Author: Georg v. Zimmermann
Date: 03:42:07 06/29/02
Go up one level in this thread
On June 28, 2002 at 15:49:46, Dieter Buerssner wrote:
>On June 27, 2002 at 13:37:44, Georg v. Zimmermann wrote:
>
>>On June 27, 2002 at 13:20:56, Roy Eassa wrote:
>>
>>>How can white shorten this mate, which takes 18 moves? :
>>>
>>>1.Rxh7+ Kxh7 2.Qh5+ Kg8 3.Rxg7+ Kxg7 4.Bh6+ Kh7 5.Bg5+ Kg7 6.Qh6+ Kf7
>>>7.Qf6+ Kg8 8.Qg6+ Kh8 9.Bf6+ Rxf6 10.exf6 Qxe1+ 11.Kxe1 Nc2+ 12.Kf1 Ne3+
>>>13.fxe3 Rd7 14.Qe8+ Kh7 15.Qxd7+ Ne7 16.Qxe7+ Kh6 17.Qg7+ Kh5 18.Qg5#
>>
>>I think white can't, you are right. Move 6 and 7 can get lost because of HT.
>
>???
>
>What do you mean by this? I think, HTs can introdroduce many sorts of artifacts
>(especially in the lines shown), but they should never show a too short mate.
>You reported mate in 15 earlier.
With the normal way to store mates (moves to mate) in the hashtable and updating
the mate score in every search() call this can not happen, you are right. I was
experimenting with scoring mates as -MATE_IN_ONE+ply/2+1 , e.g. moves from the
root. This introduces that kind of HT artifact.
It probably isn't a good idea but I am searching for ways to get rid of other
HT/mate score related problems(see the funky PV below).
I am very impressed by how well Yace does on tactical positions even though its
kn/s are rather low. This might be the first position I have found where
Sunsetter is faster ;-)
Duron800:
d:\sunsetter-chess>Sunsetter.exe
^^ @@@@@@@@@
^^ ^^ @@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@@@ ^^
@@@@@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@@@@@
~ ~~ ~ ~ ~~~~~~~~~~~~~~~~~~~~ ~ ~~ ~~ ~
~ ~~ ~~ ~~ ~~ ~~~~~~~~~~~~~ ~~~~ ~ ~~~ ~ ~~~ ~ ~~
~ ~~ ~ ~ ~~~~~~ ~~ ~~~ ~~ ~ ~~ ~~ ~
Sunsetter C2
(c) Ben Dean-Kawamura, Georg v. Zimmermann
See http://www.d2d4.de/sunsetter.asp for more info.
Created 16777216 byte transposition table and 4194304 byte learn table.
setboard 3r1r1k/1b4pp/ppn1p3/4Pp1R/Pn5P/3P4/4QP2/1qB1NKR1 w - - 0 1
analyze
7 -516 9 8684 c1g5 d8d3 f2f4 c6d4
7 -515 11 9911 e2d1 !!
7 -500 14 10844 e2d1 d8d7 f2f4 b4d3 e1d3
7 -499 19 15130 h5h7 !!
7 0 23 18933 h5h7 h8h7
8 0 32 29687 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6c1 h7g8
h5g6 g8h8 g6h5 h8g7 <draw>
9 0 44 50461 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6c1 h7g8
h5g6 g8h8 g6h5 h8g7 <draw>
10 0 73 103768 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6c1 h7g8
h5g6 g8h8 g6h5 h8g7 <draw>
11 545 311 531626 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6g5 h7g8
h5g6
12 956 761 1406516 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6g5 h7g8
h5g6
13 29982 797 1466389 h5h7 h8h7 e2h5 h7g8 g1g7 g8g7 c1h6 g7h7 h6g5 h7g8
h5g6 g8h8 g5f6 f8f6 e5f6 b1e1 f1e1 b4c2 e1f1 c2e3 f2e3 b6b5 a4b5 a6a5 b5c6 d8e8
g6e8 h8h7 c6b7 h7h6 e8h8 h6g6 h8g7 g6h5 g7g5
Georg
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.