Author: Gordon Rattray
Date: 15:18:44 11/30/01
Go up one level in this thread
On November 30, 2001 at 16:50:14, Miguel A. Ballicora wrote: >On November 30, 2001 at 16:18:24, Christophe Theron wrote: > >>On November 30, 2001 at 16:13:12, Uri Blass wrote: >> >>>On November 30, 2001 at 14:35:37, Christophe Theron wrote: >>> >>>>On November 30, 2001 at 05:16:46, Jouni Uski wrote: >>>> >>>>>... or actually not! Gandalf gains a lot rating points and Tiger loses. >>>>> >>>>>Jouni >>>> >>>> >>>> >>>>It is not surprising, because Chess Tiger has been given only partial tablebases >>>>during this match. >>>> >>>>As pointed out somewhere else, because of this lack of TB some wins have been >>>>turned into draws. >>>> >>>>It would have been better to not give any TBs to Tiger at all. I'm sure the >>>>result would have been better. >>>> >>>>This result leaves a bitter taste. Thousands of hours of work (mine and the >>>>hours spend by the SSDF testers on this match) spoiled by an inferior setup. >>>> >>>> >>>> >>>> Christophe >>> >>>I believe that other programs also used exactly the same tablebases. >>> >>>Uri >> >> >> >>That's not the problem. The problem is not that unequal means have been used by >>both programs, the problem is that using partial TBs with Chess Tiger is an >>incorrect setup. >> >>It might not be a problem for another program, but for Chess Tiger it is. >>Instead of using partial TBs it would have been better to use no TB at all. >> >>What am I supposed to do? Should Chess Tiger refuse to run when partial TBs have >>been installed? > >Either that, warn the user if it is an incorrect setup, or handle the situation >(for instance, internally you detect the absence of some tables and ignore them >all together). No program should accept an incorrect setup, IMHO. But how to you define "incorrect"? What about hash table size? Choice of opening book and settings? Modes such as "aggressive" or "suicidal"? The tester has to know how to set these correctly. It may be correct from the program's point of view, while being incorrect with what the tester is trying to achieve (e.g. optimised strength). Gordon > >Regards, >Miguel > > > >> >> >> >> Christophe
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.