Also, another issue may be the fact that most of these new players in 2020, were new, and inexperienced, and their ratings became lower and lower, dragging the average down with them.
Chess.com ratings have massively deflated since 2020
Sort:
Feb 19, 2024
0
#43
Bogo-IndianaJones wrote:
You know what else happened in 2020? A huge player increase on chess.com .There are only so many rating points to go around. Competition is fierce. For example, lets say a player starts at 400, when the average is 650. The average has now gone down, but there are more players. Now, imagines the player base doubles in size, and that causes the average rating to go down, since there are only a finite amount of rating points.
All of that I'm inclined to agree with except the phrase at the end ...
'finite amount of rating points'
maybe there's various ways to qualify that idea.
tygxc wrote:
@1
"When you literally double the population and then half of the population ranks up,
the other half must go down."
++ That is not true. A new player has a higher K and a higher RD, i.e. the new player gains more rating than the established player loses rating.
This is no deflating effect, but an inflating effect. The player pool gains rating by it.
Yes this seems correct.
robzmbi wrote:
I don't understand chess.com rating system too well, i heard it was called glickoRD and checked some wiki papers about it but.. after seeing all those formulas and things, obvious question raised in my mind.. "why?" why was it necessary? to have a seperate rating system..
new systems bring new problems with them naturally, now inflation deflation things occured, is this a side effect of glicko? how to overcome.. lots of questions
I think Glicko is inaccurate because the ideas behind it are flawed; but the idea was to have a system which adjusted faster. I always thought the Glicko guy was completely wrong regarding his claims for his system but C.c bought into it.
playerafar wrote:
The internet seems to give conflicting information.
Like this: The highest FIDE ratings of all time?
OPEN. Carlsen, Magnus. 2830.
WOMEN. Hou, Yifan. 2632.
JUNIORS. Praggnanandhaa R. 2747.
GIRLS. Assaubayeva, Bibisara. 2472.
(yes Bibi)
But another source indicates Kasparov achieved 2851 FIDE at one time.
These are the real chess ratings. But there's that 'conflict' ...
These are the current top ratings.
All time Carlesn had 2882
Women Hou Yifan 2686 (Yudit Polgar is not counted because she has played only a couple of women's events in her career)
Juniors Magnus Carlsen 2822,
Girls Hou Yifan 2673.
Feb 19, 2024
0
#47
playerafar wrote:
Bogo-IndianaJones wrote:
You know what else happened in 2020? A huge player increase on chess.com .There are only so many rating points to go around. Competition is fierce. For example, lets say a player starts at 400, when the average is 650. The average has now gone down, but there are more players. Now, imagines the player base doubles in size, and that causes the average rating to go down, since there are only a finite amount of rating points.
All of that I'm inclined to agree with except the phrase at the end ...
'finite amount of rating points'
maybe there's various ways to qualify that idea.
Ex: a new player only brings in about 400-600 points as a starter rating. As they win, they take rating points from other people and add those points to their rating. You will gain as many points as your opponent loses in online chess.
Feb 19, 2024
0
#48
Not to mention, cheaters can steal rating points, which are then stuck to their accounts forever when they are banned.
Bogo-IndianaJones wrote:
Not to mention, cheaters can steal rating points, which are then stuck to their accounts forever when they are banned.
Not true. When someone is suspected of violating Chess.com’s fair play policy, the elo points that you may have lost from that person are returned to you once said person is dealt with(suspended, banned, etc.)
Feb 19, 2024
0
#50
The idea 'finite amount of rating points' just doesn't look right.
Finite average? Looks better but too obvious.
I believe that many rating systems have been designed with a particular idea ...
but its necessary to look at the idea in reverse to understand it ...
example:
a player is 200 points stronger (and in rating) than another player - so he wins three out of four games on average between them.
Perhaps the rating system is designed with an idea like the second idea in mind leading to the first. The second idea isn't well defined because it doesn't address whether the other or fourth game is a draw or a loss for the stronger player.
IQ's were designed with an average of 100 in mind.
If there's a difference of 400 points between two players and the weaker one wins then that would be an atypically bad day for the stronger player against that competition.
Regarding adressing how rating systems work ...
it might be more useful to do game research on the typical mistakes made by players in each rating class - and against what level of competition.
I don't think I've ever seen a book title on it:
'the typical mistakes of 'C' class players.'
'the typical mistakes of Experts (the class under Master) against IM's' ....
maybe its because mistakes across classes become more similiar as the weaker player gets into time pressure.
