Author: blass uri
Date: 10:44:35 06/10/00
Go up one level in this thread
On June 10, 2000 at 11:54:17, Ed Schröder wrote:
>On June 10, 2000 at 04:23:17, blass uri wrote:
>
>>On June 10, 2000 at 01:01:54, blass uri wrote:
>>
>>>On June 09, 2000 at 18:11:37, Ed Schröder wrote:
>>>
>>>>On June 09, 2000 at 05:39:21, blass uri wrote:
>>>>
>>>>>On June 09, 2000 at 01:11:03, Paulo Soares wrote:
>>>>>
>>>>>>On June 09, 2000 at 00:08:07, Oliver Roese wrote:
>>>>>>
>>>>>>>On June 09, 2000 at 00:03:02, Paulo Soares wrote:
>>>>>>>
>>>>>>>>On June 08, 2000 at 22:49:28, Oliver Roese wrote:
>>>>>>>>
>>>>>>>>> [Event "NRW 4er-Pokal"]
>>>>>>>>> [Site "Germany"]
>>>>>>>>> [Date "2000.06.04"]
>>>>>>>>> [Round "?"]
>>>>>>>>> [White "Grimm, S."]
>>>>>>>>> [Black "N.N"]
>>>>>>>>> [Result "1/2-1/2"]
>>>>>>>>> [FEN "8/4k1pp/5p2/P1p1p3/2Qq4/1P4P1/5PKP/8 b - - 0 1"]
>>>>>>>>>
>>>>>>>>> 1...Qxc4 ({This position occured last sunday in a team event. It was the
>>>>>>>>>last game. I and a team colleague of mine were "kibitzes". Our man had white.
>>>>>>>>>Obviously black is in trouble here. But black has the chance here to swap the
>>>>>>>>>queens and go into a pawn ending. He pondered a while and finally he played
>>>>>>>>>1...Qd6. The game went on with 2.Qe4 and finally he lost. My teammate took me
>>>>>>>>>aside: "What do you think about 1.Qxc4?" he asked. "Well, white opens the
>>>>>>>>>kingside with g4 and wins.", i said. "Wrong!" he told me. We made a blindgame,
>>>>>>>>>me playing white and he took me along up to the 18th move in this analysis. "I
>>>>>>>>>see.", i said, "But maybe white has some other possibilities." I went to the
>>>>>>>>>board, puzzled a while, and came back. "What about f4?" i asked. He hesitated
>>>>>>>>>shortly and showed me the winning line for black. In a few seconds he refuted a
>>>>>>>>>line that costed me several minutes! I went to the board again, trying hard to
>>>>>>>>>refute Qxc4. Eventually i came back. "What about g4, Kg3 and then f4?" He
>>>>>>>>>ponderd a while and then quick as a flash he showed me the refutation again. "So
>>>>>>>>>this guy is tough!", i concluded for myself. After the end of the game, we
>>>>>>>>>showed them what he had found. They were both surprised. Finally we were
>>>>>>>>>interested to see if there are any winningchances in the resulting QQ-endgame.
>>>>>>>>>Should i tell you who told us 19..Kc2 ? This is one of the few examples, there
>>>>>>>>>a skilled human can outperform current hard/software. If they can solve it all!
>>>>>>>>>I tried a few minutes with crafty, but gave up at the end. Can your computer
>>>>>>>>>come up with 1...Qxc4! ? Oliver Roese } 1...Qd6 ) 2.bxc4 Kd6 3.g4 ( 3.f4??
>>>>>>>>>exf4 4.gxf4 f5 -+ 5.Kg3 g6 6.Kh4 ( 6.h4 h5 ) 6...h6 ) 3...g6 4.Kf3 ( 4.Kg3?? Kc6
>>>>>>>>>5.f4 Kb7 6.fxe5 fxe5 7.Kf3 Ka6 8.Ke4 Kxa5 9.Kxe5 Kb4 10.Kd5 g5 -+ ) 4...f5
>>>>>>>>>5.gxf5 gxf5 6.Kg3 Kc6 7.Kh4 h6 8.Kh5 e4 9.Kh4 Kb7 10.Kg3 Ka6 11.Kf4 Kxa5 12.Kxf5
>>>>>>>>>Kb4 13.Kxe4 Kxc4 14.f4 Kb3! 15.f5 c4 16.f6 c3 17.f7 c2 18.f8=Q c1=Q = 19.Qb8+
>>>>>>>>>Kc2! 20.Qc7+ Kd1 21.Qxc1+ Kxc1 22.Kf5 Kd2 23.Kg6 Ke3 24.Kxh6 Kf4 1/2-1/2
>>>>>>>>
>>>>>>>>I think 3.Kf3! is a winner move. Black have no chances.
