Author: Uri Blass
Date: 16:14:23 04/07/04
Go up one level in this thread
On April 07, 2004 at 18:52:37, Dieter Buerssner wrote:
>On April 07, 2004 at 17:42:07, Dieter Buerssner wrote:
>
>>Phantastic by Patzer. I wonder, what magic he is doing here. I first thought,
>>very aggressively extending pawn endgame positions (when the root position is
>>not a pawn endgame) should help. But it is still rather deep, when the Qs are
>>already from the board. Very impressive.
>
>Seems to work to aggressively extend pawn endgames. I added about 4 lines of
>code to Yace. When the root position is not a pawn endgame, and the current
>position is a pawn endgame, extend 1/2 ply. This is additionally to some other
>extension I already have, when trading down to a pawn endgame. Now I get:
>
> 46419606 1:00.0 0.60 13t 1.Qd6+ Kf5 2.Qd7+ Kf6 3.c6 Qe1 4.Qh3 Qa5 5.Qc8
> Qe1 6.Qd8+ Kf5 7.Qd5+H Kf6H 8.Qd4+H Ke7H 9.c7H
> Qg3+H 10.Kg1H Qe1+H {HT} {0}
> 62198227 1:19.4 0.61 13t+ 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 e2+ 4.Kxe2 Kg6 5.c6
> Kf5 6.g4+ Ke6 7.Ke3 Ke7 8.Ke4 f5+ 9.gxf5 Ke8
> 10.f4 g4 11.c7 Kd7 12.f6 Kc8 13.Kd4 g3 14.Ke5
> g2 15.f7 g1=Q 16.f8=Q+ Kxc7 {80}
> 64671720 1:22.4 1.00 13++ 1.Qxe3 fxe3 2.Kg1H Kf6H 3.Kf1H Ke6H 4.Ke2H Kd5H
> 5.Kxe3H f5H 6.c6H Kxc6H 7.f4H g4H 8.Kd4H Kd6H
> 9.g3H Ke6H 10.Kc5H Ke7H 11.Kd5H Kf6H 12.Kd6H
> Kg7H 13.Ke5H Kg6H 14.Ke6H Kh6H 15.Kxf5H {HT}
> {80}
> 66202720 1:24.4 2.00 13++ 1.Qxe3 fxe3 2.Kg1H Kf6H 3.Kf1H Ke6H 4.Ke2H Kd5H
> 5.Kxe3H f5H 6.c6H Kxc6H 7.f4H g4H 8.Kd4H Kd6H
> 9.g3H Ke6H 10.Kc5H Ke7H 11.Kd5H Kf6H 12.Kd6H
> Kg7H 13.Ke5H Kg6H 14.Ke6H Kh6H 15.Kxf5H {HT}
> {80}
> 74338752 1:33.5 7.00 13++ 1.Qxe3 fxe3 2.Kg1H Kf6H 3.Kf1H Ke6H 4.Ke2H Kd5H
> 5.Kxe3H f5H 6.c6H Kxc6H 7.f4H g4H 8.Kd4H Kd6H
> 9.g3H Ke6H 10.Kc5H Ke7H 11.Kd5H Kf6H 12.Kd6H
> Kg7H 13.Ke5H Kg6H 14.Ke6H Kh6H {HT} {0}
> 85856005 1:46.3 8.00 13t 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 Kxc6 7.f4 g4H 8.Kd4H Kd6H 9.g3H
> Ke6H 10.Kc5H Ke7H 11.Kd5H Kf6H 12.Kd6H Kg7H
> 13.Ke5H Kh6H 14.Kxf5H Kh7H 15.Kf6H {HT} {80}
> 87070103 1:47.9 8.00 13. 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 Kxc6 7.f4 g4 8.Kd4 Kd6 9.g3 Ke6
> 10.Kc5 Ke7 11.Kd5 Kf6 12.Kd6 Kg7 13.Ke5 Kh6
> 14.Kxf5 Kh7 15.Kf6 {HT} {80}
> 98130094 2:00.7 8.00 14t 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 Kxc6 7.f4 g4 8.Kd4 Kd6 9.g3 Ke6
> 10.Kc5 Ke7 11.Kd5 Kf6 12.Kd6 Kg7 13.Ke5 Kg6
> 14.Ke6 Kh6H {HT} {0}
> 124259369 2:34.5 8.00 14. 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 Kxc6 7.f4 g4 8.Kd4 Kd6 9.g3 Ke6
> 10.Kc5 Ke7 11.Kd5 Kf6 12.Kd6 Kg7 13.Ke5 Kg6
> 14.Ke6 Kh6 {HT} {0}
> 138859976 2:51.9 8.40 15++ 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 f4+ 7.Kd3 Ke6 8.Kc4 Kd6 9.Kb5 g4
> 10.fxg4 f3 11.gxf3 Kc7 12.f4 Kc8 13.g5 Kb8
> 14.g6 Ka8 15.c7 Ka7 16.c8=N+ Kb8 17.g7 Kb7 {500}
> 143707762 2:57.8 9.40 15++ 1.Qxe3 fxe3 2.Kg1H Kf6H 3.Kf1H Ke6H 4.Ke2H Kd5H