@45
"why was it necessary? to have a seperate rating system"
++ Glicko-2 is a better rating system than elo, because it converges faster to the correct rating. Glicko-2 estimates for each player at any time not only the rating, but also the RD, and calculates an appropriate K.
Elo was devised in 1970 as a simplification, because computer time was expensive then. Elo assumes the same constant RD for all players and applies only 3 different K.
Feb 20, 2024
0
#52
tygxc wrote:
@45
"why was it necessary? to have a seperate rating system"
++ Glicko-2 is a better rating system than elo, because it converges faster to the correct rating. Glicko-2 estimates for each player at any time not only the rating, but also the RD, and calculates an appropriate K.
Elo was devised in 1970 as a simplification, because computer time was expensive then. Elo assumes the same constant RD for all players and applies only 3 different K.
I have no idea what you are talking about.. what is K ? King? 
Feb 20, 2024
0
#54
11 hrs ago
0
#41Point: Nobody has ever achieved a 2900 FIDE rating.
But you'll see people here who don't have half that strength achieve a 3000+ 'Tactics Rating' ...
Regarding rapid and blitz ratings - there's such ratings online - and such ratings over the board.
Right?
A player with 'fast connection' is likely to score higher and also beat strong players at speed chess or other fast chess time controls because of that connection speed that flags or handicaps other players ...I totally agree with playerafar
robzmbi wrote:
tygxc wrote:
@45
"why was it necessary? to have a seperate rating system"
++ Glicko-2 is a better rating system than elo, because it converges faster to the correct rating. Glicko-2 estimates for each player at any time not only the rating, but also the RD, and calculates an appropriate K.
Elo was devised in 1970 as a simplification, because computer time was expensive then. Elo assumes the same constant RD for all players and applies only 3 different K.
I have no idea what you are talking about.. what is K ? King?
Assuming that you have internet, else you could not post here.
Unless Google is banned at your country, type "ELO K" in your browsers' address bar, and you will find out.
@58
"what is K ?"
++ It is the factor used in the calculation of elo or Glicko-2.
If you win a game against an equally rated player, then you gain K/2 rating.
'
K = 40 for a player new to the rating list until they have completed events with at least 30 games.
K = 20 as long as a player's rating remains under 2400.
K = 10 once a player's published rating has reached 2400 and remains at that level subsequently, even if the rating drops below 2400.
K = 40 for all players until the end of the year of their 18th birthday, as long as their rating remains under 2300.'
https://handbook.fide.com/chapter/B022024
So new players with < 40 games or young players < 18 years and < 2300 rating gain 20 rating from a win against an equally rated player.
Established players > 40 games and > 18 years old and < 2400 rating gain 10 rating from a win against an equally rated player.
Established players > 40 games and > 2400 rating gain 5 rating from a win against an equally rated player.
That is elo per FIDE.
Glicko-2 calculates a personal K for each player at any time.
Feb 20, 2024
0
#57
tygxc wrote:
K = 40 for a player new to the rating list until they have completed events with at least 30 games.
K = 20 as long as a player's rating remains under 2400.
K = 10 once a player's published rating has reached 2400 and remains at that level subsequently, even if the rating drops below 2400.
K = 40 for all players until the end of the year of their 18th birthday, as long as their rating remains under 2300.'
https://handbook.fide.com/chapter/B022024
So new players with < 40 games or young players < 18 years and < 2300 rating gain 20 rating from a win against an equally rated player.
Established players > 40 games and > 18 years old and < 2400 rating gain 10 rating from a win against an equally rated player.
Established players > 40 games and > 2400 rating gain 5 rating from a win against an equally rated player.
That is elo per FIDE.
Glicko-2 calculates a personal K for each player at any time.
thanks buddy, so it's basicly a personal multiplier. i get it.
@Mazetoskylo google is banned in china
I'm not sure where this discussion of glicko-2 is coming from, as far as I'm aware chess.com uses glicko-1, not glicko-2.
Glicko-1 calculates K based on how many recent games you have, Glicko-2 cslcuates K based on how much your rating fluctuates.
Personally Glicko-2 sounds like a better system to me, I'm not sure why chess.com sticks to Glicko-1.
You know what else happened in 2020? A huge player increase on chess.com .There are only so many rating points to go around. Competition is fierce. For example, lets say a player starts at 400, when the average is 650. The average has now gone down, but there are more players. Now, imagines the player base doubles in size, and that causes the average rating to go down, since there are only a finite amount of rating points.