Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Chess / deflation in the fide rating list
- - By Uri Blass (*****) [il] Date 2013-08-01 22:21 Edited 2013-08-01 22:28
Rating of place 2000 in the fide rating list under 20 years old

july 2004 2140
july 2013 2061

http://www.ferimex.com/icc/frl.php?list=50&country=0&title=0&sex=1&year=14&old=0&cat=1&size=9&update=yes
http://www.ferimex.com/icc/frl.php?list=0&country=0&title=0&sex=1&year=3&old=0&cat=1&size=9&update=yes

 
Edit:I deleted a mistake that I did about girls because there were no 2000 girls under 20 in the rating list in july 2004 so my comparison is not relevant.
Parent - - By Kappatoo (*****) [de] Date 2013-08-01 22:47
1) Why did you pick July 2004?
2) 2000 of how many? Maybe there are just fewer young players today than 9 years ago.
Parent - - By Uri Blass (*****) [il] Date 2013-08-02 06:59
2000 of all the players under age 20.

Today there are more players because fide reduced the minimal rating but
I have no reason to believe that today there are less young players relative to the past when there are more humans in the world.

My opinion is simply that the main reason for the drop is that fide reduced the lowest fide rating.
children who improve fast simply take rating points from other players and the result is deflation.

I expect the deflation to continue(assuming the rating rules are the same) and that means that the level that is needed to achieve fide master is going to be higher not only in chess terms (and I believe that players are improving thanks to chess programs) but also in terms of the place in the world.

In 1.5.2012 there were 8295 players with fide rating above 2300 when 5087 of them were active.
In 1.8.2013 there are 8214 players with fide rating above 2300 when 5005 of them are active

I expect these numbers to go down in the future if fide do not change the rating rules.
I believe that the rating rules did not effect immediately the top players because they play mainly between themselves but I expect the number of players with rating above 2700 also to go down in the future.
Parent - - By Ozymandias (****) [es] Date 2013-08-02 09:18
Not so sure about the elite. As you say, they mainly play among themselves, which keeps them isolated from the "deflation pandemic". There were only 11 players in the 2700 club just 12 years ago. That number has more than quadrupled since then.
Even among titled players, the rating of the player 2000 has gone up from 2390 (April 2001) to 2407 (August 2013). Not much, but indicative about them still holding their own against the new crops of younger under-evaluated players, whom they actually meet in swiss tournaments.
Parent - - By Uri Blass (*****) [il] Date 2013-08-02 10:18
I am surprised not to find my name before july 2007 when I search for fide players from israel inspite of the fact that I had fide rating from july 2006

http://ratings.fide.com/id.phtml?event=2803968
Parent - By Ozymandias (****) [es] Date 2013-08-02 10:36
That usually happens when you change banner or (with women in some countries) you get married, but neither is the case here.
Too many players and too many years, there's bound to be the occasional fluke here and there.
Parent - - By Kappatoo (*****) [de] Date 2013-08-02 10:02

> 2000 of all the players under age 20.


Yes, but is there a way to say how many these were altogether?
Or better: Since, as you point out, the minimal rating was reduced (do you happen to know what it is today?):
In 2004 (again, why did you pick 2004, when the list begins in 2001?), the minimal entry rating was 2200, I think. So is there a way to tell how many U 20 players above 2200 there were in 2004 and in 2013?
This is obviously relevant. If there are significantly less young players today than in 2004, then this alone is sufficient to explain your observation.

> My opinion is simply that the main reason for the drop is that fide reduced the lowest fide rating.
> children who improve fast simply take rating points from other players and the result is deflation.


I'm not sure I understand this. These players don't just take rating points from others, they also gain rating points.
But you are right in that the adjustment of minimal entry ratings is relevant here. A player who peaks at 2500 used to take 300 points from others, now he takes many more.
There were also other reasons why I used to suspect that the fact that there is a (pretty high) minimal entry rating at least contributed to Elo inflation. But I wouldn't have expected that lowering it would lead to deflation.
By the way: Most national ratings don't have a minimal entry point. Nevertheless, at least for German ratings I haven't observed any deflation. How would you explain that?
Parent - By Uri Blass (*****) [il] Date 2013-08-02 10:32
today the minimal fide rating list is 1000
It was 2200 only many years ago and in 2004 it was 1600 but in that time there were not many players with low fide rating because usually the weak players simply did not play in tournaments for fide rating(even today weak players who are strong enough to get fide rating that is clearly above 1000 often do not play for fide rating but there are more players with low fide rating).

