Internet ELO ratings compared to actual FIDE ratings

Hey everyone,

I was just curious about something, how accurate is your internet rating compared to an actual FIDE rating? I've heard from some chess players that in certain internet chess site like ICC (internet chess club) and playchess.com (chessbase), your internet rating is actually quite reflective of a 'FIDE' rating as compared to a smaller chess playing site. 

I myself personally don't have a official FIDE rating (I just don't have the time to travel to clubs and go for tournaments..I prefer the convenience of home. ) I do have an account on playchess.com which i play regularly online with. Would anyone out there with an actual 'FIDE" rating and a playchess.com rating care to share some light on how similar they are? Likewise, for anyone else with an actual FIDE rating and an ICC account (or other chess sites) care to throw in their comparative figures as well.  (Oh, and not forgetting chess.com as well)

Thanks so much!

I believe ICC has done a survey or two on this. Just have to google around I imagine to find it.

As for me, it was USCF+150=ICC and USCF+200=FICS...Standard ratings.

And how about this site and FIDE?

Know that but how big is the difference, let's say here I've got a rating of 1800 than if I go a club and confront a 1600 FIDE who has bigger chance to win. A formula pls

There's been many posts about this.

There is no direct correlation, you cant translate a rating from one pool to a different pool.

Ratings are relative measures based on other people in the same ratings pool. It doesn't make sense to try and compare them, especially when you are talking about ratings derived from different skill sets (correspondence vs. otb). The more games you have, the more accurate your rating will be.

To say that your rating here doesn't mean anything is to say that your games here don't mean anything, in which case I'm not sure why you're here.

Perhaps a way to find out is to checks chess.com players disclosing their FIDE or USCF ratings compared to chess.com rating. I tried to do that, and obviously there is some corelation but I am not sure what that means. My sense is that a chess.com rating is about 200 points above a USCF rating. Anyone with more knowledge on this?

The only real difference if you can put the red herrings of the stat boys aside is the timing of rating publication, net after each game / governing body less frequently.

Local concerns for me the annual September English Chess Federation grades are now released. A realignment has apparently taken place to correct a reported grading slippage from previous years. Over 220 ECF (2400 FIDE 2500USCF) not affected, 210 to 220 remain roughly the same. Below this elite level the lower the rating the higher the increase. Its an average thing and players have still gone up or down subject to previous results.

So if some are comparing ECF to net they may need to realign a bit.

Good Luck

>:)

This is not true.

Ratings only make sense compared to other ratings, yes. So you can't say that 2200 USCF == 2200 FIDE. It does not logically follow that you can't come up with a formula. On the internet it's much more difficult though. Some reasons:

1. Opponent selection
2. Computer usage & other cheating
3. Nothing really "at stake" for some people, so they don't care as much.

 There's constant debate about whether a rewards system is better than an Elo system or vice-versa.  I know that the other game I play (World of Warcraft) has tried a rewards system (start at 0, work your way up, rating doesn't go down ever) and it got mixed reviews.  But people have such a hard time understanding Elo (this thread is a perfect example why) that I wonder if we would be better off with rewards instead.

All systems are subject to manipulation and abuse, but in general, your rating on paper should reflect your skill relative to the other players that you frequently play against.  If you play about the same amount of FIDE tournaments as you do Chess.com tournaments, the numbers will have a good correlation.

USCF gives players bonus points for superior performance in a tournament, so a player's USCF rating is often 50-100 pts. higher than other rating systems.

For more info, check out:

http://en.wikipedia.org/wiki/Chess_rating

Comparing chess ratings with bridge ratings and the tennis point system, I like chess much better. Both the bridge masterpoint system and the tennis point system reward frequent players. The thing is - I want a system which tries to rank players in the sense that if player A has a higher rating than player B then they are more likely to win the game. This is not true in bridge or tennis as much as it is in chess.

Also, Chess.com uses a modified version of the Elo system, called the Glicko system.  Each player has a number called "RD" which is their ratings deviation.  This number goes down if you play a lot of games (your score is more accurate), and goes up if you are less active (older score - might not reflect your true strength).

Note that "Elo" is a person's name (Hungarian mathematician Arpad Elo), and not an acronym, even though many people still like to write it "ELO".  "Glicko" is named after professor Mark Glickman.

In the Elo system a difference of 100 rating points meant the higher rated player would score c. 66%, a difference of 200 points c. 75%.

Is this still the case?

 But it logically makes an formula suspect. You're not going to get something like 2000 chess.com = 1700 USCF. You're going to get something more like 2000 chess.com = 1600-1800 USCF. And that 200 points is a signficant range in the glicko system. It's not clear how big that range is, because there isn't good data to analyze. I'm working on some simulations to try and get a best case calculation of that range, but it's a complicated problem.

 "Hungarian-born" would be better. He moved to the US as a child.

http://www.chessclub.com/help/survey

 Intertnet surveys are unfortunately not a good source of unbiased data.

You have a better idea?

 Yes, a couple. I'm working on one of them right now, as I stated earlier in this thread.