Author: Dieter Buerssner
Date: 09:51:19 04/09/04
Go up one level in this thread
On April 08, 2004 at 15:03:37, Anthony Cozzie wrote:
>Actually, there is some logic to this.
Yes.
>The more complex the position is, the easier it is to search to higher depth.
>
>In the original position, my guess is that the tree after Qxe3 xe3 is many times
>smaller than the original - so with the extension the trees are probably much
>more balanced in terms of nodes/child.
Yes. The only fear would be, that (mainly in non pawn endgame positions) the
general search depth is hurted (in special cases only, perhaps).
>Interesting to seem to some games of Yace v/Yace w extension.
I fear, that I would not have the patience until a significant difference is
found (only few games are decided in pawn endgames).
But I reviewed some difficult positions. Some of them, I already mentioned
earlier in this thread. Actually, all positions I tried were solved in under or
around one minute. Earlier, Yace needed many hours for some, or did not solve
them. By solved I mean a convincing score (> 5 for winning) and a decent main
line - not only finding the move. Let me repeat Simon Finn's challenge 2 1/2
years ago:
in CCC you wrote on September 28, 2001:
---
>I would be very impressed if any program can solve the following
>innocent-looking position (Grigoriev 1933) in a reasonable time. By "solve" I
>mean find most of the PV (up to 13. Ka8, say) with a winning score.
>
>[D]k7/8/pp6/2p4K/8/PPP5/8/8 w - - bm c4;c0 "Grigoriev 1933";pv "1. c4 Kb7 2. a4
>Kc6 3. Kg5 Kc7 4. Kf6 Kd7 5. a5 Kd6 6. Kf5 Kc7 7. Ke6 Kc6 8. Ke7 Kc7 9. Ke8 Kf8
>10. axb6 Kb7 11. Kd7 Kxb6 12. Kc8 Kb6 13. Ka8 a5 14. Kb8 a4 15. bxa4 Ka5 16. Kb7
>Kxa4 17. Kc6 Kb4 18. Kd5";
Sorry - I missed out a move - it should be 13. Kb8 Kc6 14. Ka8 a5 etc.
---
Note, there seem to be 2 more typos. 9...Kf8 probably should mean
9...Kc8. Also, 12...Kb6 is not possible (11... was Kxb6).
Yace can solve the challenge in 1 minute (without TBs). I used a rather inferior
setup, too, with very aggressively probed TBs (idea - I wanted to see a mate
score). After 1 hour I got a mate score (mate in 38). Much of the line was
hidded because it was from the HT. After more waiting, the whole line is shown
as:
1.c4 Ka7 2.a4 Kb7 3.Kg5 Kc7 4.Kg6 Kc6 5.Kf5 Kd6
6.a5 Kc7 7.Ke6 Kc6 8.Ke7 Kc7 9.Ke8 Kc6 10.Kd8
bxa5 11.Ke7 Kc7 12.Ke6 Kc6 13.Ke5 Kc7 14.Kd5
Kb6 15.Kd6 Kb7 16.Kxc5 Kc7 17.Kd5 Kd7 18.c5 Kc7
19.c6 Kb8 20.Kd4 Kc8 21.Kc4 Kd8 22.Kd5 Kc7
23.Kc5 Kd8 24.Kb6 Kc8 25.Kxa6 {EGTB} 25...Kc7!
26.Kb5! Kc8 27.Kc5 Kc7! 28.Kd5!! a4 29.bxa4!!
Kc8 30.Kd6! Kd8! 31.c7+ Kc8! 32.a5!! Kb7
33.a6+! Kb6 34.a7 Kb7 35.c8=R Kb6 36.a8=Q! Kb5
37.Qa3 Kb6 38.Rb8#! {1501}
Note that the ! and !! in the lines above are pure formal, in the
sense: !! means the only move that is winning; ! means that it is the
only move that leads to the shortest (for the attacker)/longest (for
the defender) mate. This info is taken from endgame databases.