I took 2004 and not 2001 because it seems that the rating of the number 2000 player was the biggest in 2004 and I believe that the natural effect is increase in the number of players(because there are more humans in the world and after it increase in the result of place 2000).

Note that I have doubts about the link that I got my numbers as I explained in a different post but searching the fide rating list directly by myself also suggested that there are less players with rating above 2200 today relative to the past.
Parent - - By Speeed [de] Date 2013-08-19 09:59 Edited 2013-08-19 10:04

> My opinion is simply that the main reason for the drop is that fide reduced the lowest fide rating.


Yes. Of course there was a significant deflation in the region of ELO 2000, because for a long time there were no lower rated players. We still have some overrated players with 20xx ELO, who did not play any games (or not many games) in the last years. So I think we still have some ongoing deflation in this ELO region because of this source.

> children who improve fast simply take rating points from other players and the result is deflation.


Normally, one would not expect any inflation or deflation from this: If player A wins against player B and gains for example 8 ELO points, player B will lose 8 ELO points, so the ELO-average of all chess players remains the same.
However, there are different k-factors. Therefore: If a young player with k=30 wins 8 ELO points, his opponent with e.g. k=15 only loses 4 ELO points. This is one source of inflation! There is a simliar source of inflation in the region 2300-2500 because of different k-facter (k=15 vs. k=10). Of course players with lower and higher ratings are affected, so I do not agree with this:

> I believe that the rating rules did not effect immediately the top players because they play mainly between themselves but I expect the number of players with rating above 2700 also to go down in the future.


Generally I think it only make sense to talk about inflation and deflation in a certain ELO region. Just calculating the average does not make much sense. Weak players can get a low ELO now and therefore the ELO average of all players must be lower now of course. For me inflation/deflation actually means, what will happen to my/your ELO, if the strengh remains constant. Because of better training possibilities today, this cannot be calculated (only estimated maybe). I recently spoke with a 2600 rated GM and he claimed that his rating increased over the years (where he was not a professional player and did not play so many games as before) from 2550 to 2600 just because of the ELO inflation. I do not know if thats true or not, but its an intersting claim worth to discuss.
Parent - - By Uri Blass (*****) [il] Date 2013-08-20 15:26
I would like to know when the relevant GM got rating of 2550 and when he got rating 2600 to get a more specific claim(meaning that there was inflation when you compare year X with year Y).

