Author: chris sergel
Date: 16:16:12 03/09/00
Go up one level in this thread
On March 09, 2000 at 02:35:50, Jeremiah Penery wrote: >On March 08, 2000 at 18:52:32, chris sergel wrote: > >>During my last chess lesson I studied a position from a game >>Nimzovich-Capablance. > >[D]2r5/1p3pk1/1q2p1p1/p2pPnPp/2rN1P2/P1PR3Q/1P2R1KP/8 w - - > >>After the lesson, out of curiosity, I let my computer program analysze >>the position for a whole hour. At this time, my program wanted to play >>NxN (which loses in just a few moves). > >Can you tell us how it loses? I couldn't see it. I don't mean to say it loses in sense of blunder or quick tactical refutation. I mean it loses because black has all the play - on the queenside, on the c file, and because the h pawn will be forever poisoned (computer would not take h pawn either). Well I didn't see that Nxf5 lost either, (though I am not that good, but I do have one IM norm so I am not that weak either), Nimzovich didn't see it lost, and neither did program(s). It lost because in this bad position, (note that black can penetrate queenside with b pawn, or, is some variations that didn't happen, with Queen only while leaving b pawn at home), white can still defend by just being stubborn. When white does penetrate the queenside, he can trade off some pieces, relieve pressure, and hope that he can defend inferior endgame. White is still alive. But he can only do this if that is all he has to do. As soon as trades Knights, he not only has queenside to worry about, but also the center (Re4 and Rc4 are coming) and even weakened Kingside. Can't do it. Not against a player like Capablanca who understood positional play like he did. >> This is not so bad, because also >>Nimzovich played that move, and he was fine player. The problem with the >>program is that it thought white had the advantage of almost 3/4 of a pawn. >>If Junior would evaluate the position the same, then there are still >>problems that programs have in evaluating this kind of position. >>With careful preparation it might be possible >>for a human to defeat even a program like Junior. >>So everyone here is sure Junior will win, but since human has had Junior >>for a while, and being a much better player than I am, can find this >>sort of weakness - maybe the outcome of this match is not so certain. >>I suspect it all might depend on Junior's opening book and whether or >>not human has taken the time to probe for this sort of positional weakness. >>Or it could be that my program is weak and that Junior would evaluate >>the position differently. >>Someday a program will come along that would suggest moving Rd2 and >>evaluate it's position as -1.25 or something. That would be impressive. > >Here's what my modified Crafty thinks after a while: > >15-> 110:10 0.09 1. Nxf5+ gxf5 2. Qh4 Re4 3. Red2 Rh8 > 4. h3 Qd8 5. Rd4 Qc7 6. Kg3 Kg6 7. > Kh2 Rg8 8. a4 Rxd4 9. Rxd4 Qc5 > The program and the human games differed quite a lot, but most of the ideas are similar. The same ideas are made, but often at slightly different times. The human game continued 2. Qf3 (same idea as Qh4 - protect pawn), Re4 3. Rd2 (also played) Kg6 (also played) 4. Rd4 Rxc5 5. Kg3 Rxd4 6.c3xd4 Qc4 (penetration) 7. Kg2 b7-b5 (even now my program preferred white - maybe I should buy a different program - I would shudder to think of it playing Capablanca) 8. Kg1 b5-b4 9. a3xb4 a5xb4 10. Kg2 Qc1 11. Kg3 Qh1 12. Rd3 Re1 Resigns >So it likes Nxf5 more than other moves, but it's barely more than a draw score. >And the score has gone down on almost every iteration so far. I'll see what it >is after 16 and maybe 17 a bit later, if I can stand to let it run that long. :) It looked to me like your program understood the position better than mine did, but I think it started to lose track by moving Queen from Queenside. I don't understand that. Black has already started the gears, he just has to wait for the meat to be ground.
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.