Why the elo change?

Like we know the chess elo in internet is different of the real elo.

In another sites change too, In the site i been playing i have a elo of 1500, but in chess.com have 1300, and maybe change in another sites.

But here is my question

Why the elo change?

What type of factors make change the elo.

We can have 1500 here but in the site i play you may can have 1800.

I dont understand what makes change elo, how strong are the players?

Chess.com uses Glicko but the biggest difference is that the pools of players are different. Different pools will generate different ratings.

For example, USCF and FIDE are different pools and if you have a player rated in both, those ratings will differ for most players. I believe in general, a 2200 USCF is likely around 2100 FIDE.

How does this get asked so often? Even worse, some people seme to be unaware that the ratings aren't directly comparable!

So, imagine a new chess site was starting up now and everybody who registered started with a rating of 2200. You register on that website and play enough games to get a stable rating. It would be higher than your rating(s) here. Do you understand why? If so, why did you ask this question? If not... No comment.

New scenario, imagine chess.com got a new sister site that only allowed people with an long time established blitz rating of 2K+ here on chess.com to join. Everybody on that new site started with a rating of 600. After the new site had plenty of time to get going, what percentage of members do you believe would have ratings on that website that were higher than their ratings here? If your answer is <1% then why did you ask this question?

 obviously pool of players plays a role, no argument there, but i was just thinking...could it also be length of time site is established, or essentially accuracy of elo.  For example, let say god forbid all chess.com members decided that chess.com was no longer suitable (perhaps the trolling finally took its toll) and they all simultaneously leave to a new site made just to capture this new market share.  (you catch my drift) same pool of players, but provisional ratings wont be nearly as accurate as a site that has been established for so long.

Thoughts?

woops i guess my scenario is practically the same as chessoath's first scenario

No it's not.

Wait...but isnt your first scenario you gave, where people started at 2200 the reason why people will have generally a higher rated elo? i feel you are being contradictory to your original statement.  Its a 0 sum system, so in a pool of x players, after ratings are established, (sum of elo)/x=the starting elo.

I meant it wasn't "another" factor since I had already used it in both of my examples.

That's not totally true. Rating systems are not zero-sum, especially Glicko (I guess some others might be). The amount of rating gain/lost by one opponent won't necessarily equal the amount the other player lost/gained.

What the starting rating can do is alter your rating distribution curve. Most of the ratings here are close to the 1200 rating historically used. In a few years, when v3 has been running for a while, where starting ratings can be different, it will be interesting to see how that changes things.

There's nothing wrong with what you've said, Martin, but I wish people wouldn't bring Glicko into these threads. If the person in question isn't naturally sure of the basic concept that two entirely seperate ratings aren't directly comparable then do you really want to confuse them further with Glicko?

It's relevant and it's true, but I think a little bit more appreciation for what's possibly going to resonate with this particular audience is in order. Nothing good can come of mentioning Glicko under these circumstances. There are plenty of simpler and more widley applicable ways to explain.

But you said yourself above that if site A starts players out with a different rating than site B, then the stable ratings of a given player on the two sites are expected to be different, or have I misunderstood?

Edit: nevermind, post #11 seems to have cleared that up.

Even within the USCF, for example, it's difficult to compare ratings.  To begin with, ratings are not absolute, they have a variance (not published), and they are dependent on the competition that a player has faced in the past.

I remember playing in a tournament where a 1600+ could not beat any 1300+.  Probable reason:  he came from a small city where the competition was weak,  The tournament was in a major city where the competition was strong.  His rating was inflated because of the weaker competition that he regularly played.  His competitors' ratings were suppressed because of the stronger competition that they regularly played.

Note: This happened decades ago when the rating formula was different.  Also, read about the USCF supposedly freezing Jude Acers rating when they found that it was approaching 2400 and he hadn't played any strong competitors.

If this works, i dont care of how much elo you start maybe 1500 ir 1200, i have a 1500+ elo and i play, maybe in the another site there are more weak players

Yes, something similar happened to me... I spent three consecutive days mindlessly catching pokemon... Came back and my rating dropped over 200 points...

Pokemon is dangerous, kids.

Hasn't Chess.com also changed the starting rating in V3? So you can nominate yourself whether you want to start with 800, 1200 or 1800?

Yes. There are 4 or 5 ratings I think. That's why I think there may be some shifts in the distribution curves in the future.

The FAQ has a summary article that also links to this blog post from Erik, which I think is pretty helpful: https://www.chess.com/article/view/chess-ratings---how-they-work

  For me when i belonged to uscf my rating there was very close to my rating here. On other online sites though my rating was usually much better than my uscf rating. I remember once someone started a thread for uscf players to list both their rating here and there uscf rating. It seemed like there was roughly as many players that had higher ratings here as lowwer ratings, so i think the uscf rating system is fairly close to the one we have here.