>>>>>>>>
>>>>>>>>1... Qxc4 2. bxc4 Kd6 3. Kf3!
>>>>>>>>
>>>>>>>>Paulo Soares, from Brazil
>>>>>>>
>>>>>>>But _why_ do you think that?
>>>>>>
>>>>>>I was wanting to place the answer quickly in the forum and I analyzed the
>>>>>>position believing in the evaluation of the program, without giving the
>>>>>>necessary time for a good evaluation. My mistake, sorry.
>>>>>>
>>>>>>1... Qxc4 2. bxc4 Kd6 3. Kf3 f5 {!}
>>>>>>4. g4 g6 5. gxf5 gxf5 6. Kg3 Kc7 7. Kh4 h6 8. Kh5 e4 9. Kh4 Kb7 (9... f4 10.
>>>>>>Kg4 e3 11. fxe3 fxe3 12. Kf3 {+-}) 10. Kg3 Ka6 11. Kf4 Kxa5 12. Kxf5 Kb4 13.
>>>>>>Kxe4 Kxc4 14. f4 Kb3 15. f5 c4 16. f6 c3 17. f7 c2 18. f8=Q c1=Q {=}
>>>>>>
>>>>>>Paulo
>>>>>
>>>>>The problem is that programs do not know to evaluate unstoppable pawns
>>>>>
>>>>>try the following position that can happen after
>>>>>3.Kf3 f5 4.g4 g6 5.gxf5 gxf5 6.kg3 Kc6 7.Kh4 h6 8.Kh5 e4 9.a6 Kb6 10.Kxh6 f4
>>>>>
>>>>>
>>>>>[D]8/8/Pk5K/2p5/2P1pp2/8/5P1P/8 w - - 0 1
>>>>>
>>>>>Programs cannot see at evaluation time that black is winning because of
>>>>>unstoppable pawn.
>>>>>
>>>>>If I give them to analyze at 1 ply depth they give a big advantage for white.
>>>>>
>>>>>part of the programs know that h2 is unstoppable but they do not know that e4 is
>>>>>unstoppable and that e4 can be a queen faster than h2
>>>>>
>>>>>Uri
>>>>
>>>>You can solve a problem in 2 ways, a) with chess knowledge or b) by search.
>>>>Very often search is much more powerful than adding complex and processor
>>>>time stealing complex chess knowledge. This position is a typical example
>>>>that search is the right solution to play the best move.
>>>>
>>>>Ed
>>>
>>>The problem is not this position but the position many plies before.
>>>
>>>programs did a mistake in the evaluation of the position many plies before
>>>probably because they did not give the right static evaluation to this position.
>>>
>>>I believe that it is important to see big positional scores at evaluation time
>>>because of this reason.
>>>
>>>Uri
>>
>>Here is another test position that shows better the importance of knowledge
>>about passed pawns.
>>
>>I composed this position
>>
>>[D]Q3q3/4k3/8/5p2/2p5/8/P6P/4K3 w - - 0 1
>>
>>avoid Qxe8+
>>
>>Century1.2a(p450) can see in less than 5 minutes that Qxe8+ is losing but cannot
>>see a better move at least in the first 23 minutes.
>>
>>I believe that a good evaluation can help to see that white is losing after 2
>>plies based on the squares of the pieces.
>
>Maybe the opposite is true. It depends per program how the programmer
>looks at things. For this position I would say that having 2 outside
>passers usually is a great advantage and as such is rewarded by a chess
>program. If so then this position is an exception to the rule. And the
>end-game is full of exceptions much more than the mid-game.
>
>Ed
Having 2 outsides passed pawns is an advantage but having 2 advanced passed pawn
is also an advantage.
The problem is to know which advantage is bigger.
I agree that it is not a simple problem and I understand the fact that
programmers prefer to waste more time about other problems.
I still believe that it is possible to see it at evaluation time by an array
64*64*64 of distance to promotion.
I think that white distance to promotion is 6 for white because white need to do
at most x+6 pawns moves when x is the number of king moves of black and there
are cases when white cannot avoid x+6 moves.
By the same consideration black distance for promotion is 5 so black is closer
to promotion.
Uri
This page took 0.01 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.