Winboard Forum

Hi all ,

CEGT downloads and rating lists 40/40 (2 Ghz) are updated. Hopefully everyone will find a list telling most .

http://www.husvankempen.de/nunn/downloads.htm

http://www.husvankempen.de/nunn/rating.htm

Best Regards
Heinz

Hi Heinz,

I am not sure your rules allow this, but if they do, I would advise you to count all Glaurung versions from 060105 to 1.0.1 as a single version on the CEGT list. The changes from 060105 to 1.0.1 are mostly limited to new features which have no influence on strength.

Tord

Hi Tord ,

okay will be done with next rating list. I already guessed that the rating differences for the versions could be explained by relatively few games still for the latest two versions. Anyway it helps that the rating lists will not be cluttered up with many versions.

Best Regards
Heinz

Tord Romstad wrote:Hi Heinz,

I am not sure your rules allow this, but if they do, I would advise you to count all Glaurung versions from 060105 to 1.0.1 as a single version on the CEGT list. The changes from 060105 to 1.0.1 are mostly limited to new features which have no influence on strength.

Tord

I remember that there is one change in the default parameter of pruning when you changed it from 1 to 2.

I think that the best is to have 2 lists when in one of them they are a single version and in one of them they are not a single version.

I can also say that versions 332-337 of movei can be counted as a single version because there are only very small changes but I also prefer to see result of all versions.

Uri

Uri Blass wrote:

Tord Romstad wrote:Hi Heinz,

I am not sure your rules allow this, but if they do, I would advise you to count all Glaurung versions from 060105 to 1.0.1 as a single version on the CEGT list. The changes from 060105 to 1.0.1 are mostly limited to new features which have no influence on strength.

Tord

I remember that there is one change in the default parameter of pruning when you changed it from 1 to 2.

This is correct. All versions from 060105 to 1.0 should play exactly the same moves in all positions, but 1.0.1 has a different default value for the "Static pruning depth" parameter, and will therefore play somewhat differently.

I would still prefer that they are all counted as a single version. I don't like to use the CEGT list as a tool to compare the differences between various parameter settings. Instead of having two or three almost identical Glaurung versions on the list and trying to get an accurate rating for each of them, it would be better to count them as one version in order to get a significant number of games more quickly, and to use the time saved to test a bigger number of engines.

Tord

Hi Tord, hi Uri ,

in order to avoid cluttering up the lists and to give a statistically relevant amount of games for fast updating authors it is advisable to combine some very similar versions.

I can tell you that I have a second database being identical with the only difference that all versions are still separated. So information about any version and rating can still be given by email, although the rating lists show only the combined ratings.

Hope this would be a compromise. Another proposal would be to eliminate games from the rating list when there are more than 5 versions for each engine and maybe only give some "eternal" rating list from time to time. This one will be almost unreadable when we continue with CEGT for some more time because of so many versions . For Chessmaster I already started to give only a few in the all version list.

Best Regards
Heinz