My rating dropped 200 points! What is going on?
Standard Ratings Boost
Sort:
Lol, looks like someone failed at doing teh SQL. 
All it does is make them random numbers.
They are no longer ratings.
Well, to be bland, sadly, it would appear that this cannot be undone anymore.
Oct 28, 2014
0
#246
LongIslandMark wrote:
erik wrote:
Optimissed wrote:
It may be to stimulate interest, pure and simple.
nope. this was done 100% in an effort to make all of our ratings more accurate. it obviously did not work for everyone. very sorry.
Erik: I respectfully suggest that making the chess.com ratings "more accurate" - assuming that means reflective of other systems (FIDE), is an unrealistic goal. chess.com ratings are reflective of the pool on chess.com - trying to make them match something else shouldn't work in the long-term.
"More accurate" does not mean more "reflective of other systems (FIDE)".
In this case, there was a logjam at around 2100. That no longer exists. Chess.com was right to fix the broken standard ratings system, but they went overboard on the top end. The top rated player now has a losing record at 2794! The rating graph shows that something went wrong in standard ratings.
http://www.chess.com/livechess/players?type=Standard
Perhaps over time, the graph will normalize, but it will take a very long time. One possible solution would be to try to give incentives for players in the top 5% to play much more. Right now, the ratings actually provide a clear disincentive to risk those shiny new points.
Proposal to incentivate the top players to play more standard rated games.
Most of the top rating players have very few games played. And mostly of them are unrated. One idea could be to use, as USCF, one provisional rating when the number of rated games is less than a specific number (25, as USCF, or 9, like FIDE, or whatever number between them). And only appear at the top list (ego is important, no doubt) if you have the minimum of rated games and at list one rated game (one game per year don´t seems to be a real problem) in the last 365 days, for instance. Ah, the statistics of win, draw and loses must only reflects the rated games, not the "cofee games". And, like FIDE ratings, you must gain at list 1 point for whatever victory (that could incentivate the stronger player to play against not so strong ones). Now, if the difference is two high, the stronger player has nothing to gain (rating speaking). At FIDE ratings, if Carlsen win a rated 1000 player (1000 is the minimum at FIDE ratings), Carlsen gain 0,08 * 15 = 1.2 points (400 points difference rule)
Oct 28, 2014
0
#248
If a different rating algorithm has been used in standard play than in the other forms, then the problem can be solved by switching to the algorithm that isn't broken.
Otherwise, this action displays a profound misunderstanding of what ratings measure and how ratings work.
I was 1700 only a few days ago, now I am 2000. Do these rating boosts happen all the time?? :D Also, now all the opponents from 1900-2200 are easier, so it seems like the inflation was way too big...
Oct 28, 2014
0
#250
This sort of thing has only occurred a few times in the history of the site. I think this is the first time such a ratings boost has happened in standard ratings.
jaadlh_6691 escribió:
Proposal to incentivate the top players to play more standard rated games.
Most of the top rating players have very few games played. And mostly of them are unrated. One idea could be to use, as USCF, one provisional rating when the number of rated games is less than a specific number (25, as USCF, or 9, like FIDE, or whatever number between them). And only appear at the top list (ego is important, no doubt) if you have the minimum of rated games and at list one rated game (one game per year don´t seems to be a real problem) in the last 365 days, for instance. Ah, the statistics of win, draw and loses must only reflects the rated games, not the "cofee games". And, like FIDE ratings, you must gain at list 1 point for whatever victory (that could incentivate the stronger player to play against not so strong ones). Now, if the difference is two high, the stronger player has nothing to gain (rating speaking). At FIDE ratings, if Carlsen win a rated 1000 player (1000 is the minimum at FIDE ratings), Carlsen gain 0,08 * 15 = 1.2 points (400 points difference rule)
Sorry, my mistake. The Carlsen´s K is 10. If Carlsen win a rated weak player he gain 0,08 * 10 = 0.8 points. Sorry
¿How long is Chess.com history?
If the cheaters are the main reason of deflation of the standard ratings, I suppose that this type of things will happen every several years...
Oct 28, 2014
0
#253
erik wrote:
Optimissed wrote:
It may be to stimulate interest, pure and simple.
nope. this was done 100% in an effort to make all of our ratings more accurate. it obviously did not work for everyone. very sorry.
Have you read the zillion threads about "accurate" ratings? Ratings are accurate only within a given pool.
My own online ratings at different sites range from a high near 2400 to a low near 1600. That's just correspondence chess.
Which of these is accurate? All.
Accuracy? Fix?
Math makes it accurate. Leave that to mathematicians. Marketing makes it pretty, and that's all this is.
Not that I wouldn't have made the same decision.
All the same
Thanks for explaining, my rating makes more sense now; its now closer to my USCF rating.
Oct 29, 2014
0
#257
Now the ratings are even more inaccurate given that expert players on this site are not FM/IM level players 99% of the time.
Well, you gotta inflate the ratings from time to time. Now a newbie has 2000+.
The government does the same. Couple of years ago I bought a kebab for 2.50 € or 3 €, today they cost 4 € (and more).
Oct 29, 2014
0
#259
Then they should be inflated by the same ratio to all members of the site. There's no real evidence that players 1600 or less were less underrated than 1800+ or expert players.
There was a psychological problem. When my blitz rating dipped below 1550, which it often did, no-one above 1550 would play me and since there were others in the same boat, many of the other under-rated players would beat me and it was really hard to get a proper rating. You go on a decent run, get it up to 1650 and then lose a couple. It seemed that 1550 was a psychological threshold that damaged the ability of the rating system to reflect true playing strength.
Incidentally my blitz rating didn't get a boost. Should it have? Only my standard rating did, of 200. 15 minutes is my favourite time limit and I'd recently got a standard rating. Online rating didn't get a boost. Was this limited to standard ratings?
My theory on why I, personally, saw no standard rating boost is that my blitz rating is/was lower than my standard rating.
So you guys conditionally boosted standard ratings for those with standard ratings lower than blitz ratings!?