OK, ratings seem genuinely weird right now

Sort:
Matt_Rouge

I used to play around 1400 pretty regularly on daily. For the past year, however, I've been struggling to get up there. Meanwhile, I think my game has improved over the past couple years I've played here.

But before you say it's just me... Several people I play on daily here have had their ratings crater. People who used to be regularly in the high 1400s are now in the low 1300s. So it doesn't seem like just me.

But wait, there's more! Now this is really weird to me: for the past couple weeks, when I go to daily Open Challenges, everything is unrated. I used to pick up rated challenges all the time, and unrated standard games were strictly in the minority.

Are these two phenomena connected? I have no idea.

I also don't connect the apparent ratings collapse to anything else like cheating. If anything, the initiatives to stop cheating have seemed successful to me (I used to get points back from cheaters once in awhile; now, I never do).

So what are your thoughts? It really doesn't seem to be my imagination...

llama

Your rating graph shows you fluctuating between the high 1200s and the high 1300s over the last few years... so I don't understand how right now is any different.

(everyone who plays regularly fluctuates at least 100 points up and down and this is for any time control)

 

Matt_Rouge

You can see that I pretty regularly was above 1400 until September 2019 and have not been able to get back there.

But more telling is that people I've known here are also much lower, which then makes it harder for me to raise my rating again as I used to depend on nicking a few games from them.

llama

I think you're exaggerating, but sure, it seems your rating is lower even though you may have gotten better.

 

In the first 20 months you went above 1400 10 times. In the last 9 months you haven't done it once.

Your last 5 peaks average to 1380
The 5 peaks before that average to 1413

The difference is:  33.6

Matt_Rouge

I think it's a pretty significant difference.

But you haven't yet commented on the hard crash of some of the people I've been playing frequently, at least one for well over a year. Dude used to play at around 1500 and now is in the low 1300s. It's been that way for awhile. And it's not just him.

llama

I've seen enough of this type of topic to know that 99.9% of people exaggerate. If you want to talk about a "hard crash" post their username.

llama

Nothing I've done yet has used a program.

I've never written a program that takes data from webpages... I'm sure it wouldn't be hard to learn how and I could have it written in a day or two (moving average, graph various results, etc), but that's a lot of work for a single datapoint (one user whose rating appears to have dropped less than 50 points).

That's too much work to be worth it. Anyone else is welcome to do that though.

MorphysMayhem
llama wrote:

Your rating graph shows you fluctuating between the high 1200s and the high 1300s over the last few years... so I don't understand how right now is any different.

(everyone who plays regularly fluctuates at least 100 points up and down and this is for any time control)

 

 

Your rating graph shows that you have averaged about 1300 for a very long time. I see nothing out of the ordinary here.

Matt_Rouge

Users I have played that are down are:

javieroscarsi

kalinych69

ironbokken

Matt_Rouge

I'm not sure about "exaggerating." I feel that the site/app has changed since I joined in 2018. Certainly not all for the worse. I feel that a lot of cheating has been eliminated. I rarely feel, in the middle of a game, that my opponent has suddenly turned on an engine and is kicking my rear (which used to happen).

I just feel that there has been some mildly accelerating ratings deflation over the past 6 months to a year.

llama
Matt_Rouge wrote:

Users I have played that are down are:

javieroscarsi

kalinych69

ironbokken

In a number of years @ironbokken has hit a low in May. Maybe his job or kids cause that, and then this year with the virus it started earlier? I don't know. His lows tend to be in the mid 1300s... he's currently 1341.
Conclusion: looks totally normal

@javieroscarsl was bouncing between 1450 and 1550 for years. In the year 2018 he mostly stayed in the 1400s though, and then in the year 2019 he lost 200 points. This year he's recovered about 150 so far. So that doesn't line up with your data which shows a steady rating until this year.
Conclusion: I don't know what causes a person to have a bad year in terms of rating, but your bad years aren't the same (if you can call being about 30 points under a bad year).