> 5.Kxe3H f5H 6.c6H Kxc6H 7.f4H g4H 8.Kd4H Kd6H
> 9.g3H Ke6H 10.Kc5H Ke7H 11.Kd5H Kf6H 12.Kd6H
> Kg7H 13.Ke5H Kg6H 14.Ke6H Kh6H {HT} {0}
> 188002510 3:48.9 10.00 15t 1.Qxe3 fxe3 2.Kg1 Kf6 3.Kf1 Ke6 4.Ke2 Kd5
> 5.Kxe3 f5 6.c6 Kxc6 7.f4 g4H 8.Kd4H Kd6H 9.g3H
> Ke6H 10.Kc5H Ke7H {HT} {0}
>
>For general positions, the idea might need some tuning, here it seems to work
>exceptionally well.
>
>Regards,
>Dieter
I guess that you also have some special extensions in pawn endgame that are not
in Yace padderborn.
Programs cannot find that white is winning after Qxe3 fxe3 in reasonable time
based on my testing.
You found >0.60 before Qxe3 in less than 80 seconds.
I have slower hardware than your hardware(A1000) but I do not think that it is
more than 3 times slower and I could not find >0.60 with some programs even
after 4 minutes.
I tried Fritz8 and yace padderborn(I do not have most of the tablebases but I
have the KPP vs Kp tablebases).
Here is analysis by Fritz8.008
New game
[D]8/5p2/6k1/2P3p1/8/4pP2/6PK/8 w - - 0 1
Analysis by Fritz 8:
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 Kd5
² (0.41) Depth: 7/14 00:00:00 14kN
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 Kd5
² (0.41) Depth: 8/16 00:00:00 18kN
2.Kg1--
= (0.13) Depth: 9/18 00:00:00 25kN, tb=1
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2
= (0.09) Depth: 9/18 00:00:00 33kN, tb=1
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5
= (0.13) Depth: 10/19 00:00:00 50kN, tb=2
2.Kg1 Kf6 3.Kf1 Ke5 4.g4 Kd5 5.Ke2 Kxc5 6.Kxe3 Kd5 7.f4 f6 8.Kf3 gxf4
= (0.16) Depth: 11/21 00:00:00 90kN, tb=15
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 Kd5 5.Kxe3 f6 6.f4 Kxc5 7.fxg5 fxg5 8.Ke4
² (0.41) Depth: 12/24 00:00:00 174kN, tb=111
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 Kd5 5.Kxe3 f6 6.f4 Kxc5 7.fxg5 fxg5 8.Kf3 Kd5
² (0.41) Depth: 13/26 00:00:01 269kN, tb=256
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 f4+ 6.Kd3 Kd5 7.c6 Kxc6 8.Ke4 Kd7 9.Kf5
² (0.59) Depth: 14/27 00:00:02 571kN, tb=697
2.Kg1--
² (0.31) Depth: 15/28 00:00:03 759kN, tb=878
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 15/28 00:00:04 1047kN, tb=1333
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 16/29 00:00:07 1725kN, tb=2285
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 17/33 00:00:11 2665kN, tb=3882
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 18/33 00:00:18 4551kN, tb=6380
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 19/37 00:00:29 8570kN, tb=10262
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 20/37 00:00:51 16831kN, tb=16212
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.g4 Kxc5 7.gxf5
= (0.00) Depth: 21/41 00:01:36 34486kN, tb=25850
2.Kg1 Kf6 3.Kf1 Ke5 4.Ke2 f5 5.Kxe3 Kd5 6.c6 Kxc6 7.f4 g4 8.Kd4 Kd6 9.g3 Ke6
10.Kc5 Kf7 11.Kd5
= (0.06) Depth: 22/46 00:03:57 98431kN, tb=57522
(Blass, Tel-Aviv 08.04.2004)
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.