Disagree with everyone, focus on improving your games, try videos, lessons anything, but don't complain like a deprived child if your elo is under 2000
If you have nothing of value to add feel free to get lost and unselect the "follow" option below. No one will miss you but I'm sure you're used to it.
Lol,
Exactly, smart guy
It's true that a bunch of weaker players joining the pool would lead to inflation of ratings for existing players, but it wouldn't change the average rating. The playing strength at a given rating would be lower than it used to be, but the distribution curve would be centered in the same place that it used to be (and that place would be equal to the starting rating for new accounts). It would be a case of "1200s aren't what they used to be," not "the average used to be 1200, but now it's 850."
I think what complicates matters is that, yes, a bunch of weaker players have joined the pool, which has likely created inflation as described above, but also many of these new players have joined the pool with starting ratings of 400 or 800, both of which are below the average rating, and this has created deflation. Both the inflationary force and the deflationary force are acting at the same time, and I can't say which force is stronger with respect to any particular person's rating.
However, the drastic decrease in the average rating (like 175 points over the course of the past six months) concerns me more than the natural inflation that would arise from the composition of the pool changing; the latter is kind of baked into the concept of Elo and Glicko style ratings, which are relative to the pool of players that are being rated.
As far as what action ought to be taken, I think that every account should start at the same starting rating and that that starting rating should be set equal to whatever the administrators of the site want the average rating to be. After this change is made, then maybe a one-time adjustment could be applied to existing accounts. If the change isn't made, then I think that any such adjustments would just be temporary bandaids. New players will continue to join, most of them will choose "new to chess" or "beginner", and this will pull down the average again, necessitating further adjustments in the future.
Read all of the above as qualified with "to be best of my knowledge" and "in my opinion." However, I'd pose the following questions to anyone claiming an alternate explanation:
1. Imagine chess.com, but with a pool consisting only of titled players. Each player is given a starting rating of 1200. Each player plays 10 games against random opponents each day for a year. No one enters or leaves the pool during that time. At the end of the year, what is the average rating?
2. Same scenario as question 1, except that instead of titled players, the pool consists only of players who have just learned the rules of the game and have never played before. At the end of the year, what is the average rating?
3. To what do you attribute the decrease in average ratings on chess.com over the past six months?
4. Another popular chess site is known to start accounts at 1500 rating and to have an average rating of approximately 1500, while chess.com (formerly) was known to start accounts at 1200 rating and to have an average rating of approximately 1200. It is (or was) common knowledge that players could expect about a 300 point difference in their rating across the two sites. To what do you attribute this difference?
As far as numbers 1 and 2 go, if Chess.com uses the same k-factor for all ratings (just varying it by activity) then I'd expect the ratings to average 1200. If the k-factor varies by rating then I would expect the ratings to average slightly higher unless all games were drawn (if player A beats player B as 1200s then they would both gain or lose X points and if player B than beat player A you would see player B gain back slightly more that X points and player A lost slightly less than X points). Since all of them have the same activity then that would have no effect on the k-factors (well, maybe a little if one player waited until everybody else had already played nine games for the day and then played the 10 games for the day).
As far as number 3 goes, it may be that prior activity has a strong affect on the k-factor. I've seen that rating changes in my games do not add up to zero for both players and (unexpectedly when used to US Chess k-factors) are sometimes higher for the higher rated player. If there are a lot of new players coming in and they start out losing a lot of games (and becoming almost inactive) then, even if they started at the site average, I would expect the average rating to go down as they lose more points than others gain from them.
As far as number 4 goes I would expect that 300 difference to be maintained if all else is equal unless the rating minimums of the two sites did not also have that 300 difference (if the 1500 site had a ratings minimum more than 300 points above the 1200 site's then the eventual difference would be more than 300 points and if the 1500 site had a ratings minimum less than 300 points above the 1200 site's then I'd expect a difference of less than 300 points). For that matter, if both sites started with the same average rating and new players were brought in at that average rating then I would expect the site with a higher minimum rating to have a higher average rating.
Thanks for your answers.
The equivalent of k-factor in the Glicko rating system used here is called rating deviation (RD). As far as I know, it varies only based on activity and not current rating. It used to be displayed on people's stat pages, but it's been (temporarily?) removed in the recent UI update.
