Universal Rating System

Discuss anything you like about chess related matters in this forum.
Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Wed Jan 04, 2017 11:15 pm

Nick Grey wrote: Grey, Nicholas D
1793
99074 in the top 100,000
1256 in ENG list.

By comparison, FIDE's Elo method has you at 1745 for a world rank of 72568 and an English rank of 1042

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: Universal Rating System

Post by MartinCarpenter » Thu Jan 05, 2017 9:49 am

This is one case where I'd maybe back the Universal system to be working better than raw FIDE - low level FIDE grades in the UK have been a total mess for a while and we don't play enough games for the system to correct itself at any sort of speed.

If the Universal people are doing a total recalculation based purely on the raw results of the games it could well produce a non trivial improvement in terms of accuracy.

Ian Thompson
Posts: 3551
Joined: Wed Jul 02, 2008 4:31 pm
Location: Awbridge, Hampshire

Re: Universal Rating System

Post by Ian Thompson » Thu Jan 05, 2017 10:14 am

MartinCarpenter wrote:If the Universal people are doing a total recalculation based purely on the raw results of the games it could well produce a non trivial improvement in terms of accuracy.
Bad news for some players though - a quick perusal of ENG titled players shows a few very large drops, e.g.

Code: Select all

Levitt, JP		2404 FIDE, 2143 URS
Martin, AD		2364 FIDE, 2194 URS
Wicker, KJ		2275 FIDE, 1762 URS
Kenworthy, G		2249 FIDE, 1633 URS
Bennett, GH		2320 FIDE, 1439 URS
According to URS Bennett played 7 games in December; FIDE has no record of this. I thought he stopped playing about 20 years ago.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Thu Jan 05, 2017 10:40 am

Ian Thompson wrote: According to URS Bennett played 7 games in December; FIDE has no record of this. I thought he stopped playing about 20 years ago.
It looks as if they have some problems with player identification. I don't think Kevin Wicker has played in recent times either.

FIDE, the ECF and others who run rating systems make it possible to see the results included in a calculation. That makes corrections to player identification relatively straightforward.

It's a lesson of established rating and grading systems that their credibility drops very rapidly if you don't maintain the quality of player identification and result reporting.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Thu Jan 05, 2017 10:51 am

Chessbase now has an article on the subject.

http://en.chessbase.com/post/grand-ches ... ing-system

It doesn't do any more than repeat the material from the GCT press release.

I don't think we've seen anywhere a proper description as to how a difference in rating between two players translates into their expected results. The FIDE Elo uses tables derived from an underlying assumed distribution, whilst USCF Elo uses a mathematical formula directly. The ECF uses a straight line approximation.

Alistair Campbell
Posts: 379
Joined: Sat Mar 06, 2010 12:53 pm

Re: Universal Rating System

Post by Alistair Campbell » Thu Jan 05, 2017 2:13 pm

Ian Thompson wrote:Bad news for some players though - a quick perusal of ENG titled players shows a few very large drops...
I note IM Mark Condie languishing in 181,103rd place with a rating of 1532.

I'm intrigued at the thought process behind this calculation. I don't think Mark has played competitively for 12 years or so, but I would have him a little higher than this. :lol:

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Thu Jan 05, 2017 3:59 pm

Alistair Campbell wrote: I'm intrigued at the thought process behind this calculation.
With the prospects of an exciting new rating system, they haven't done basic plausibility tests. With the exception of those awarded titles as prizes, there should not be titled players with low ratings.

If you go to http://universalrating.com/ratings.php, there's limited history of ratings, but not events included.

Mark's history is
http://universalrating.com/player.php?PlayerID=576137

appearing for the first time in January 2017.

The downloadable list doesn't show any games having been played.

It seems there's a certain amount of garbage associated with this new initiative, see Ian Thompson's notes on inactive English players as another example.

Veteran US IM Bernard Zuckerman has the lowest IM rating at 1117
http://universalrating.com/player.php?PlayerID=617885

The lowest GM is Serbian Stanimir Nikolic at 2013, which as he was born in 1935 has an air of plausibility.

User avatar
Christopher Kreuzer
Posts: 8806
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Universal Rating System

Post by Christopher Kreuzer » Thu Jan 05, 2017 4:21 pm

Roger de Coverly wrote:The lowest GM is Serbian Stanimir Nikolic at 2013, which as he was born in 1935 has an air of plausibility.
Plausibility? You do talk utter rubbish sometimes, Roger. Just because a GM is 81 years old doesn't mean he or she has gone ga-ga. He was still playing rated games just over two years ago, with those games appearing in the November 2014 FIDE rating list. Rated at 2243, so no, not at all plausible. More likely he is just inactive or retired. The URS system clearly doesn't cope well with that.
Last edited by Christopher Kreuzer on Thu Jan 05, 2017 4:28 pm, edited 1 time in total.

Ian Thompson
Posts: 3551
Joined: Wed Jul 02, 2008 4:31 pm
Location: Awbridge, Hampshire

Re: Universal Rating System

Post by Ian Thompson » Thu Jan 05, 2017 4:25 pm

Roger de Coverly wrote:they haven't done basic plausibility tests.
They haven't done what they said they would do in their FAQs:
http://universalrating.com/faqs.php wrote:Players who have not played for more than a year will still have their rating recalculated as new data comes in, but are considered by the system to be Inactive. Inactive players are removed from all published lists on the URS™ website.
Doing what they said they would do would deal with long inactive players. As they haven't done that, the obvious questions are how have they calculated a rating for anyone who has played 0 games in the last 6 years, and why don't all such players have the same (arbitrary) rating, if they are to have a rating at all?

