Rating Formula Break Down

superchessmachine

Hello my friends. Here I am to try and figure out the rating difference of chess.com's ratings and USCF's ratings. 

superchessmachine

SO first we need the formulas. Anyone got either on hand? USCF and Chess.com?

Preggo_Basashi

USCF uses Elo, chess.com uses Glicko.

 

Not a big difference, but Glicko takes into account how reliable a rating is (new players and inactive players gain and lose more points). This is the RD value.

 

Knowing the formulas doesn't let you calculate a comparison though (like chess.com 1000 = USCF 1072) because it's always relative to the other players. For example both chess.com and USCF could use the same formula, but its players would still have different ratings.

superchessmachine
Preggo_Basashi wrote:

USCF uses Elo, chess.com uses Glicko.

 

Not a big difference, but Glicko takes into account how reliable a rating is (new players and inactive players gain and lose more points). This is the RD value.

 

Knowing the formulas doesn't let you calculate a comparison though (like chess.com 1000 = USCF 1072) because it's always relative to the other players. For example both chess.com and USCF could use the same formula, but its players would still have different ratings.

I know that stuff but what I am asking for is the formulas. Do you have them on hand?

madratter7

Here is the approximate formula used by the USCF.

http://www.glicko.net/ratings/approx.pdf

Preggo_Basashi
superchessmachine wrote:
Preggo_Basashi wrote:

USCF uses Elo, chess.com uses Glicko.

 

Not a big difference, but Glicko takes into account how reliable a rating is (new players and inactive players gain and lose more points). This is the RD value.

 

Knowing the formulas doesn't let you calculate a comparison though (like chess.com 1000 = USCF 1072) because it's always relative to the other players. For example both chess.com and USCF could use the same formula, but its players would still have different ratings.

I know that stuff but what I am asking for is the formulas. Do you have them on hand?

But you said you need the formulas to figure out the rating difference... so I'm not sure you do know all that stuff. The formulas don't matter at all. What'd you want to do is get a large sample of players who are active in both, (which is basically impossible to do for free).

superchessmachine
Preggo_Basashi wrote:
superchessmachine wrote:
Preggo_Basashi wrote:

USCF uses Elo, chess.com uses Glicko.

 

Not a big difference, but Glicko takes into account how reliable a rating is (new players and inactive players gain and lose more points). This is the RD value.

 

Knowing the formulas doesn't let you calculate a comparison though (like chess.com 1000 = USCF 1072) because it's always relative to the other players. For example both chess.com and USCF could use the same formula, but its players would still have different ratings.

I know that stuff but what I am asking for is the formulas. Do you have them on hand?

But you said you need the formulas to figure out the rating difference... so I'm not sure you do know all that stuff. The formulas don't matter at all. What'd you want to do is get a large sample of players who are active in both, (which is basically impossible to do for free).

I see. What I was trying to do is see what would happen if one 100 rated player wins 10 games against 200 rated players and then compare the end result

wanmokewan

Have you tried this revolutionary thing called an "Internet search"?  Go to google, type in chess rating formula, and see what you get.  Sorry for being smarmy but, come on, that's basic.  Heck, Wiki has some stuff.

superchessmachine
wanmokewan wrote:

Have you tried this revolutionary thing called an "Internet search"?  Go to google, type in chess rating formula, and see what you get.  Sorry for being smarmy but, come on, that's basic.  Heck, Wiki has some stuff.

I do not have access to google.

Preggo_Basashi

They're both really similar when it comes to stuff like that.

In either one, if you beat someone rated the same as you, you get about 10 rating points.

The big difference is how they deal with e.g. new players. Glicko is better at getting new players to their correct rating faster (I guess). 

Preggo_Basashi

So stuff like this (I don't know what you can open or not)

 

Glicko:

https://en.wikipedia.org/wiki/Glicko_rating_system

http://www.glicko.net/glicko/glicko2.pdf

 

Elo

https://en.wikipedia.org/wiki/Elo_rating_system#Mathematical_details

superchessmachine
Preggo_Basashi wrote:

They're both really similar when it comes to stuff like that.

In either one, if you beat someone rated the same as you, you get about 10 rating points.

The big difference is how they deal with e.g. new players. Glicko is better at getting new players to their correct rating faster (I guess). 

Yeah. I agree with all of that

superchessmachine
Preggo_Basashi wrote:

So stuff like this (I don't know what you can open or not)

 

Glicko:

https://en.wikipedia.org/wiki/Glicko_rating_system

http://www.glicko.net/glicko/glicko2.pdf

 

Elo

https://en.wikipedia.org/wiki/Elo_rating_system#Mathematical_details

Doesn't FIDE use elo as well?

Preggo_Basashi

Pretty sure FIDE uses Elo.

Umm, but I guess USCF uses some modified version of Elo. Maybe FIDE does too. Is that what "glicko2" is? I don't know.

 

In chess.com ratings there's literally a state labled "Glicko RD" so I'm pretty sure they use Glicko.

superchessmachine
Preggo_Basashi wrote:

Pretty sure FIDE uses Elo.

Umm, but I guess USCF uses some modified version of Elo. Maybe FIDE does too. Is that what "glicko2" is? I don't know.

 

In chess.com ratings there's literally a state labled "Glicko RD" so I'm pretty sure they use Glicko.

Yeah I think USCF has some thing with bonus points and such. As for glicko I think you are tight.