@kalinych69 looks completely normal. About 30 rating points under his/her lifetime average... which is obviously very comfortably within normal rating fluctuation.

llama
Matt_Rouge wrote:

I'm not sure about "exaggerating." I feel that the site/app has changed since I joined in 2018. Certainly not all for the worse. I feel that a lot of cheating has been eliminated. I rarely feel, in the middle of a game, that my opponent has suddenly turned on an engine and is kicking my rear (which used to happen).

I just feel that there has been some mildly accelerating ratings deflation over the past 6 months to a year.

Maybe true. I don't know.

llama
GMproposedsolutions wrote:

Here's what I get for Matt_Rogue's rating using a ten-game moving average and one page worth of data. Doesn't it look more appealing?





Red line is obviously the moving average. Less fluctuation. Less thinking there is a problem.

It's fun to do stuff like that happy.png

Matt_Rouge

K, perception is important. If you guys don't think anything unusual in daily is going on, that's an important set of data points for me (not saying you're right yet, but I'm taking it into consideration).

Any idea, though, of why rated games have disappeared from Open Challenges?

llama

Covid has caused a lot of new players to join. I've been looking at a few blitz stats since about May.

From May to June the total number rose from 8 to 9 million, and now it's over 10 million (active players in the last 90 days is significantly lower, at 2.6 million).

Anyway, a bunch of new players playing unrated is my guess.

Or maybe chess.com has changed the visibility of rated seeks for some reason.

Matt_Rouge
GMproposedsolutions wrote:

Matt, what do you think if u only saw a 10-game moving average of your rating? as I said earlier to me it makes more sense. at the highs or the lows those are only momentary and thus a "true" rating cannot be at these boundaries. I wish those in charge would realize this.

I don't know. I don't fully understand the implication of that metric. I appreciate any further guidance you can provide.

Matt_Rouge
llama wrote:

Covid has caused a lot of new players to join. I've been looking at a few blitz stats since about May.

From May to June the total number rose from 8 to 9 million, and now it's over 10 million (active players in the last 90 days is significantly lower, at 2.6 million).

Anyway, a bunch of new players playing unrated is my guess.

Or maybe chess.com has changed the visibility of rated seeks for some reason.

You mean that the influx of new members has caused the ratings deflation?

llama
Matt_Rouge wrote:
llama wrote:

Covid has caused a lot of new players to join. I've been looking at a few blitz stats since about May.

From May to June the total number rose from 8 to 9 million, and now it's over 10 million (active players in the last 90 days is significantly lower, at 2.6 million).

Anyway, a bunch of new players playing unrated is my guess.

Or maybe chess.com has changed the visibility of rated seeks for some reason.

You mean that the influx of new members has caused the ratings deflation?

No... at least that's what I was going to say.

I was going to say they've moved the average down, but they haven't caused rating deflation... I purposefully didn't mention that idea until now because it might be hard to explain why new players can do one but not the other tongue.png

However, in the middle of typing my explanation to you, I realized there is a plausible way they could cause rating deflation in this case.

First of all, long story short, rating deflation due to new members happens when new players are better than their rating after their first 20 games or so, and then they keep playing.

This is plausible because new players improve very quickly. Someone who started chess a few months ago may have correctly been rated 400, but after a few months it's not hard to imagine them completing a few 100 games and gaining a few 100 points.

When two active players play, it's essentially a zero sum rating exchange. I.e. if I win 10 points it means you lose 10 points... so if I gain a few 100 points as an active player that means the rest of the population has to pay me for it. Multiply this by 1 million new players, and it's enough to hit everyone for a few points.

llama

Covid has created a situation where...

 - there are a lot of new members
 - new members are uncommonly unskilled
 - new members have an uncommon amount of free time

So yeah, it's reasonable to think they can cause some rating deflation.

Matt_Rouge

That does make some sense. And it could subtly and gradually drag everyone down.