Author: Mike S.
Date: 15:47:19 10/20/03
Go up one level in this thread
On October 20, 2003 at 16:48:35, Enrique Irazoqui wrote: >On October 20, 2003 at 13:36:59, Mike S. wrote: >>(...) >>The problem of running The King in the Fritz GUI (and in other Auto232 capable >>GUIs, like Shredder classic and Arena) has been solved since ages! So, using the >>King engine and Auto232 is possible. I'm sure SSDF ist aware of this. The Wb2Uci >>topic is well known; lots of people use it constantly. So, the original book is >>the only problem, but for that I think a good "neutral" CTG book could be >>created from IM/GM games, to be used in these test runs. >>A little compromise - much better that not to test it at all... >>http://www.talkchess.com/forums/1/message.html?322394 >>(about recommandable settings etc., as the defaults would be nonsens to test) >>Blown in the wind. :-) >If the SSDF would do as you suggest and Chessmaster wouldn't top their list they >would be criticised for: > >- not using the right book >- using CB's UCI, which is undoubtedly faulty >- using the secret Chessbase autoplayer, that as everyone knows makes Fritz win >all its matches. >- not playing manually with the original CM9000, feasible since the SSDF tried >to get Deep Blue. > >Etc. You name it. > >What you propose would mean putting the SSDF in a lose-lose situation. It would be a compromise... I think, the computerchess fans would appreciate it more than to have The King 3.23 not in the list at all. Now it's definetely a "lost" situation, because The King 3.23 simply is missing at all. I wouldn't demand manual tests, nowadays (except a few important chess computers or handhelds, occasionally). Also, I'd be quite confident that the number of people who would expect too much, and than eventually would critizise in the way you mention, is small. Currently, Rank #7 can mean 4th-best program (Shredder, Fritz, Junior, X). Important would be IMO, to care very well for the details mentioned before starting the tests: - cooperation with Wb2Uci and "King in Fritz" experts, to ensure proper setup - generate a "good quality" book, maybe from an up-to-date public game collection (I would not recommend a very big move depth, but that may be a matter of discussion, or J.d.Koning could recommend something) - minimum setting changes are selectivity = 12 and larger hash size (usually 32+ MB) - I'd recommend king safety 150/150 in addition to that, because most successfull custom personalities have a higher ks setting in common - since King doesn't use it's own endgame table during the search, the Nalimov access of the Fritz GUI basically can replace the CM tables with the same result (AFAIK both are DTM "distance to mate" endgame tables). It is important for computerchess that SSDF can continue to be the testing authority that it is (or was...), and not to test a major program properly, or not at all, is a constant problem since CM8000. Btw. another problem is, that the software publishers seem to be required to "send in" the programs. I understand that the SSDF people should at least get the progs for free in return for all the important testing work they do, but OTOH, this puts a big question mark behind the word "independant" IMO... Life is full of compromises. :-) The big information value of the SSDF list is beyond doubt, with or without The King! It just would even be a bit better with The King 3.23, than without... (And *much* better for the King fans :-)) Regards, Mike Scheidl
This page took 0.01 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.