Author: Bernhard Bauer
Date: 00:36:06 06/28/99
Go up one level in this thread
On June 27, 1999 at 02:08:18, blass uri wrote:
>
>On June 27, 1999 at 01:57:51, kurt wrote:
>
>>Fritz 5.32 can not find the winning move Ra8 [Capablanca] in 20plys.
>>White:Kd7,Rb8,b5
>>Black:Ka7,Nc7,b6
>>White to move
>
>of course Fritz5.32 cannot see it because it is a null mover.
It's not that easy.
A program is not doomed to fail on this position only because it's a
null mover. One may ask: why does it use null move in a position which
contains only 6 pieces?
Have a look at another 'null mover' Crafty.
8-> 0.20 1.33 1. Rf8 Nxb5 2. Kc6 Na3 3. Rf3 Nc2 4.
Rf7+ Ka6 5. Rb7
9 0.33 1.37 1. Rf8 Nxb5 2. Kc6 Nd4+ 3. Kd5 Nb3
4. Kc4 Nc5 5. Kb5 Nd3
9 0.38 ++ 1. Ra8+!!
9 0.39 7.63 1. Ra8+ Kb7 2. Ra7+ Kxa7 3. Kxc7 Ka8
4. Kxb6 Kb8 5. Kc6 Ka7 6. Kd6
(4) 9-> 0.41 7.63 1. Ra8+ Kb7 2. Ra7+ Kxa7 3. Kxc7 Ka8
4. Kxb6 Kb8 5. Kc6 Ka7 6. Kd6
(4) 19-> 3:40 9.60 1. Ra8+ Kxa8 2. Kxc7 Ka7 3. Kc6 Kb8
4. Kxb6 Kc8 5. Ka7 Kd7 6. b6 Ke6 7.
b7 Kf5 8. Kb6 Ke4 9. Kc5 Kd3 10. Kd5
Kd2 11. b8=Q
So Crafty finds Ra8 after 0.4 sec and the evaluation goes up at least to ply 19.
Using end game table bases leads to:
8-> 0.61 7.83 1. Ra8+ Kb7 2. Ra7+ Kb8 3. Rxc7 Ka8
4. Kc6 Kb8 5. Rc8+ Kxc8 6. Kxb6
9 0.63 ++ 1. Ra8+!!
9 0.63 Mat18 1. Ra8+ Kxa8 <EGTB>
9-> 0.77 Mat18 1. Ra8+ Kxa8 <EGTB>
time=0.98 cpu=467% mat=2 n=21876 fh=93% nps=22322
ext-> checks=1207 recaps=0 pawns=0 1rep=56 thrt:235
predicted=0 nodes=21876 evals=1886
endgame tablebase-> probes done=2241 successful=247
hashing-> trans/ref=60% pawn=99% used=0%
SMP-> split=71 stop=8 data=5/64 cpu=4.58 elap=0.98
mate in 18 moves.
So the position is a mate in 18 found in a second.
My point is: Null movers doesn't have always to fail. Null move is a good
heuristic, but you have to use it 'right'. Here I admit that we may currently
not know what that mean.
Kind regards
Bernhard
>black is losing after Nxa8 Kc8 because of the fact that black has to move and
>not because white has a threat.
>Fritz prunes line with no threats.
>
>If you want a program to solve problems then do not buy a program that is a null
>mover and that there is no option to turn the null move off.
>
>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.