Author: Uri Blass
Date: 00:06:27 01/24/05
Go up one level in this thread
On January 24, 2005 at 02:35:22, Uri Blass wrote:
>On January 24, 2005 at 01:40:44, Will Singleton wrote:
>
>>[d]8/5p2/6p1/5k1p/5P1P/6PK/5rB1/8 b - - bm Rxg2
>>
>>Amateur finds this in < 3 sec, faster than some others. A null-move issue?
>>
>>14 255 198 911657 Rc2 Bd5 f6 Bf7 Rb2 Bd5 Re2 Bc6 Rd2
>>15 255 236 1091636 Rc2 Bd5 f6 Bf7 Rb2 Bd5 Re2 Bc6 Rd2
>>15 328 277 1280278 Rxg2 Kxg2 Kg4 Kf2 Kh3 Kf3 f5 Kf2 Kh2 Kf3 Kg1 g4 hxg4
>>16 264 289 1341561 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf4 Kg2 Kxf5 Kxg3
>>17 334 312 1455185 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf4 Kg2 Kxf5 Kxg3
>>18 344 340 1592589 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf2 Kg4 Kg2 f4
>>19 348 389 1826593 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf2 Kg4 Kg2 f4
>>20 348 457 2147297 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf2 Kg4 Kg2 f4
>>21 431 569 2673404 Rxg2 Kxg2 Kg4 Kf2 Kh3 f5 gxf5 Kf3 f6 Kf2 Kg4 Kg2 f4
>
>No null move because almost no program use null move pruning in pawn endgames.
>
>I guess that it is evaluation issue and most programs do not have special
>evaluation for pawn endgame and they need some plies to evaluate the following
>lines as more than +2.5 for black
>
>Rxg2 Kxg2 Kg4 Kf2 Kh3 Kf3 f5 Kf2 Kh2 Kf3 Kg1 g4
>
>Movei needs 6 plies after the following line to see a score that can convince it
>to play Rxg2.
>
>Uri
I can add that I guess that some pruning based on evaluation probably prevent
movei to find it in few minutes but this pruning only delay the solution and
does not prevent it and public version find it at depth 18.
Analyzing suggest that Movei even prefers some losing KPP vs K that is evaluated
only as -2.26 like the following line and it makes finding Rxg2 harder.
Things will probably be better after I use nalimov tablebases for debugging my
evaluation function(I already got permission to use them few weeks ago and I
believe that the main benefit of them is to debug the evaluation(I delayed
implementing knowledge about endgame and one of the reason was not having
permission from nalimov to use the nalimov tablebases).
[Event "Edited game"]
[Site "URI-AMD"]
[Date "2005.01.24"]
[Round "-"]
[White "-"]
[Black "-"]
[Result "*"]
[FEN "8/5p2/6p1/5k1p/5P1P/6PK/5rB1/8 b - - 0 1"]
[SetUp "1"]
{--------------
. . . . . . . .
. . . . . p . .
. . . . . . p .
. . . . . k . p
. . . . . P . P
. . . . . . P K
. . . . . r B .
. . . . . . . .
black to play
--------------}
{
--------------
. . . . . . . .
. . . . . p . .
. . . . . . p .
. . . . . k . p
. . . . . P . P
. . . . . . P K
. . . . . r B .
. . . . . . . .
black to play
--------------
}
1... Rxg2 2. Kxg2 Kg4 3. Kf2 Kh3 4. f5 gxf5 5. Kf3 f6 6. Kf4 Kg2 7. Kxf5
Kxg3 8. Ke4 Kxh4 9. Kf4 Kh3 10. Kf5 Kg3 11. Ke4
*
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.