Author: Thorsten Czub
Date: 01:48:32 12/16/01
Go up one level in this thread
On December 15, 2001 at 08:00:18, Uri Blass wrote: >On December 15, 2001 at 06:19:39, Thorsten Czub wrote: >>this is the game and the passage you want to talk about. you don't like the tal >>score of 5.54. Why ? Tal sees a big advantage and gives a high score. >>How much is a win worth ? 0.9 pawns ? 0.8 ? 1 pawn ? 2 pawns ? >The point is that it is impossible to see a win and it is possible only to guess >a win Chess System Tal is NOT created to measure accurate, if this would be possible somehow, what the value of a position is. WHO can say this ? We can all only guess, only if it would be a mate or trasnposition into a tablebase, we can say precise. Chess System Tal evaluates the position, or certain parts in the position. And it measures how STRONG the chances are to win for black. You seem to mix this up with the old paradigm point of view to try to score the position by counting pieces and giving a bonus for positional. but this is not the way cstal works. it gives a very high score for the situation white is in. no matter if it WORKS or NOT. And this is excatly the point. The CD-Fritz version said it is -0.04, and this was ONE reason it lost the game. Because if you play this position as if this position is a draw, you will lose. CSTal is playing this position as if it is a win (winning scores indicate this very good, or ?) and when you know you win, you play different then playing for a draw. >A guess is not something that you can be sure about and it is better to be more >careful(+1 for black seems to me more logical) and at least I expect humans to >be afraid that the white king may somehow escape Kf2-e1-d2 when the material >advantage of white(bishop against 2 pawns) can be decisive. it is not better to be careful- whats this for a rule ? if you want to create a plan, and you think you have a better position (CSTAL thinks) you should NOT be careful. Otherwise you cannot create a plan. the "beeing careful" is the reason todays programs of old paradigm do not plan, because they are careful not to lose a pawn, not to lose a game. but this stops them from playing chess. instead they count material. and this is the difference between cstal and other programs. as i always told you. for years by now. >>We talk about chess. what your "chess programs" do is they count the material >>and say: 0.8 for black. Is this chess ? I don't think so, your "chess programs" >>do not play chess. they emulate it, and they do it very weak. >No YES. >If you count the material you are going to see that white has a bishop for 2 >pawns after capturing e7 so programs consider some positional advantage for >black. now relate the SOME POSITIONAL ADVANTAGE they give to the score CSTAL gives. and you see the difference. because THEY do play the way they score. and that is wrong. because this way they cannot play good, they can only play like somebody beeing very very careful not to lose. >You forgot the fact that white has 2 bishops against 1 bishop of black so by >only material it is possible to see bishop for the 2 pawns that is 1 pawn >advantage for white. wow. ><snipped> >>the move is ok. Hiarcs would also play f5. > >The move may be winning but I do not see why it is better than the game move. ok. >You are right that the main line is wrong but I doubt if the win for black is so >easy. why is YOUR program not seeing that its main line is bullshit although it computed 45 minutes ? i tell you. because your program has no idea about the things it is doing there. it is a pocket calculator. it is a system immanent problem. it does not know about chess. it does not even try to know. its enough to show ONE move in the main line, the rest are stupid moves. but how will such a machine plan ? never.
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.