Author: Ed Schröder
Date: 01:16:22 11/26/01
Go up one level in this thread
On November 26, 2001 at 03:45:26, Uri Blass wrote:
>On November 25, 2001 at 20:58:06, Ed Schröder wrote:
>
>>On November 25, 2001 at 17:05:54, Mike S. wrote:
>>
>>>On November 25, 2001 at 15:14:16, Victor Fernandez wrote:
>>>
>>>>(...)
>>>
>>>>Yes, an example: an ending with few pieces .
>>>>
>>>>Kotov-Botvinnik
>>>>[D]8/8/4b1p1/2Bp3p/5P1P/1pK1Pk2/8/8 b - - 0 1
>>>>The correct move is ...g5 !!! Few programs play this.
>>>
>>>You're right. This position was used as #18 in the BS-2830 computer test suite.
>>>I don't have detailed results available at the moment, but Shep lists #18 as
>>>unsolved on his site:
>>
>>Unsolvable?
>>
>>My latest one needs only 18 seconds.
>>
>>Ed
>>
>>==========================================
>>
>>00:00:04 13.00 0.73 1..Kg3 2.Be7 Bf7 3.Kb2 Be6 4.Kc3
>> Bf7 5.Kb2 Be6 (3)
>>
>>00:00:10 14.00 0.73 1..Kg3 2.Be7 Bf7 3.Kb2 Kf3 4.Bc5
>> Kg4 5.Be7 Kf5 6.Bc5 Be6 7.Bd4 Kg4
>> 8.Bf6 Kf5 9.Bd4 Kg4 10.Bf6 Kf5 (7)
>>
>>00:00:18 14.15 0.73 1..g5
>>00:00:18 14.15 0.79 1..g5 2.fxg5 Kg3 3.Kxb3 Kxh4 4.Kc3
>> Kxg5 5.Kd4 Kf5 6.Be7 Bf7 7.Bd6 h4
>> 8.Be5 (18)
>>
>>00:00:28 15.00 0.82 1..g5 2.fxg5 Kg3 3.Kxb3 Kxh4 4.Kc3
>> Kxg5 5.Kd4 Kf5 6.Be7 Bg8 7.Bd8 Bf7
>> 8.Be7 Bg8 9.Bd8 Bf7 10.Be7 (18)
>>
>>00:01:06 16.00 0.89 1..g5 2.fxg5 Kg3 3.Kxb3 Kxh4 4.Kc3
>> Kg3 5.Kd4 h4 6.Bd6 (35)
>
>I think that the second move of black in the pv is wrong.
>I believe that after 1...g5 2.fxg5 Kg3 3.Kxb3 the position is a draw
You are right about 2..Kg3.
However this is common for chess programs and not such a big deal, actually
the above case is fascinating. A chess engine might sac a pawn and produce
a wrong main-line as a first result and I often see it labelled as, "playing
the right move for the wrong reason". That is a wrong way of seeing things.
In the above case based on eval there obviously is a good reason to sac a
pawn, so the program knows something important is going on and although the
engine hasn't figured it all out it still decides it is worthwhile to sac.
>The question is how much time does Rebel need to find 2...d4+?
Of course that is the crucial question, the program must see all the right
moves after the sac to justify its speculative sac.
Rebel needs 1:13 to find 2..d4+ which is an acceptable time. Things would
be different for me if Rebel had needed 10 minutes or more to find the
right continuation, then obviously eval is too optimistic and need some
changes.
Ed
========================================================
00:00:09 14.00 0.89 2..Kg3 3.Kxb3 Kxh4 4.Kc3 Kg3 5.Kd4
h4 6.Bd6 Kg2 7.Ke5 Bg8 8.Kf6 h3
9.Be5 (6)
00:00:18 15.00 0.95 2..Kg3 3.Kxb3 Kxh4 4.Kc3 Kg3 5.Kd4
h4 6.Bd6 Kg2 7.Ke5 Bf7 8.Kf6 Bg8
9.Be5 Kf3 (11)
00:00:37 16.00 0.73 2..Kg3 3.Kxb3 Kxh4 4.Kc3 Kxg5 5.Kd4
Kf6 6.Bd6 Kf5 (23)
00:01:13 16.10 0.83 2..d4+ 3.Bxd4 Kg3
00:01:58 17.00 0.88 2..d4+ 3.Bxd4 Kg3
00:04:42 18.00 1.02 2..d4+ 3.Bxd4 Kg4 4.g6 Kxh4 5.Kd3
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.