that. It is, however, a common misconception that it can be used in such a fashion.
I have looked at a couple of dozen cc games here. Not all that many, to be sure, but in each case I see some quite bad play that does not correspond with what I know of USCF otb-ers with similar ratings. The cc games are much worse quality than what I see otb, with similar ratings. I can only conclude that the ratings are very inflated as they express chess skill/knowledge.
Live chess. Haven't looked at one game so I have no opinion.
This is a common misconception about Chess ratings. They don't measure chess skill or knowledge. They measure the likelihood that one player will beat another within a given population.
Since the point of the original poster was how the ratings here equate with USCF, we had to try to answer even if they are two different populations, playing under different time controls and in very different settings. Ultimately an otb chess player with some years of experience notices when players here with a 1600 cc rating play with nowhere near the skill of the 1600s he is familiar with otb. So the comparison is made and the conclusion is that the 1600 here is "inflated" (not technically the right term but the one we use without much of a stumbling block) relative to the 1600 USCF. We also note that the chess skill of the obt-ers vs. that of the onliners isn't reflected in their similar ratings, as expected.
While ratings don't directly reflect a level of chess skill, as was stated, those adults with long-established ratings have an excellent idea what those numbers imply for those like them and serve as a fair predictor for the result of many games vs.many opponents of likewise established status.
Is it possible to study and find something meaningful between the two ratings in a sense.
We have a population of people on chess.com who have both an online(or live) and USCF rating, so it's only natural to ask about the intersection between the two populations since there's a large group of members who exist in both. I think this idea can be extended a bit, since it isn't a terrible assumption that <2000 FIDE=USCF to include some international members into our sample populace as well.
While ratings don't directly reflect a level of chess skill, as was stated, those adults with long-established ratings have an excellent idea what those numbers imply for those like them and serve as a fair predictor for the result of many games vs.many opponents of likewise established status.
Actually, the original post didn't ask how they related. The original poster noted a difference in their ratings and asked if that was correct. There was no need to make a comparison, or to state that Chess.com ratings are inflated (something which is more than technically incorrect). All that was needed was to state that they are different populations and different skill sets, and therefore the ratings are not comparable.
Let's all stop talking about the USCF comparison to CC chess here and let's start talking about LIVE chess.
I don't think you can compare online rating with board play. I do both and at the board you have to deal without the analysis/cheat board(you need to visualize/think ahead, try it it is actually pretty hard to think ahead like +4 moves).
Also time stress, stress in general, noise around, concentration and so on influence your strength. I know a few clubmembers with low rating who are actually good on the internet, but they easily get beaten by +1700 club players in real life chess.
The biggest problem imo with online chess it actually force you to think about lots of games the same time(This is hard for visualizing, at least for me it is). When i play one game for +1.30h long i play a lot stronger and don't have to visualize every time a new game.
At least this is my opnion =)
While ratings don't directly reflect a level of chess skill, as was stated, those adults with long-established ratings have an excellent idea what those numbers imply for those like them and serve as a fair predictor for the result of many games vs.many opponents of likewise established status.
Actually, the original post didn't ask how they related. The original poster noted a difference in their ratings and asked if that was correct. There was no need to make a comparison, or to state that Chess.com ratings are inflated (something which is more than technically incorrect). All that was needed was to state that they are different populations and different skill sets, and therefore the ratings are not comparable.
You're just being argumentative now. A comparison was called for and given.
Nice ad hominem, goldendog. I'll have to remember that one.
Nice ad hominem, goldendog. I'll have to remember that one.
I didn't attack you. Where in my post do you say I made an ad hominem?
Enjoy playing with statistics. I have no problem with that or you. If you wish to insist on "inflation" as only a term of art for statisticans to define here then you are in the wrong forum perhaps. I and others used the term usefully here, if colloquially, and advanced some understanding of what ratings here mean in terms of relation to USCF otb ratings.
Just relax.
You say I'm just being argumentative, that I'm "playing with statistics", and that I need to relax. Those are all ad hominem arguments. If that's how you argue, I'm not going to argue with you anymore.
You say I'm just being argumentative, that I'm "playing with statistics", and that I need to relax. Those are all ad hominem arguments. If that's how you argue, I'm not going to argue with you anymore.
Playing with statistics, playing with chess...both ok. You were being argumentative. Take it or leave it it's no big deal.
When you play the Gatekeeper and try to tell me that I should have just answered as you have, re different populations etc., and no more, then you lay yourself open to some rebuttal and perhaps even criticism.
If "just relax" qualifies as an ad hominem for you, then I suggest you are being hyper-sensitive.
thank you for all of your wonderful posts =]
I believe the answer is +42
Online rating is pretty irrelevant. a rl FM might be blitzing out his move on intuition after a party, and he might lose against a 1600 who has been looking at the position for days. If you play a 1200 you could be playing a beginner who only knows the basic moves, or you could be playing a 2000 who only just started.
Who cares about online rating anyway..
However after 1900 or so live chess ratings the competition is very thin and so I think that those ratings will be a bit inaccurate.
Phelon (online) 1949 (live) 1712 (USCF) 1655
Zug (online) 2534 (live) 1837 (USCF) 2208
etc. It seems in general if you see a player with an accurate USCF rating (if he played 4 games 10 years ago it may not be exactly right ;p) it will coincide very well with his most accurate live chess rating, (again if he has only played 2-3 live chess games this rating will be very inaccurate) however his online CC chess rating will be very inflated.