The potential asymmetry in rating changes at the end of a game that you mention in your answer to number 3 is correct and results from one player having a significantly higher RD than the other, meaning that the first player's rating is less certain and more provisional.
A user in the previous thread to which I linked above ran a simulation and concluded that "The average rating of a pool heavily depends on how quickly you can get new users to their correct rating" with the idea that the goal should be to get players to an appropriate rating while their RD is still high/their rating is still provisional, because afterwards the asymmetry that you mentioned is gone and the average is fixed because point gains/losses become zero sum.
Another factor of interest that seems related to this is what fraction of the pool is new. It seems to me that, if two new players are playing one another, then they both will have high RD and there won't be any (or as much of) an asymmetry in gain/loss after the game. So it's possible that it's not just that millions of players are joining with starting ratings of 400 and 800, but that there are so many of them relative to the size of the preexisiting pool that provisional ratings aren't converging to a rating reflective of a player's strength before starting to count (i.e., before going zero sum).
Your point about rating floors in number 4 is also interesting. The floor here is 100, while the floor on the other site is higher. A quick search suggests that it's 600, but I'm not certain.
And you don't think that the national master in question might have a deal with this already rather corrupt site? Or are these players immune to corruption. No they arent. The ratings are deflated. Stop. Ring so gullible
Comparing the rating distribution of chess.com with other rating distributions (FIDE, etc.) aside, I think there is a rating deflation over time on chess.com itself. I started here a few years ago around 1800 and per year I tend to drop around 100 points. (Blitz). While - at the same time I am improving my game knowledge and see the success of that in other places. So I improved (maybe not by as much) but my rating dropped here by 300 points. And no - while I am not the youngest anymore - I am far from altzheimers (I hope).
One theory how this can be is (in my humble opinion) the way, arenas work.
(I play nearly exclusively in 3+2, 5+0 and 3+0 arenas).
There is no "rating based match making" from my experience. 90% of the time I play much lower rated opponents (when I am around 1500, I get 800 rated players quite often, lots of 1000 and 1100). So every misclick and every cheater costs a lot of points and it does not average out, since I 90% of the time play high risk games (in terms of rating loss).
Even from a methodological point of view, this arrangement is bad, because you actually tend to measure only the lower bound, not the average rating.
Those kind of comments are not helpful.
Mathematically, there is a problem with rating deflation for players whose ability is plateauing or improving more slowly than individuals who are earlier on the learning curve. Chess.com should address this problem, as should the USCF.
It is very hard to appreciate rating numbers when they vary so much over different rating systems and different time controls.
The review system assigns ratings 2000-2300 for player strenghts based on the review. Where the actual ratings are 1600-1700 for players that have 1800-2000 blitz ratings.
Ah well. Shouldnt care abnout rating. It's just a handicap.
GM Aman Hambleton made a youtube series Building habits 2021. His rules took him up steps in blitz gradually. He re-tried the same concept 2025. It is obvious that the same basic rules and simple skills that took a player to 800 elo back in 2021 can today take them up to 600 elo. The basid skills of player have improved drastically.
2021 most 800 rated players did not notice a mate in two. 2025 they do notice.
2021 most 800 rated players hanged a piece at least once a game. 2025 they might hang a pawn.
If old 800 rated games from 2021 are analyzed with game review, the average computer estimate is closer to 600 ranking.
In the opposite. An inflation occured.
First of all, if such "250-300" deflation was true, it'd make me some 2200 Blitzer I'm not and never was.
Then, an unknown percentage of newcomers, having chess skills comparable to your 3 yrs old sister, registered for some obscure reason, under the label "expert", and began their first game in every rating category at Elo 2000, distributing free Elo points to their opponents (including me).
Like here: (the opponent was 2000 at the beginning of the game)
He played only two more games since then, distributing more Elo points to a 1700 and a 1600 who had no trouble crushing him in 21 and 16 moves.
The said newcomer is now at 1500+, but I trust he'd be much lower if he insisted on playing more games.
That's a valid point. But deflation could happen only in certain parts of the distribution. I believe this was the case in bullet and chesscom was upfront about it even though they issued 150 points to all active bullet players. I really don't like how chesscom starts new accounts at various ratings. Everyone should start at the same rating and work their way up (or down).