Author: Albert Silver
Date: 16:40:16 12/12/05
Go up one level in this thread
On December 12, 2005 at 19:27:47, Chessfun wrote:
>On December 12, 2005 at 19:23:26, Albert Silver wrote:
>
>>On December 12, 2005 at 19:21:04, Chessfun wrote:
>>
>>>On December 12, 2005 at 19:09:32, Uri Blass wrote:
>>>
>>>>On December 12, 2005 at 18:29:26, Chessfun wrote:
>>>>
>>>>>CEGT Blitz Rating List updated:
>>>>>
>>>>>BLITZ:
>>>>>http://www.husvankempen.de/nunn/
>>>>>
>>>>>8 New engines added:
>>>>>
>>>>>Rybka 1.0 Beta 64-bit
>>>>>Chess Tiger 2004 Normal
>>>>
>>>>It seems that chess tiger 2004 normal played 120 games out of 520 games against
>>>>ktulu7.1
>>>>
>>>>I think that when one engine play so many games against single opponent the
>>>>rating may be misleading because it is possible that one program has good result
>>>>against specific opponent.
>>>>
>>>>I suggest that you reduce the number of all games between 2 opponents to 30
>>>>games for rating calculation and it means that tiger2004 got 14.75 out of 30
>>>>against ktulu7.1 and not 59 out of 120 against it because you should not let one
>>>>player to decide about the rating list because by this idea if you play 1000
>>>>games between Fritz9 and TogaII1.0 and Fritz9 scores less than 45% you can tell
>>>>fritz9 to go down in the list and togaII1.0 to fight for the first place.
>>>>
>>>>Of course if there is data about more games I do not say not to use it but you
>>>>should use it only to get better estimate for the result in 30 games.
>>>>
>>>>
>>>>Tiger also played only against 6 opponents so I think that for calculation you
>>>>should assume that it played 180 games by reducing the number of games of it
>>>>against every opponent to 30 and 180<300.
>>>>
>>>>Uri
>>>
>>>
>>>I don't agree. Testing of Chess Tiger 2004 Normal is limted in that only a
>>>couple of CEGT testers have it. It was agreed to run 80 games per engine match
>>>however there was a mistake regarding Ktulu 7.1.
>>>
>>>Testing of Chess Tiger 2004 Normal is also limited in that it only runs in the
>>>Chess Partner interface and therefore the variety of engines to play against
>>>isn't so great.
>>>
>>>Further the method of testing within Chess Partner don't allow easy use of match
>>>play gauntlets as Chessbase GUI's do.
>>>
>>>Last assuming it will end up playing some 1500 games ultimately it's rating may
>>>get a bit skewed at the beginning.
>>>
>>>Sarah.
>>
>>What are the openings settings? Do they use a specific set of opening positions
>>à la Nunn, or do they use their own books? If so, what books do the bookless
>>engines use?
>>
>> Albert
>
>
>All use generic books in the case of Chess Tiger 2004 Normal I believe they were
>tested using the Harry Schnapp testsuite.
>
>Sarah.
What is the HS testsuite? Where can I find it?
Albert
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.