Author: Georg v. Zimmermann
Date: 06:54:00 07/11/02
Go up one level in this thread
[D]8/8/8/2pK3p/8/5N2/1P4pq/5Rbk w - -
Sunsetter DuronMobile800,
Rb1 after 3 secs and sticks to it, for Kc6 it takes almost half a minute, the
rest is found quickly again.
If you do well here this might even mean that your engine's pruning can still be
dramatically improved :)
Georg
^^ @@@@@@@@@
^^ ^^ @@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@@@ ^^
@@@@@@@@@@@@@@@@@@@@
@@@@@@@@@@@@@@@@@@@@
~ ~~ ~ ~ ~~~~~~~~~~~~~~~~~~~~ ~ ~~ ~~ ~
~ ~~ ~~ ~~ ~~ ~~~~~~~~~~~~~ ~~~~ ~ ~~~ ~ ~~~ ~ ~~
~ ~~ ~ ~ ~~~~~~ ~~ ~~~ ~~ ~ ~~ ~~ ~
Sunsetter C2experimental
(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 8/8/8/2pK3p/8/5N2/1P4pq/5Rbk w - -
analyze
9 -61 7 11833 f1b1 h5h4
10 -61 20 29485 f1b1 h5h4
10 -60 22 30456 f1a1 !!
10 4 24 31146 f1a1 h5h4 a1b1 h4h3
10 5 26 32208 f1c1 !!
10 9 27 32641 f1c1 h5h4
11 9 48 78857 f1c1 h5h4 d5e4 h4h3 c1c2 c5c4 f3h2 h1h2
11 10 57 97487 f1a1 !!
11 24 59 99706 f1a1 h5h4
11 25 64 106386 f1e1 !!
11 25 72 121895 f1d1 !!
11 29 73 123208 f1d1 h5h4
12 29 102 190597 f1d1 h5h4
13 9 199 416415 f1d1 c5c4 d1a1 h5h4 a1c1 h4h3
13 10 284 604449 f1b1 !!
>>
>>In this position Chessmaster 8000 (running on an AMD Athalon 1800+ system) set
>>on default with a selective search of 6 takes 3'7" to find Rb1 with a score of
>>3.45.
>>
>>Chessmaster set on 0 selective search only takes 2'41" for the same problem.
>>
>>Chessmaster set on 12 selective search cannot find the Rb1 move even after a
>>half hour.
>>
>>This is what Graham Burgess says about the position in his book "The Mammouth
>>Book of Chess," page 394.
>>
>> "Unfortunately, the standard algorithms to prune the variation tree are very
>>likely to prune out the winning line, as the moves only make sense when you have
>>seen to the end of the line.
>> After more than 7 hours, Junior 6 (using tablebases, which one would expect to
>>help speed up the assessments in some of the sidelines) considered 1 Rb1 to be
>>winning, but it also thought 1 Rd1 to be equally good! Crafty 17, after several
>>hours, opted for 1 Re1 or Ra1, and considered the position in either case
>>(correctly) to be drawn, but did not see the winning line.
>> Fritz 6, also using tablebases and with its Selectivity option set to zero,
>>found the solution, including 1 Rb1 and 4 Ka8 in less than three minutes. In
>>fact, it takes the same time without tablebases, so this isn't a significant
>>factor here.
>> The same engine, with a normal selectivity setting, is far less successful.
>>1 Rb1! c4 (1...h4 2 Kc6 h3 3 Kb7 c4 comes to the ssame thing.)
>>2 Kc6! h4 3 Kb7! h3
>> Even at this point, when the calculation is not so deep, some of the engines
>>still take a while to find White's next move (Junior 6 and Hiarcs 7.32 are
>>quickest - about 10-15 seconds to find the win; others, including Fritz 6 and
>>Crafty 17, take longer, though Fritz 6 with Selectivity set to zero finds it in
>>a split second).
>>4 Ka8!
>> Only when this position is reached do most of the engines' assesments jump
>>dramatically in White's favour."
>>
>>I would be interested in how Chessmaster 9000 handles this problem.
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.