User avatar
Matt Mackenzie
Posts: 5206
Joined: Tue Mar 31, 2009 11:51 pm
Location: Millom, Cumbria

Re: Universal Rating System

Post by Matt Mackenzie » Thu Jan 05, 2017 4:30 pm

Tbf all that RdC said was that it had an "air of plausibility" in a way that the (for instance) Zuckerman rating clearly does not.

Not that it was actually correct :)

I think one or two veteran GMs (Gurgenidze, Bisguier) *have* fallen a bit below the old FIDE rating floor of 2200.
"Set up your attacks so that when the fire is out, it isn't out!" (H N Pillsbury)

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Thu Jan 05, 2017 5:20 pm

Ian Thompson wrote: As they haven't done that, the obvious questions are how have they calculated a rating for anyone who has played 0 games in the last 6 years, and why don't all such players have the same (arbitrary) rating, if they are to have a rating at all?
My theory would be that the data has become contaminated in some manner, either through the identity codes being scrambled or some test data remaining in the live system. Once FIDE and the ECF started to publish the detailed results on their respective websites,it greatly increases the credibility of the calculations.

Another thought is that the data has been "borrowed" from FIDE. Is it possible the odd bit of invented data was included so as to thwart users of "research" data from publishing a rival rating list? A bit like maps supposedly including the odd bit of fictional material.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Universal Rating System

Post by E Michael White » Thu Jan 05, 2017 5:59 pm

I suspect all this is a lot of nonsense but which are the best links to read for the basis and final ratings ?

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Universal Rating System

Post by Roger de Coverly » Thu Jan 05, 2017 6:17 pm

E Michael White wrote:I suspect all this is a lot of nonsense but which are the best links to read for the basis and final ratings ?
There's a website containing both the data and what little they are prepared to reveal about the calculation methods and how they should be interpreted.

http://universalrating.com/

There's also material at chess.com
https://www.chess.com/news/view/grand-c ... cards-3703

which includes this somewhat strange statement.
He added that a purely random ratings system would only predict the winner of a decisive game 50 percent of the time, while a perfectly accurate system would predict the winner of a decisive game 100 percent of the time. His goal for URS is "75-80 percent," which is better than current prediction systems according to him.
Elo's original formulation, as I understand it, was that a game of chess between two equally ranked players could have three results although he never came up with a model for draw percentage. He went on to observe that for players with different ratings, the results would be tilted towards success for the higher rated player from which he built the "actual - expected" model for modifying ratings.

Lewis Martin
Posts: 296
Joined: Wed May 28, 2014 11:45 am

Re: Universal Rating System

Post by Lewis Martin » Thu Jan 05, 2017 9:14 pm

Ian Thompson wrote:
Roger de Coverly wrote:There's an underlying historic premise in rating systems that two players of the same rating should have a 50-50 chance. In that context I really don't see what the Rgap and Bgap are trying to achieve.
Are they not, effectively, giving you Rapid and Blitz ratings. Take Adams and Short, for example. Their ratings and gaps are:

Code: Select all

PlayerName		URating	RGap	BGap
Adams, Michael		2730	29		78
Short, Nigel D		2672	54		112
So Adams is 58 points better at classical time limits, 83 points better at rapidplay and 92 points better at blitz.

If the method had been disclosed, you could then calculate Adams expected score against Short at each of these time limits.
I don't understand this point. Not only are you assuming that the Gap is a positive difference, it does not appear to consider Adams being inactive in Rapid and Blitz. In Rapid, the current FIDE rating has Adams as 2741(i) and Short as 2744, so the gaps are somewhat bizarre. I know it has various weighting allowed for past performances and ratings at the time, but you are almost suggesting that the expected score is similar to a rapid match Short's 2658 vs Adams's 2741.

Take my ratings:

URating RGap BGap
2111 29 77

FIDE has my ratings at: Standard 2102, Rapid 2156, Blitz 1994.

If you consider my blitz rating as inactive

((it was over 4 years ago - and played on one evening with varying results due to different efforts and playing standard (beating the defending champion first round, and got mated by a certain forumite in about 10 moves... - you know who you are!) - Note that at the time my classical rating was 1999, so it was around par then.)

which they probably did, then the URS probably isn't too far off as I have played more FIDE standard (as low as 2050-odd at one point) than FIDE Rapid in 2016. However how the figures arise for their corresponding gaps is a mystery. You would think that this means an estimate of +29 in Rapid and -77 in Blitz in comparison, but that is not necessarily indicated here. So the gaps do not make any sense to me.

E Michael White
Posts: 1420
Joined: Fri Jun 01, 2007 6:31 pm

Re: Universal Rating System

Post by E Michael White » Thu Jan 05, 2017 9:30 pm

Thanks for those Roger. Think I can guess what they're up to but need to read all the info first. Wonder if they've assessed whether under normal load it contains internal inflation/deflation, spread/contraction and whether these are uniform over the whole range of grades.