A score distribution curves comparison

Hi everybody,

as I texted in another post, I recently started taking chess lesson and I'm studying now in a structured way.

I use to play on Chess.com and on Lichess, but I noticed a very different progression of my growth on the two sites.

If, on Lichess, I had a pretty regular progression (today I'm at a 1400 score level, with a variance of 20-30 points), on chess.com I'm not stable at all. I have been able to reach 1000-1100, and the I dropped back to 900 and also less.

My perception is that on chess.com, the score doesn't reflect at all the level of my opponents. Sometimes I play with 1100 that I beat without issues, sometime I meet a sequence of 890-900 player that are devastating.

I had also the weird perception that on chess.com, every day the average raise more.

So for curiosity I compared the distribution curves of the players on the two sites:

Now, where on chess.com the average is at 780, on lichess is at 1500. 720 points of difference. We all know that the two score system are different and usually the gap should be around 300-400 points. Where the rest of the gap come from?

Possible answers can be:

1. on chess.com there is a bigger number of beginners vs lichess, which keep the avg score pretty low
2. on chess.com there avg strength is really higher than lichess, so it's far harder getting score

 I wanted just to share my thoughts and read your opinion on that.

What rating are you talking about? Rapid? Blitz? Bullet?

Sorry, you are right. I missed to specify. I’m talking about Rapid scores

@3

Maybe you should investigate also the blitz and bullet ratings.
Thus you can check hypotheses.
One difference is the seeding of new players at a different level. That is 2 different seedings.
Another difference is that the player pools are different. That is 6 different pools.

Can you please elaborate little more about these seedings and pools? I am new to such terms. Just curious to know about this 🙂

@5
A new player gets a rating to start with, and that start rating is different per site.
Some players play rapid only, some blitz only, some bullet only, so they never meet in a game.
Some play on one site only, some on another site only, so they never meet in a game.
So there are 3 * 2 = 6 different pools of players.

Got it! Thank you

Partially disagree.

its true that I’m comparing just rapid and not others, but in the rapid “basket” of players, we should have a similar distribution on two web sites with an important amount of players. And we should have also an amount of people playing on both websites. 
So, or on chess.com use to play a better average players, or for some reasons, chess.com it’s becoming harder and harder. 
This increasing of difficulty in chess.com is something I noticed many people underlined in the forum. Hard to prove, yes, but the general and diffused perception, associated to the comparison with other similar sites, make me think that it is real.

@8

"I’m comparing just rapid and not others"
++ Comparing 6 pools instead of 2 allows you to draw conclusions about the cause: either the difference in sites and seeding new players, or just the different pools.

"in the rapid “basket” of players, we should have a similar distribution on two web sites"
++ They seed new players differently.

"we should have also an amount of people playing on both websites."
++ You can compare ratings of known players on both sites.
However, most known players play bullet only, or maybe blitz, but no rapid.

"on chess.com use to play a better average players" ++ That is unsustained.

"for some reasons, chess.com it’s becoming harder and harder"  ++ That is unsustained.

"This increasing of difficulty in chess.com" ++ Unsustained.

Chess.com is a beast. 😀

Let’s see blitz scores

Very similar situation. 
It means that, taking into consideration that the average of players in chess.com and Lichess should be similar, a 1500 on Lichess is equivalent to an 800 on chess.com.

The computation method and the starting point for sure are part of the problem, but cannot be the only reasons. After a certain amount of games played the effect of the starting score and computation method should gradually disappear.

Interesting thing is that I found the chart of roughly  6 months ago….

the average was 797. 21 points less on 25 M players in 6 months, taking into consideration that number of players didn’t increase (so we can exclude the effect of new beginners) its pretty much. 

@11

"The computation method and the starting point for sure are part of the problem, but cannot be the only reasons. After a certain amount of games played the effect of the starting score and computation method should gradually disappear."
++ That is not true. Whevever a player gains rating, another player loses rating.
When site A starts new players at 1500 and site B starts new players at 800, then the same players will be around an average 1500 on site A and an average 800 at site B.
The computation method is the same: they all use Glicko-2.

inaccurate, we choose our ratings here

which obviously leads to masters at 800 and newbies at 2000

and technically, 1200 should be the "average" starting rating of chess.com anyways

I didn't get. Independently by where I start from, after a certain number of games in a community, I should be correctly ranked according to my chess skills. With a score close to other people with similar level.

Number of games I need to be correctly ranked, depends on the points the system takes or gives for each game. 

So, if both sites use the same system, it means that speed of ranking for each player is the same. 

Therefore, if I have a so bigger ranking difference in two web sites (with big amount of players), the reason can be just that in one website there is a bigger amount of players stronger than me. That's it. No other possibilities.

According to me, on chess.com the number of stronger players is increasing, definitely. The question is how is it possible?

@15

"1200 should be the "average" starting rating of chess.com anyways"
++ If chess.com starts at 1200 and another site starts at 1500, then it is only normal that the same person is rated 300 higher at the other site.
Let us take an example: Wesley So: blitz: 3148 on the other site, 3006 here.

@16

"Independently by where I start from, after a certain number of games in a community, I should be correctly ranked according to my chess skills." ++ Yes, but in getting there you draw rating from the pool if you rise or you donate rating to the pool when you fall.

"Number of games I need to be correctly ranked, depends on the points the system takes or gives for each game."
++ The system does not give or take points: your opponents give or take your points.

"So, if both sites use the same system, it means that speed of ranking for each player is the same." ++ That is right, but the absolute level can be different. 

"in one website there is a bigger amount of players stronger than me" ++ There are other possibilities. Let us assume the same players enter both sites. One site starts all players at 1200 and the other site at 1500. Then all ratings of all players will be 300 higher at the end.

"chess.com the number of stronger players is increasing, definitely."
++ Maybe they get good forum advice.

I don't want to insist, but the overall rapid average keep decreasing everyday...

it was 776 when I open this thread just 6 days ago... on 26 M players is a CRAZY effect. It means that the main part of the players in chess.com is not able anymore to keep it own ranking.

A positive (probably testable) hypothesis would be that there is a slow improvement in the rapid play quality. It is said that slower chess is better at improving quality!