Author: blass uri
Date: 22:08:22 05/21/98
Go up one level in this thread
On May 21, 1998 at 21:54:07, Thorsten Czub wrote: >On May 21, 1998 at 19:40:07, Mark Young wrote: >> I think what Junior, Fritz 5, and Nimzo 98 have shown >>that its not amount of knowledege that important. Its understanding what >>positional knowledege that the program does not need. > >Why ? I don't see that these 3 programs are much stronger than >Hiarcs or Mchess. I do not believe that these 3 programs understand >chess better than Hiarcs or Mchess. Do you ? >>This leaner code gives >>the program its faster search rate. Giving the program greater search >>depth. With the greater search depth I think the programs sees over the >>board what positional has to be done. > >How can it see what positional stuff has to be done when its knowledge >about positional is very rudimental ? >Also - does it even HAVE knoweldge in deep lines ? >Or does it only have piece-square tables from the root ? > > > >>This may give Junior, Fritz 5, and >>Nimzo 98 a better more flexable positional understanding. > >Big words. How can you prove them ? >I have not seen Fritz showing positional understanding. >Maybe Junior, sometimes Nimzo. But Fritz5 ? yes fritz5 is showing sometimes positional understanding for example it found a5(move of Junior in the third game of the swiss tournament that Junior play) is a weak move Amir ban said he was surprised that fritz understood it while Hiarch did not understand the position. from my experience fritz5 knows how to attack when the opponent's king is in trouble even before it sees a tactical forced variation. in my nunn match(1000 seconds per move) it does not know to evaluate the position correctly like other programs (like Genius3 and Junior4.6) but it find good moves to attack the opponent's king (I posted some examples to it in the past). of course it is weak in positional understanding like all programs but its main weakness is not its static evaluation function but the fact that it does not see the position clearly for example in paris it lost against stobor because it did not understand it is a bad idea to trade queens but after it traded queens it understood stobor's position is better. I do not like fritz5 because they want to be number 1 by unfair means so I want fritz5 to be a weak program but I cannot ignore the facts. > > >> Then some >>static positional programmed knowledege that may help in some positions >>and kill you in others. > >STATIC POSITIONAL programs have a search-tree too. In opposite to the >others, they use their full knowledge within the tree. >I don't understand why you believe that knowledge chess programs are >weaker. >As I said, in positions where there is nothing to find, nimzo98 or Fritz >do stupid moves, mainly moving repetition moves. Moving back good >developed pieces. Doing kind of "null moves". They have no plan. No >idea. They only see something, when the opponent makes it possible. This >is pretty passive. >I don't think that having no plan is worth much. >For how long do you believe will Fritz5 be no.1 ? I believe a better idea is that a program will have more than 1 evaluation function and when there is nothing to find it will use more knowledge and when there is something to find it will be faster and use less knowledge. 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.