Author: Jeremiah Penery
Date: 15:36:06 08/21/03
Go up one level in this thread
On August 21, 2003 at 09:05:52, Omid David Tabibi wrote:
>On August 21, 2003 at 07:25:04, Uri Blass wrote:
>
>>On August 21, 2003 at 07:18:47, Uri Blass wrote:
>>
>>>On August 21, 2003 at 07:06:20, Omid David Tabibi wrote:
>>>
>>>>On August 21, 2003 at 06:12:12, Jouni Uski wrote:
>>>>
>>>>>[D]8/1p6/1Pp2N1q/p1Ppk2p/P3p3/3PPpPp/3K1P1P/1R6 w - -
>>>>>
>>>>>From CSS forum: Machine should, within a million hours, be able to find the draw
>>>>>in this position (with Ng4+). But Shredder 7.04 found it in one second!!
>>>>>
>>>>
>>>>It is not a draw. The black queen will penetrate into the white cvamp via a6.
>>>>For the past year I've been working on endgame blockage detection. I noticed
>>>>that Shredder has some very limited form of blockage detection, but it is very
>>>>inaccurate, and in many positions I tried, that inaccurate blockage detection
>>>>wreaked havoc to the search tree.
>>>
>>>Wrong
>>>
>>>It is a draw
>>>I also believed that it is not a draw but trying to win it
>>>against shredder I got the following game
>>>
>>>[Event "?"]
>>>[Site "?"]
>>>[Date "????.??.??"]
>>>[Round "?"]
>>>[White "New game"]
>>>[Black "?"]
>>>[Result "*"]
>>>[SetUp "1"]
>>>[FEN "8/1p6/1Pp2N1q/p1Ppk2p/P3p3/3PPpPp/3K1P1P/1R6 w - - 0 1"]
>>>[PlyCount "11"]
>>>
>>>1. Ng4+ hxg4 2. d4+ Ke6 3. Rh1 Qf8 4. Ke1 Qa8 5. Kf1 Qa6+ 6.
>>>Kg1 and black cannot make progress
>>>
>>>Uri
>>
>>Here is some analysis of shredder in a non draw position
>>Shredder believes that black advantage is very small
>>until seeing the win by search.
>>
>>I believe that shredder's score for maybe fortress position
>>is too small and +1 for black is more realistic for
>>a position that may be in significant part of the cases a win
>>for the stronger side.
>
>When the program is 100% certain about the existence of an unbreakable blockage,
>it should give it the draw score (0.00) regardless of material advantage. For
>example, in the following position Genesis is certain about the blockage and
>gives it 0.00 instantly:
In the original position given for this thread, Uri's line yields a draw, so
Shredder is correct there. My Crafty also gives a drawscore for Ng4 in that
position.
>[D]8/1k4pp/pP1pb3/P1p2pP1/2B2P2/3P3P/1K1P4/8 b - - 0 1
>
>
>depth time nodes score variation
> 1/ 4-> 0.02 0k 1.65 1...e6d7
> 2/ 6-> 0.02 0k 1.61 1...d6d5 2.c4b3
>
>Without the block detection knowledge, it is very hard to find the correct move
>1...Bxc4.
The block detection I wrote in Crafty a while back also sees a draw here right
away.
8/1k4pp/pP1pb3/P1p2pP1/2B2P2/3P3P/1K1P4/8 b - - 0 1
White(1): 8/1k4pp/pP1pb3/P1p2pP1/2B2P2/3P3P/1K1P4/8 b - - 0 1
analyze.White(1): end-game phase
clearing hash tables
time surplus 0.00 time limit 22.50 (3:30)
depth time score variation (1)
1 0.00 1.41 1. ... Bd7
1-> 0.01 1.41 1. ... Bd7
2 0.01 1.61 1. ... Bd7 2. Kc3
2 0.01 1.57 1. ... Bxc4 2. dxc4
2 0.01 1.46 1. ... d5 2. Bb3
2-> 0.01 1.46 1. ... d5 2. Bb3
3 0.01 1.43 1. ... d5 2. Bb3 h6
3 0.01 ++ 1. ... Bxc4!!
3 0.01 0.00 1. ... Bxc4 2. dxc4 g6
3-> 0.01 0.00 1. ... Bxc4 2. dxc4 g6
4 0.01 0.00 1. ... Bxc4 2. dxc4 g6 3. Kc3
4-> 0.01 0.00 1. ... Bxc4 2. dxc4 g6 3. Kc3
5 0.01 0.00 1. ... Bxc4 2. dxc4 g6 3. Kc3 Kc6
5-> 0.02 0.00 1. ... Bxc4 2. dxc4 g6 3. Kc3 Kc6
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.