Another question is if the relevant GM had in the past a significant number of games when his rating did not improve from 2550 and in this case what is the number of games.
Parent - - By Banned for Life (Gold) Date 2013-08-21 00:13
It strains credulity to imagine that you could get any useful information from the experience of one GM...
Parent - By Ozymandias (****) [es] Date 2013-08-21 11:21
:twisted:
Parent - - By Ozymandias (****) [es] Date 2013-08-02 10:23
Likewise, the rating of the 2000th player in the world above 50 years has gone down form 2218 to 2166 since April 2001. I guess the decline is smaller because they are higher up in the ratings and most of them are inactive.
But this are all partial data from sub-groups of chess players. You can't look at the whole picture using the tool you provide in those links. When you take all chess players into consideration, deflation rears its ugly head. Boris Schipkov took the time to compile all the data back in January 2012 (http://www.chessbase.com/Home/TabId/211/PostId/4007913/-population-growth-some-interesting-facts-about-chess.aspx), and the most revealing information contained therein was the 120 ELO points decline in average rating for all world chess players in the short span of four years.
Obviously, with the recent change in ELO admission, a good number of new players with really low ratings entering the ELO pool took its effect in the overall average. The real problem is that as they progress, be it quickly or not, they will take ELO points from the rest of the players, distorting the statistical value of the ratings in the process.
The only way this wouldn't affect the ratings in the long run, would be for those new players to wait until they have reached their initial rating before retiring. This was something chess aficionados had more or less been doing all along, but one thing was to leave when you were getting below 2000 or 1800, and another one altogether is to expect new players to play until they reach again 1000 ELO points.
Parent - - By Banned for Life (Gold) Date 2013-08-02 17:55
As you indicate, it is unrealistic to expect the players to enter and exit at the same Elo to prevent their existence from perturbing the overall pool, Elo being a zero sum game. But it would be fairly easily to periodically normalize the field based on the ratings of the top x% of the players, where x would be chosen to be a point where enough players were in the pool to keep statistical noise to an acceptable level. Of course this would preclude the system from measuring improvements (or degradations) in the collective Elo over time, but in reality the current system doesn't support this anyway...
Parent - - By Ozymandias (****) [es] Date 2013-08-02 18:54
You can't expect it now, but for quite some time it was pretty much a given. Players who didn't reach a "reasonable" ELO just weren't rated and played unrated games until (and if) they reached said level. Players who reached higher ELOs continued playing rated games and eventually retired, some of them with a higher rating than the one they had when they entered the ELO pool, others with a lower rating. In general, I would say that for the majority of chess players they were neither leaving nor taking ELO points with them on the way out.
All of this is from my perception of the ratings, of the people I've known over the past 20 years in Spain. The top percentage of players are said to "suffer" from inflation, rather than deflation, but that would be due to problems in the ELO table, according to Jeff Sonas (http://www.chessbase.com/Home/TabId/211/PostId/4007114/the-elo-rating-system-correcting-the-expectancy-tables.aspx). Other countries may also not be affected as much by the FIDE lowering the ELO floor. And even what I perceive here among the players I know may be due to something else, I haven't really investigated it, but common sense points to FIDE and those blasted children who go for their ELO right after the second chess class.
Parent - - By Kappatoo (*****) [de] Date 2013-08-02 19:37
Here is one problem with having a minimal entry rating: Let's say that n is this minimal entry rating. There will be a great number of player whose playing strength is n minus 50 or n minus 100. Inevitably, some of them will one day get lucky: They will play a great tournament and be rated n plus 50 or whatever. Most likely, when these people play more tournaments, they will just drop out of the rating list at some point, leaving rating points in the system.
Parent - By Ozymandias (****) [es] Date 2013-08-02 22:11
It's rather difficult to get ELO rating with just one tournament, IIRC you need at least 9 rated games. But I guess the scenario you describe it's possible, just one of several ways of leaving ELO points in the pool. As long as its balanced out by others taking them as they leave, there's no problem. Right now there's absolutely no balance at all, people are losing rating (fast) despite their playing skills remaining the same.
Parent - - By Uri Blass (*****) [il] Date 2013-08-03 09:14
This was correct when the minimal rating was high, so there was an inflation because the players with rating slightly above the rating floor usually did not improve and reducing the rating floor initially caused them to lose more rating points,
but I think that when the rating floor is only 1000, the opposite effect is going to be bigger(meaning that players who get slightly above 1000 improve, and practically their real strength is more than slightly above 1000, and they take rating points from stronger players).
Parent - By Ozymandias (****) [es] Date 2013-08-03 10:38
"they take rating points from stronger players" not so much stronger as higher rated.
Parent - - By Kappatoo (*****) [de] Date 2013-08-03 10:49
This sounds reasonable, but again: Shouldn't we be seeing massive deflation in national ratings then?
Parent - By Uri Blass (*****) [il] Date 2013-08-03 22:11
national rating has different rules and are different for different nations.

For example
the israeli rating rules also add points for playing for players with israeli rating below 2300 based on the formula
(2300-r)/1000 that means that a player with a israeli rating 1300 can get 1 rating point for every game that he is playing

I am sure different countries have different rules and I doubt if there are countries when national rating is a zero sum game that means that the rating that one player wins another player lose(it is also not always 0 sum game with fide but only because of different K for different players and fide does not add rating points to weak players based on the assumption that they probably improve based on experience).

Fide have a bigger K for players who did not play at least 30 games if my memory is correct and I believe that they have also a lower K for very strong players(above 2400) that is K=10 instead of K=15.

Note that the israeli rating has always K>=20
K=20 for rating above 2200
K can be also 24 or 28 for weaker levels and K is multiplied by 1.5 for players with less than 30 games.
Up Topic The Rybka Lounge / Chess / deflation in the fide rating list

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill