Author: Uri Blass
Date: 22:11:38 12/11/00
Go up one level in this thread
On December 11, 2000 at 19:49:35, Terry Ripple wrote: >On December 11, 2000 at 19:21:56, G. R. Morton wrote: > >>Century 3 is touted as a superb (the best?) positional chess player, which my >>own non-scientific tests on well known GM actual game positions seem to bear >>out. Are there are any formal positional test results? Does any strong player >>(besides the reviews at the Rebel site) have an opinion Century 3’s positional >>play as the best (whether or not it may make it to the SSDF top few)? >------------- >The SSDF list only proves that what ever program is on the top of the list was >the best at that time in beating other chess programs under similar conditions >but doesn't prove that it's the best at winning against very strong players at >tournament time controls of 40 moves/2 hours. > >Rebel Century 3.0 was fine tuned to play it's best against strong human >opponents and there for it is possible that it might not be the best program >when it is matched against one of the other top programs. I believe that one of >the reasons for this is that the more knowledge a program has it can't think as >deep into the position because it is a slower thinker compared to the tactical >monster "Fritz" that has less knowledge to slow it down, so it has more time to >think deeper because of it's faster search. ( Just my opinion ) > >Regards,Terry >Regards,Terry I disagree. My common sense tells me that the more knowledge the program has it can calculate deeper. One reason is because having knowledge can help to prune the correct lines. I know that a lot of programs(Rebel is not one of them but Fritz is one of them) use null move pruning. A program with good evaluation function can evaluate threats better so it can prune the right lines better. A program with bad evaluation function will not see a lot of positional threats. It is going to prune them and the result will be that they will be weaker in tactics because I believe that a lot of tactical ideas are based on positional threats if you search to smaller depth. Another reason that I expect a program with more knowledge to be better in tactics is the fact that it can evaluate positions with material disadvantage more correctly so in order to see the tactical idea they do not need to see the final position. For example if your program know that 2 pawns in the 6th rank are winning against a rook in the endgame it does not have to see that the pawns queen. 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.