Author: Mark Young
Date: 23:33:20 08/24/98
Go up one level in this thread
On August 24, 1998 at 17:48:13, blass uri wrote: > >On August 24, 1998 at 16:38:37, Mark Young wrote: > > >>I agree. I don’t understand people’s needs to test only the engine. When you >>play a chess program, you play the whole program. And in my opinion the opening >>play is a big percentage of any programs overall strength. And the sooner we >>come to realize this, and stop treating it like an ugly step child, the sooner >>we can address this area of chess programming that has been neglected by so >>many. > >I agree the opening play is important but I think that a good program >can find the right moves by analysis and not by an opening book. > >If this is not the case people can get the program out of theory and win >because the book cannot have all the lines. > >It is possible today to get out of book every program by a small number of >moves. >kasparov did it against deeper blue when he played with black >1.e4 c6 2.d4 d6 >This is the reason I believe that books are not very important. > I was not talking about opening books, but opening play in general. I think programmers need to go much farther then just opening books. We see steps being made now in this area like book learning coupled with position learning, but this tends to only help after the program has made a fool of itself a few times in a given position. Even with this type of learning the program is still very vulnerable to the same theme but a different position. This is a programs greatest weakness. Now from time to time I will see how much I can spike a chess programs rating online, or what Bob and others call a cyborg. When I do this, I do it through opening play only. Like keeping the program away from setups that I feel that the program does not play well. Forcing the program to sidestep dubious pawn sacs in the opening. Keeping the program from repeating bad openings. And other common senses things so that the program can start off the game on at least an equal footing. I do this when playing computers and grandmasters alike. I have been able to spike a chess programs rating over 100+ rating points doing this. And I am just a C player. So I think there should be away for programmers to program a little common sense into the computers opening play. To avoid some very silly and stupid pitfalls that the chess programs right now are very vulnerable to. Right now the chess programmers are just scratching the surface in this area of computer chess programming. And it is costing them much in potential rating points. >I believe there will not be a very big difference between the engine without >opening book and the engine with an opening book. >I expect the difference in rating to be less than 100 if the program is a good >program. > >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.