Forums

Current average chess rating

Sort:
llama44

So if I tried to understand economics (which I'm ignorant about) via chess ratings (ok, I'm already lol at myself) then inflation is due to routinely overestimating the value of commodities.

By the time you find the correct value, it's too late, everything in the system has become "more" expensive (without altering the relative value between them). In other words the price of a car may be 10000x more than the price of a loaf of bread no matter what. But instead of 1 cent vs 100 dollars it becomes 1 dollar vs 10000 dollars.

WeylTransform
llama44 wrote:

Ok, but I mean... mathematicians are sort of famous for coming up with solutions 100+ years early, before people even knew there was a problem to be solved at all.

I know very little about both the theory behind statistics and economics... but doing this little program made me think they're connected.

Imagine you have a certain amount of gold. Its value depends on the market you enter, and commodities fluctuate day to day like a person's rating shadows their true skill.

In other words there were practical problems related to this since the dawn of civilization. So I'm surprised a mathematician didn't come up with something similar to Elo 1000 years ago

 

Well, yes, economists make extensive use of statistics, which is just about applicable in any form of academic research involving sample data. Both discIplines, while distinct, employ mathematical methods to represent theories and analyse problems.

I'll warrant that the conceptual notions revolving around Elo were of interest before Mr Arpad Elo. But the intricacies, the underlying mathematical aspect of it likely was not, and the mathematical aspect contributed a significant lot more (else, on which ground would the ratings be justified). 

I'm not here to discuss math history, but might I add that 1000 years ago, mathematicians were concerned to a greater degree with pure math and the application of some of their problems, rather than mathematics employed for zero sum games like chess. They probably had to invest more time into ploughing the fields or hacking at the grass than develop a framework for the primitive zero sum games.

That goes said with John Conway's statement that all games are inherently mathematical. Such a statement on its own, of course, will not seem very much, but there are quite a significant array of implications, e.g. in combinatorics, game theory, logic, etc. Sadly Conway passed away just very recently on April 11 due to COVID 19. We all know of the incredible contributions he made to math. A natural problem solver, known universally to computer programmers as the creator of the Game of Life, and yes, I can't enumerate the massive impact he made on finite groups, knot theory, algebraic number theory and what not.

WeylTransform

Speaking of mathematicians who came with the solution 100 years early, there's always the thrilling George Dantzig, who is also known for his simplex algorithm. And let's not forget me! I solved the Riemann hypothesis when I was 0 (my age may sound trivial but really if it does, you've just found yourself into the essence of the problem). Take that octogenarians! grin.png

nklristic
llama44 wrote:

I mean, there's probably some interesting psychology... maybe the names hardly matter at all... if I recall 3 of them are synonyms (beginner, novice, new).

People don't want to say they're terrible, so when faced with 3 synonyms people will usually choose one of the higher ones... and if the two higher ones straddle your average then most new users will be put there... which is statistically what you want.

Then the last 2 are something like "intermediate" and "advanced"

Intermediate is perhaps the OTB average (in the United States at least) of ~1500 which may be ~1700 on this site.

Then they choose advanced to balance the distribution (whatever it is).

Again, maybe I'm over thinking stuff, but it's kinda of fun to think about.

Well, it depends. For instance, when I started, I stated that I am a 2 out of I don't know 5, or 7 in chess (can't remember how many values were there), and I was set to a rating of 1 000. A little more than 2 months later, I am over 1 300. Not that great, I know, but it seems that my initial estimation was somewhat low.

On the other hand, I had an opponent 2 or 3 days ago. I saw that his rapid rating was 1 400 and I was a bit hyped about the game, as I had a couple of games against 1400-1450 players and those were mostly nice games (a few terrible losses, a few nice wins and wild games as well). I didn't checked his profile before the game. It turns out I have beaten him easily in around 20 moves, and I even missed great opportunities before that. His rating plummeted to below 1 200 so I checked his profile and saw it was his first rapid game. In his case he actually set himself to high in the beginning.

An_asparagusic_acid
llama44 wrote:
An_asparagusic_acid wrote:
llama44 wrote:

Hmm, I wonder how many employees the USCF has and what they're paid. I think it's mostly volunteers, and probably not very tech savvy. Just a guess.

A 3rd party could use a generic algorithm to create a better rating formula, or they might even use a neural network* to make a formula.

* to run a neural network you will need a good GPU, which is not widely available.

I assume the formula itself is 100% fine. That's the pure math side of it.

The problem is the real world  You're trying to represent people's current skill with their past results. There will always be some amount of error.

I remember an article on chessbase about the mathematical flaws of the elo system, the gliko rating system is almost certain to have a flaw.

SocialistEgypt

You can't, it's talent. some people are born with it. FACTS.

An_asparagusic_acid
SocialistEgypt wrote:

You can't, it's talent. some people are born with it. FACTS.

My IQ is 80, I am 81 years old, and I am 2000+.

SocialistEgypt

IQ=80 ?
LOL this the lowest I have heard for a while, don't give up, kid.
By the end of 2050 you can hit 85 IQ with constant training.

An_asparagusic_acid
SocialistEgypt wrote:

IQ=80 ?
LOL this the lowest I have heard for a while, don't give up, kid.
By the end of 2050 you can hit 85 IQ with constant training.

I was a Soviet national mastering my 40s. Unfortunately my chess playing abilities have declined.

SocialistEgypt

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

An_asparagusic_acid
SocialistEgypt wrote:

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

Is that why the 11 year old @scaryninja has a blitz rating that's 400 rating point's higher than yours?

