Hi,


Player 1 1 Xie, George New rating 2500 New RD !!

+40 Rating points

BTW if you are interested you can verify this information using Rincewinds Glicko Calculator here Barry's Web Site

is the new rating Glicko or something decent? Will George now be entitled to be a GM with that rating?

No and NO

The Glicko system ought to be removed permanently from Australian chess. It has some serious conceptual flaws with its pre-determined premises.

Grand Master Titles are not awarded on Mickey Mouse ACF rating systems. George has to get his only important rating, the Fide Rating, over 2500.

It seems reasonable to ditch glicko or perhaps any local rating system in favour of the FIDE rating system.

What is the argument against doing this ?

Cost ?
Mathematical-statistical sampling issues ?

One would assume there is a perfectly rational answer .... and if someone could encapsulate it for us , i'm sure it would be very useful .

Hi,



26 Kolak, Chris VIC 2059 2011 Open 4 28:W 37:W 12:L 13:L 25:W 10 16 7:L 0:L

Ok Chris had a very poor tournament, but it is simply ridiculous what is going to happen to his rating. The faulty Glicko premise is that he is rusty and therefore will perform badly. But this is a ridiculous presumption because Chris regularly plays chess at different time controls. In fact last week he beat FM Michael Baron and drew with IM James Morris in the MCC allegro.

His RD will be Blank therefore he performed at 1787 or -224 points.
It is worth considering what would have occurred with other RD factors

!! 1941
! 1879
? 1666
?? 1542

Does his total historical tournament record mean nothing? This is being done to a guy who lost an impromptu 2 minute match against GM Ivanchuk 3:2 at MCC, the stuff of folk legend.

There goes 224 points from the Cup weekender playing pool!

Hi Kerry,


I think that if the only issue facing an ELO system is underated players deflating the pool, then their are many simple approaches that could remedy the situation.

1) Set a standard cut off at 1000 where a players rating cannot fall below this threshold. This will prevent massive distortion from young juniors. It will also stabilise the junior pools in an effective manner.

2) Make the requirement of an adult rating a certain amount of wins against players not games.

3) Only take maximum points from the playing pool when the new player has a historical record of X amount of games (X=100 for example)

4) Weight a win/loss record of the new player to reduce the loss of points to established players. Factor this as being more important then the actual rating for deducting points.

But in any event, the premise that inactive players with proven historical samples ought to be penalised for the sake of imagined rating accuracy in the moment ought to be abolished immediately.

Perhaps this 'official' calculation will be closer to the December1 release of ACF ratings than the 224 estimate you have produced.

I looked at the cross-table to see who Chris Kolac dropped points against and noted that a loss and two draws were against 3 youngsters who are all so active that their ACF ratings would have to be judged as extremely reliable, not under-rated newcomers.

I had a quick look at the Vic OPEN field for 2011.
The list below is (junior) players who would suddenly be base-lined at 1000, if I understand your first proposal.

I eye-balled the cross-table and found none of the top half of the field dropped even 1/2 a point against the listed players in all of 7 rounds.
Not sure your proposal has any relevance to protecting the top half of the field from abnormal rating decreases.




62. Foo, May-Yi 1000
63. Maligin, William 858

66. Chin, David 774
67. Chew Lee, Alanna 732
68. Loucas, Carl 702
69. Krstevski, Tristan 683
70. Bhat, Vishal 676
71. Chan, Kris 633
72. Ebeyan, Thomas 600
73. Chan, Luis 592
74. Guo, Zhi Lin 532

76. Middleton, Jody 478
77. You, Jennifer 433
78. Fan, Phoebe 328
79. You, Jessica 219
80. Chan, Vincent
81. Cotton, Sam
82. Herath, Thusantha
83. Li, William


86. Ren, Kevin

Hi Kerry,


I think that if the only issue facing an ELO system is underated players deflating the pool, then their are many simple approaches that could remedy the situation.

1) Set a standard cut off at 1000 where a players rating cannot fall below this threshold. This will prevent massive distortion from young juniors. It will also stabilise the junior pools in an effective manner.

2) Make the requirement of an adult rating a certain amount of wins against players not games.

3) Only take maximum points from the playing pool when the new player has a historical record of X amount of games (X=100 for example)

4) Weight a win/loss record of the new player to reduce the loss of points to established players. Factor this as being more important then the actual rating for deducting points.

But in any event, the premise that inactive players with proven historical samples ought to be penalised for the sake of imagined rating accuracy in the moment ought to be abolished immediately.

Moz, It is time to examine your thinking on this matter. The real issue is about the deflation of the overall playing pool.

In the context of this discussion it was pointed out by Kerry Stead that rating pools calculated using ELO are rating neutral, in that they balance out in equilibrium.

To show how this works I will select 2 names on your random list of players.
62. Foo, May-Yi 1000
and
72. Ebeyan, Thomas 600

Let us now imagine that we construct a rating period where these two players play each other in a match over a large enough sample (25-100 games) where we can get a true rating indication.

Let us, for arguments sake say that, they break even at 50%.

Both players will now have a rating of 800 points. May will lose 200 points and Thomas will gain 200. The pool is in balance.

However, if May is not allowed to drop below 1000 points, utilising my suggestion as listed as item 1, then Theoretically May will remain 1000 and Thomas if we had to rate his games would be 800.
The difference is that the pool gained 200 points at the bottom, but didn't deflate at the top.

Of course, under my suggestion Thomas would not have a rating of 600. It would also start at 1000 so that at the end of the match they should just be exchanging 7.5 points back and forwards. Therefore they would both end up on 1000.

This is exactly how the Fide system used to operate when players ELOs fell under 2000.

cheers,