Author: Thorsten Czub
Date: 15:47:42 10/02/02
Go up one level in this thread
On October 02, 2002 at 12:54:34, Marc van Hal wrote: >It is is standard set too 48MB >But I did set it to 96MB ok. than the hash is not the reason. >It must play good chess! right. >It is also for this reason that chessknowledge is necasery to improve >chessprograms ! >Instead of being able to beat a certain program. >Or solving epd diagrams. correct. >Some ideas to improve chess programs could be >1)Maybe it could be intressting to start all over again and start with >pawnstructures as the main object and only then start twiggeling with the pieces >and/or their valeus. >like 4pawns on Whiteskingside and the king on the kingside versus 3pawns on >Blacks kingside and the kingside is a draw put the pawns to the queenside and >leave the kings on the kingside and it's a win. > >It will avoid manny unintresting material gaining in that way. right. >2.}But more intresting is the usage of pawnstructures to build up a position >Or make strong outpost for the pieces this also is a way of decreasing the >mobilety of your oponents pieces >Only Amir took something over from that. rebel XP has this knowledge too. ed build in 3 new knowledge features: Center Outpost KS = king safety. these new 3 features in the evaluation function of rebel XP have been switched on with alpha stage of rebel xp. see my rebel XP diary where i give example games of the 3 styles. the only problem when you implement NEW KNOWLEDGE is: you must tune again the weigthings. this is a very time investive and difficult thing. IMO Ed spared this time and instead tuned against century4 eng-eng play. this way he found a style that beats century4, but IMO the weigthings of the 3 new knowledge bases are completely unbalanced and make rebel.eng default play very lousy. in my style macheide i switched the 3 new kowledge on like in the alpha stage and set the weigthings NEW. of the material. of the pawn stuff. and of all other values. This is exactly what makes my style play better than rebel.eng default. >3.} At the time it is even of importance which are the right kingsafety setting >I think they should be increased a litle more so it will use it in the right way >Rather then freezing by a threatening check > >Infact it also is for this reason why a pawnstorm can be so effective against >chessprograms. done in my style. >The program will not take advantage of it by trying to weakening the >pawnstructure of it's oponent so the oponents has an easy way to contineu that >plan. realized in my style. >4.}It also could be verry intresting to make a program which can play closed >positions verry wel just for experimenting the reward will later on be great. >Because the positional understanding will be verry good. > >Then you have the term isolany how to play against them >Most of the time by blocking it with a piece preferably a knight CSTAL had specific isloani knowledge and played IMO very good with it, blocking, pushing or attacking it depending if playing with or against the isolani. >And which isolany >because it might be a positional weaknes but it also can serve as a tactical >weapon as seen in many Gary Kasparov games >Or even more intressting the qualety sacrefice just too improve the >pawnstructure for Tactical purposes as seen in some games by R Ponnomariov i have overworked the material values of rebel XP as well. I guess you would like my style. it works in all the fields you talk about BETTER. ONLY ONE THING can IMO made better in my style. that is the search is still not omtimized enough- for that you need special knowledge how rebels search works.
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.