SocialistEgypt
An_asparagusic_acid wrote:
SocialistEgypt wrote:

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

Is that why the 11 year old @scaryninja has a blitz rating that's 400 rating point's higher than yours?

18 Years old + Played against Bobby Fischer ? Didn't you catch the troll? That's really an 80 IQ.

An_asparagusic_acid
SocialistEgypt wrote:
An_asparagusic_acid wrote:
SocialistEgypt wrote:

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

Is that why the 11 year old @scaryninja has a blitz rating that's 400 rating point's higher than yours?

18 Years old + Played against Bobby Fischer ? Didn't you catch the troll? That's really an 80 IQ.

Lol, I new that you were trolling, I was trying to troll you by not raging.

SocialistEgypt
An_asparagusic_acid wrote:
SocialistEgypt wrote:
An_asparagusic_acid wrote:
SocialistEgypt wrote:

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

Is that why the 11 year old @scaryninja has a blitz rating that's 400 rating point's higher than yours?

18 Years old + Played against Bobby Fischer ? Didn't you catch the troll? That's really an 80 IQ.

Lol, I new that you were trolling, I was trying to troll you by not raging.

I knew you were trolling so I would pretend that I didn't know and troll about the troll you made.
Peace my friend, Together we can run chess.com forums. MY COMRADE.

An_asparagusic_acid
SocialistEgypt wrote:
An_asparagusic_acid wrote:
SocialistEgypt wrote:
An_asparagusic_acid wrote:
SocialistEgypt wrote:

I'm 18 years old with IQ of 127.
I remember playing against Bobby Fischer once and he told me that I'm a prodigy.

Is that why the 11 year old @scaryninja has a blitz rating that's 400 rating point's higher than yours?

18 Years old + Played against Bobby Fischer ? Didn't you catch the troll? That's really an 80 IQ.

Lol, I new that you were trolling, I was trying to troll you by not raging.

I knew you were trolling so I would pretend that I didn't know and troll about the troll you made.
Peace my friend, Together we can run chess.com forums. MY COMRADE.

Good luck

llama44
An_asparagusic_acid wrote:
llama44 wrote:
An_asparagusic_acid wrote:
llama44 wrote:

Hmm, I wonder how many employees the USCF has and what they're paid. I think it's mostly volunteers, and probably not very tech savvy. Just a guess.

A 3rd party could use a generic algorithm to create a better rating formula, or they might even use a neural network* to make a formula.

* to run a neural network you will need a good GPU, which is not widely available.

I assume the formula itself is 100% fine. That's the pure math side of it.

The problem is the real world  You're trying to represent people's current skill with their past results. There will always be some amount of error.

I remember an article on chessbase about the mathematical flaws of the elo system, the gliko rating system is almost certain to have a flaw.

IIRC it was failing to predict matches when the players were ~400 points apart. So sure, it's not a useful tool for that case, so we should consider modifying it for those rare cases.

An_asparagusic_acid
llama44 wrote:
An_asparagusic_acid wrote:
llama44 wrote:
An_asparagusic_acid wrote:
llama44 wrote:

Hmm, I wonder how many employees the USCF has and what they're paid. I think it's mostly volunteers, and probably not very tech savvy. Just a guess.

A 3rd party could use a generic algorithm to create a better rating formula, or they might even use a neural network* to make a formula.

* to run a neural network you will need a good GPU, which is not widely available.

I assume the formula itself is 100% fine. That's the pure math side of it.

The problem is the real world  You're trying to represent people's current skill with their past results. There will always be some amount of error.

I remember an article on chessbase about the mathematical flaws of the elo system, the gliko rating system is almost certain to have a flaw.

IIRC it was failing to predict matches when the players were ~400 points apart. So sure, it's not a useful tool for that case, so we should consider modifying it for those rare cases.

I know that glicko tried to fix that, but glicko can be improved.

m_connors

When originally posted in 2018, this topic generated 1 response. In the last 2 days it has generated 81+. I think that helps explain the ratings drop as much as anything . . . happy.png

LCIINN48

Eso quiere decir que si juego en liches tendria unos 1600?  wow

LePoneyFringant
llama44 wrote:

Ok, here's the graph after 5 million games. We can see the rating system has done a good job.

 

 

But what's the average?

It hasn't changed!

We see the average started at 1502 and after the games is 1500.

Why?

It's because as the red group moved right through the graph, eventually their high RD disappeared. When their RD was about the same as everyone else, at that moment the new average was set. The red group continues to move to the right, but the average will be unchanged. I could run it for 500 million games and it wouldn't change the average.

Pretty cool isn't it?

Hi llama44 !
I do disagree with the conclusion.
I think that if the average is 1500 ELO points, and we add a group of news players starting at 1500 ELO, the average will stay the same, no matter the level of news players (as you conclude).

But if the same group starts at, let's say 0 ELO. 
The average will goes down (no matter the level of news players).

In the other hand, if the same group starts at 4000 ELO, the average will goes up.

 

I would like to get your point of view on this idea. If you can, I would really appreciate a simulation (by the way, which program do you use ?).


So, if every player gets 800 ELO when they sign in on chess.com, the average should be 800, no matter their level, no matter the number of games they play, no matter the number of players. Cheaters may unbalanced this though.
This average is an invariant, the ELO system is build like this.


But we can see that it is not the case, Today, Avg Rating Blitz : 866.63.
I assume that is because players started with more than 800 back in time on chess.com.

The average was at 1200 and it goes down due to news players starting with 800.