Free online Chess server. Play Chess now in a clean interface. No registration, no ads, no plugin required. Play Chess with the computer, friends or random opponents.
Play
Create a game Tournament Simultaneous exhibitions
Learn
Chess basics Puzzles Practice Coordinates Study Coaches
Watch
Lichess TV Current games Broadcasts (beta) Video library
Community
Players Teams Forum Questions & Answers
Tools
Analysis board Board editor Import game Advanced search
Sign in
Reconnecting

Could we have a Bell graph of all player rating ?

Asked by DragonFly31
Tags rating rank ranking glicko-2 glicko elo
Activity Viewed 38535 times, last updated
116
Since it is so hard to compare a player's Glicko rating to that of FIDE (Elo), it would be useful to have a distribution graph of all player rating... This way it would be possible to compare each and every one of us to ... each and everyone else.
Ederich commented :
Yeah! Upvote
I´d really like to see this too.
Please make it happen dear Lichess Team
3 Answers
49
Answered by Beeco76
I've discovered this link:
http://de.lichess.org/stat/rating/distribution/blitz

There's a graph for every chess variation.

This seems to be one of the new features of the last update.

DragonFly31 commented :
Indeed, LiChess team seems to have heard us! Thanks Lichess!
LDog11 commented :
Yes, I see the graph. I like it. Way to respond to the most frequently viewed question.
Double cheers.
14
Answered by Splike
What you're looking for is called a histogram and you're presupposing that it will be a bell curve (although you're certainly correct)

Well I Lichess doesn't do this directly, so I created a python snippet to do it.

https://gist.github.com/MarkDunne/c0f89988837b9f8de9f1

The little program uses lichess' API to first fetch a list of users then queries the ratings for each user. Because of the way the API is structured and the fact that the servers don't like getting too many queries too fast, the program works slowly through the list of users (sleeping for 1 second between each user).

Someone with some free time to spare can run the program for a large number of users and generate the plots that you want
DragonFly31 commented :
Hey, thanks for your answer. How do I run your program?
Ntune commented :
The code is written in Python so you should use a Python-Console.
TotalBoron commented :
code doesn't work for 100+ players (api restriction?)
Phexxer commented :
When can we see a graph?

Text is too short.
LDog11 commented :
It won't be a bell curve either. The curve will be skewed wickedly to the low side. I was looking at the rankings, because I just got one, and there are about 2,000 people between 1700 and 2400. (The top here.) A spread of 700 points. Going back 700 more, between 1000 and 1700 there are about 12,000 people!

I suppose if you call any curve with one lump in it a bell curve, OK, but it's not close to the normal curve.
anubiann00b commented :
@LDog11 Well if the lump is in the middle, then it is a bell curve :)
MishaON commented :
If the "lump" is in the middle it is called a normal distribution. Otherwise it is skewed.
7
Answered by static_shadow
While a graph like that would be nice, there's no need to compare the rating to that of FIDE. FIDE's rating is relevant only to the pool of players that is within FIDE, and the lichess Glicko rating only is relevant to the pool of players that are within lichess. There is no point in comparing the ratings to players of other sites who also use Glicko because their pool of players are different. As Clarkey has pointed out, ratings should be seen as a relative indicator of skill, not an absolute one.

Also, unlike ELO, given enough time (which lichess has achieved) the bell curve for ANY Glicko-2 rating system will result in generally the exact same curve. You will have a starting point at 0 and an ending point at 3000. No one will be at the absolute farthest ends. You will then have a steep curve between roughly 500 and 2500 where players begin to distribute, with the median point somewhere very close to 1500. The largest number of players will be clumped between about 1200-1800. So there isn't really much point in creating a graph when you already know what it will look like.
DragonFly31 commented :
I disagree - the FIDE and Glicko bell curve will be evidently different for both these systems because the relative values we are provided with using these two systems changes.

Thus, we can have interesting questions like - what is the mean ELO using Glicko-2 here at Lichess? How is ELO distributed ? What is the shape of the bell curve and so-on...
TurkishCuirassier commented :
I don't agree at all with static_shadow. Everything he said is dubious logic.

It's worth comparing the two graphs and the numbers do have meaning outside lichess.
MishaON commented :
Many people making these comments do not know how to use stats terms properly.
For example a normal distribution is a normal distribution regardless of the rating system used. Also some people are trying to state something which is known as a standard deviation.

Only registered members with one week of lichess activity can contribute to the Q&A.