Author: Uri Blass
Date: 12:09:12 03/01/01
Go up one level in this thread
On March 01, 2001 at 14:18:13, Robert Hyatt wrote: >On March 01, 2001 at 08:21:38, Sune Larsson wrote: > >>On February 28, 2001 at 23:19:12, Robert Hyatt wrote: >> >>>On February 28, 2001 at 13:48:00, Sune Larsson wrote: >>> >>>>On February 28, 2001 at 13:00:53, Robert Hyatt wrote: >>>> >>>>>On February 28, 2001 at 11:12:09, Sune Larsson wrote: >>>>> >>>>>> >>>>>> [D]4n3/1b3k1p/p2p2p1/3Pp3/2P3P1/6PP/2BB4/6K1 w - - 0 1 >>>>>> >>>>>> This is Hort-Ciocaltea, Budapest 1973. Only two moves are required >>>>>> to immobilize the black knight and pave the way for the victorious >>>>>> march of the white king to the deserted queenside: >>>>>> >>>>>> 1.g5! Bc8 2.g4! and black's position is ripe for resignation. >>>>>> >>>>>> Test: The above move order should be rewarded with big +scores for white. >>>>>> >>>>>> Sune >>>>> >>>>> >>>>>There may be more than one winning plan here. Crafty thinks that Ba5 wins a >>>>>pawn, for example... eval = +1.26 after 18 plies, almost 2 minutes: >>>>> >>>>> 17 1:01 0.99 1. Ba5 Ke7 2. Kf2 Bc8 3. Kf3 Bd7 4. >>>>> Ke4 Kf6 5. Bb4 Nb6 6. Bd3 Nc8 7. c5 >>>>> dxc5 8. Bxc5 a5 9. d6 >>>>> 17-> 1:09 0.99 1. Ba5 Ke7 2. Kf2 Bc8 3. Kf3 Bd7 4. >>>>> Ke4 Kf6 5. Bb4 Nb6 6. Bd3 Nc8 7. c5 >>>>> dxc5 8. Bxc5 a5 9. d6 >>>>> 18 1:40 1.26 1. Ba5 Ke7 2. Kf2 Bc8 3. Kf3 Bd7 4. >>>>> Ke4 Kf6 5. Bb3 h5 6. Bb4 hxg4 7. hxg4 >>>>> Ke7 8. c5 dxc5 9. Bxc5+ Kf6 10. g5+ >>>>> Kxg5 11. Kxe5 >>>>> 18-> 1:53 1.26 1. Ba5 Ke7 2. Kf2 Bc8 3. Kf3 Bd7 4. >>>>> Ke4 Kf6 5. Bb3 h5 6. Bb4 hxg4 7. hxg4 >>>>> Ke7 8. c5 dxc5 9. Bxc5+ Kf6 10. g5+ >>>>> Kxg5 11. Kxe5 >>>> >>>> Interesting line. Can Crafty win the above position after 11.-Kg4? >>>> >>>> Sune >>> >>> >>>Good question. A brief note about chess programs. When they produce a PV >>>as above, the first move is the one they play. The second move is the one >>>they expect, but it is less accurate as it is the result of a search that is >>>one ply shallower than the search to choose the _first_ move. By the time >>>you reach the end of the PV, it is _highly_ likely that those moves will never >>>be played in a real game. The last move is based on a 1 ply search, for >>>example, and no one would consider that to be reasonable chess... >>> >>> >>>IE when a GM gives me a variation to look at, _every_ move generally is >>>a sound move. When a computer gives me a variation, the deeper the move >>>in the variation, the less likely it is to be the best move. >> >> Okey, from your statement I conclude that we cannot trust the Crafty line >> above. Furthermore it doesn't even come close to GM Hort's choice or >> the theme of immobilizing the black knight with 1.g5! and 2.g4! >> >> Sune > > >Here is a quote from a very good GM on ICC (Not Roman this time). > >"I notice that when crafty is whispering analysis, the first move or two >are _always_ very strong. But the further down in the PV you go, the less >sense the moves make at times. why is this?" > >It is based on the reasoning I gave... I rarely fault the move the thing >produces, as the moves are good enough to give human GM players fits. But I >am _certain_ that if Crafty had to play the entire PV in a game, it would >end up in much difficulty... most of the time.. It is clear. I think that it is also the case with humans in a game because they care not to blunder in the first move but they care less not to blunder many plies after the root. It is not truth with GM analysis only because GM's do not give lines during the game and when they give lines after the game they have no reason to care more to be correct in the root position. I think that people are wrong to think that the main line is the computer analysis after the game. If they want their chess program to generate analysis of a position after the game then it is better to give their chess program to play some moves against itself. 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.