Author: Uri Blass
Date: 05:34:25 10/03/02
Go up one level in this thread
On October 03, 2002 at 08:16:55, Andreas Guettinger wrote: >On October 02, 2002 at 23:54:16, Robert Hyatt wrote: > >>On October 02, 2002 at 18:35:06, Uri Blass wrote: >> >>>On October 02, 2002 at 16:22:44, Hartmut Woldeit wrote: >>> >>>>On October 02, 2002 at 14:07:59, Uri Blass wrote: >>>> >>>>>On October 02, 2002 at 13:02:08, Hartmut Woldeit wrote: >>>>> >>>>>>Hallo, >>>>>> >>>>>>it's no new information that Crafty1714 is one of the best Crafty-versions. >>>>>>Better than most of his followers. >>>>> >>>>>I do not know it. >>>>>I only know that there is not big difference and you need hundreds of games to >>>>>prove. >>>> >>>>Hallo Uri, >>>> >>>>I do not know "hundred of games". I'm watching engine-engine-games. >>> >>>I did not watch games of Crafty versions but saying that Crafty17.14 is clearly >>>better than later crafty seems to me an insult to Bob who continue to work on >>>crafty. >> >> >>Note that I don't take it as an insult. I've heard this (and other such >>things) more than once. The problem is that it is very difficult to play >>two programs against each other and conclude which is best. Often the >>result will be different than if each of the programs is played against a >>group of _other_ programs... >> >>I have played 17.x vs 18.15 and I noticed that it won more than it lost >>(the 17.14 version mentioned). However, I then played them _both_ on ICC >>and noticed very little difference against other computers, with 18.x doing >>just a little bit better. however, against humans, 18.x was significantly >>better, most likely the result of endgame knowledge that has been added... >> >>It is just another example of how hard it is to say X is better than Y beyond >>a shadow of a doubt. If you only want to play X vs Y then it is not so hard >>to show that. But if you want to compare X to Y in _other_ ways, it becomes >>much messier... > > >It's all the same discussion over and over again. Some people are just fixed on >the results of engine - engine matches. They don't care if e.g. there is not >much endgame knowledge necessary for such games, because the other engine >doesn't know anything about engines. The goal of a programmer should be the >performance against humans, not the performance in engine-engine games. Why? I see no reason that the goal of the programmer should be what you ask for. There are a lot of people who are interested in engine-engine games. I'm sure >that engines like Hiarcs or Crafty etc. are not weaker than the top shots >Tiger15 or Deep Fritz7 in engine-human games. Just because the programmers are >not interested as much in tuning their engines against other engines simply >doesn't mean they perform weaker against humans. > >Andreas I believe that endgame knowledge can help also in engine-engine games. There are cases when engines do mistakes in endgames. I believe that the anti symmetric evaluation help crafty to get good results against humans but I do not like antisymmetric evaluation and I suspect that it may be a problem at 120/40 that is not played on ICC. Uri
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.