Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is there a rating inflation?

Author: Dann Corbit

Date: 19:23:05 05/30/02

Go up one level in this thread


On May 30, 2002 at 22:11:11, Rolf Tueschen wrote:

>On May 30, 2002 at 20:05:35, Dann Corbit wrote:
>
>>On May 30, 2002 at 19:47:52, Mike S. wrote:
>>
>>>On May 30, 2002 at 19:29:45, Dann Corbit wrote:
>>>
>>>>On May 30, 2002 at 19:08:49, Chris Carson wrote:
>>>>
>>>>>On May 30, 2002 at 17:59:35, Amir Ban wrote:
>>>>>(...)
>>>>>>Most strong players agree that the level of play is higher than 30 years ago,
>>>>>>and that's a good enough reason why today top ratings are higher.
>>>>>>(...)
>>>
>>>>>ELO said that ratings can be compared, one of the reasons he created this
>>>>>system.  Ofcourse you are right.  However, this will continue to be a debate.
>>>
>>>>The argument is flawed.
>>>>If players never died, were never added and never subtracted from the list then
>>>>the notion would work.
>>>>Illustration:
>>>>Take a pool of players where one guy is GM level and you have 1000 IM's.
>>>>Let the pool stabilize.  You will see the GM with 100 ELO over the IM's.
>>>>Now add 10,000 patzers to the pool.
>>>>Let the pool stabilize.  You will see the GM with 100 ELO over the IM's. (...)
>>>
>>>I have questions about elo rating inflation.
>>>
>>>1. Does it exist, and if yes
>>
>>We do not know for sure if it has inflated.  We *do* know for sure that it has
>>moved.  It might also be lowered as an absolute number.
>>
>>>2. Where does it come from?
>>
>>Adding weaker players to the pool will inflate the ratings of the higher players
>>in an absolute number sense.  Adding stronger players will lower it.  But the
>>differences will stay the same.
>
>I like to remind you of the fact that another factor of inflation is exactly the
>opposite. If very few strong players play against each other repeatedly, the Elo
>score is mounting. However against weaker players a top player can't win much
>points. So the level will tend to decrease in that case for the best. Ah, you
>meant that for the rest down under the rating will increase nevertheless. Yes, I
>agree. :)

However, the top players almost never play against the worst players.  So the
net effect is that points slowly perk towards the top and away from the bottom.

>>>I had one idea: Since there are more very strong GM's "available" than i.e. were
>>>in the seventies, an even stronger "Super GM" can reach higher performances. -
>>>Just because he doesn't have to play that many opponents which are much lower
>>>rated, like it was unavoidable probably in the 70's (when there just weren't so
>>>many 2650+ players at all).
>>>
>>>If this is true, it would mean that you can reach *higher elo performances with
>>>the same strength* today (because you have more stronger opponents available to
>>>beat).
>>>
>>>If this is so, then the top ranks of the SSDF list are also affected by that,
>>>most probably (?).
>>
>>I think a simulation would be a good idea.  The actual problem is incredibly
>>complex and I am not sure if my model to understand what would or should happen
>>is correct.  Since the values are not binary as won/loss but we also have draws,
>>that complicates the issue.  What does it mean when the best player dies {or
>>retires} to the pool?
>
>Oh excuse me, I thought you were talking about human chessplayers. Sorry.

Works the same either way.  Man/Woman/Machine/TheTurk



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.