Author: Dieter Buerssner
Date: 13:02:09 05/24/05
Go up one level in this thread
On May 24, 2005 at 15:43:27, Heiner Marxen wrote:
>On May 24, 2005 at 15:32:41, Dieter Buerssner wrote:
>
>> But you still would not have a real proof. I assume
>>Chest and company store the whole position inside the HT, to be really sure.
>
>Yes, you are right, Chest stores the complete board in its HT entries,
>(including castling rights & ep status), for exactly this reason:
>to be absolutely sure. Currently, Chest's HT entries are 40 bytes large.
Heiner, thanks for confirming this. I have a debug mode, which will store the
complete board including ep and castling. When showing a mate, we can happily
ignore the other severe problems of HTs considering 50 moves rule and
repetitions. I guess, I can prove mate in 18 in this position with the secure HT
scheme in about 15 minutes. But it needs user interaction (although, really no
chess knowledge is needed - just wait a second, do the suggested move on the
board, until you see the mate score, go back "slowly"). Of course, no way to
prove a shortest mate in this way in general. Without the whole position stored
in HT, I can get a mate in 18 in about 10 minutes. That debug mode makes the
engine about 20% slower; and of course less hash entries can be used.
With the unsecure HTs, after 10 minutes of interactive analysis, no TBs, I got:
1127 0.020 Mat18 12t 1.Ke8 Bd4 2.h8=Q+H Bxh8H 3.Bf8+H Kh7H 4.Rb7+H
Kg8H 5.h6H Bb2H 6.Bg7H a1=QH 7.h7+H Kxh7H
8.Bxb2+H Kg6H {HT} {-151}
8998695 4.887 Mat18 12. 1.Ke8 Bd4 2.h8=Q+ Bxh8 3.Bf8+ Kh7 4.Rb7+ Kg8
5.h6 Bb2 6.Bg7 a1=Q 7.h7+ Kxh7 8.Bxb2+ Kg6
9.Bxa1H Kf5H 10.Rb4H Ke6H 11.Rb5H Kd6H 12.Bd4H
Kc6H 13.Rc5+H Kb7H 14.Kd7H Kb8H 15.Rc4H Ka8H
16.Kc6H Kb8H 17.Ra4H Kc8H {HT} {850}
8998696 4.887 Mat18 13t 1.Ke8 Bd4 2.h8=Q+H Bxh8H 3.Bf8+H Kh7H 4.Rb7+H
Kg8H 5.h6H Bb2H 6.Bg7H a1=QH 7.h7+H Kxh7H
8.Bxb2+H Kg6H 9.Bxa1H Kf5H 10.Rb4H Ke6H 11.Rb5H
Kd6H 12.Bd4H Kc6H 13.Rc5+H Kb7H 14.Kd7H Kb8H
15.Rc4H Ka8H 16.Kc6H Kb8H 17.Ra4H Kc8H {HT}
{850}
101162769 57.002 Mat18 13. 1.Ke8 Bd4 2.h8=Q+ Bxh8 3.Bf8+ Kh7 4.Rb7+ Kg8
5.h6 Bb2 6.Bg7 a1=Q 7.h7+ Kxh7 8.Bxb2+ Kg6
9.Bxa1 Kf5 10.Rb4 Ke6 11.Rb5 Kd6 12.Bd4 Kc6
13.Rc5+ Kb7 14.Kd7 Kb8 15.Rc4 Ka8 16.Kc6 Kb8
17.Ra4 Kc8 {HT} {850}
I did not check the PV, which is mostly reconstructed from the HT, and may be
garbage. The score should be correct always (unless there is a serious bug in
the engine). When showing this score, the engine will be in a state, where it
always can show the "best" move (any move, that will at least guarantee a mate
in 18) very fast to any reply of the defender.
Cheers,
Dieter
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.