The line is a transposition of the line given by Simon until 9. Ke8. If I input
Simon's moves instead, I get the same line as he gave. After 9...Kc8 (which is
meant above, I suppose, I get:
29093668 3:32.8 Mat25 23t 10.axb6 Kb7 11.Kd7 Kxb6 12.Kc8 Kc6 13.Kb8 Kb6
14.Ka8 a5 15.Kb8 Kc6 16.Ka7 Kc7 17.Ka6 Kc6
18.Kxa5 {EGTB} 18...Kc7 19.Kb5! Kd6! 20.Kb6!
Ke6 21.Kxc5! Ke7 22.b4 Kf7 23.b5 Ke7 24.b6 Kd7!
25.Kd5! Kd8 26.c5 Kc8! 27.c6! Kd8 28.Kd6! Ke8
29.c7! Kf8 30.c8=Q+ Kf7 31.Qe6+ Kg7 32.b7 Kh7
33.b8=Q! Kg7 34.Qbg8#!
So, a somewhat faster mate. My engine is not a mate proofer, so it
really cannot guarantee, that these are the fastest mates, and
possibly the first line I had shown leads to a shorter mate than the
second line. But that does not really matter, I think. I
hope, it fulfills the challenge (although 25 plies of the given line were
wanted; I gave some more plies, that are correct - in the
sense it is certainly winning).
It might also be interesting to note, that in the first line above,
white is still one pawn down after 30 plies.
For the brain of the earth challenge, see
http://members.aol.com/MuellerLamprecht/BrainsoftheEarthengl.html
All positions are solved without TB in under one minute (the 4th position is one
move into the Grigoriev study). Not what Frederic Friedel wrote about it:
"So the total solving time for Garry Kasparov in John Nunn's the Brains of the
World test was one hour and fifteen minutes. I have advised him to start
training for a trip into space. His results are the best we have received so far
-- by quite a big margin."
The last position, I also tried with the inferior setup with the aggressively
probed TBs:
80981342 4:34.2 9.86 20++ 1.Kh2 f6 2.Kh1 f5 3.Kg1 Kf6 4.Kf1 Ke5 5.Ke1 Kd5
6.Kd1 Ke5 7.Kc2 Ke4 8.Kc3 Ke5 9.Kd3 Kd5 10.Ke3
Ke5 11.Kf3 h5 12.Ke3 f4+ 13.gxf4+ Kd5 14.g3 Ke6
15.f5+ Kf6 16.Kd3 Ke7 17.f6+ Ke8 18.f4 Kf8
19.f7 Kg7H 20.f8=Q+H Kg6H {HT}
127797185 12:44.2 Mat45 20t 1.Kh2 f6 2.Kh1 f5 3.Kg1 Kf6 4.Kf1 Ke5 5.Ke1 Kd5
6.Kd1 Ke5 7.Kc2 Kf6 8.Kd2 Kg5 9.Kd3H Kh5H
10.Ke2H Kg4H 11.Ke3H Kg5H 12.Kf3H h6H 13.Ke3H
Kg4H 14.Kd2H Kg5H 15.Kd3H Kh5H 16.Ke2H Kg6H
{HT} {80}
Solution times without TBs:
1s 2s 5s 10s 20s
60s
bote.1: 1 (Kf4) 1 (Kf4) 1 (Kf4) 1 (Kf4) 1
(Kf4) 1 (Kf4) 0.290
[score > 5 after 0.3 s]
bote.2: 1 (Kb2) 1 (Kb2) 1 (Kb2) 1 (Kb2) 1
(Kb2) 1 (Kb2) 0.038
[score > 5 after 23 s]
bote.3: 1 (Kg5) 1 (Kg5) 1 (Kg5) 1 (Kg5) 1
(Kg5) 1 (Kg5) 0.011
[no draw score found without TBs, but -0.03]
bote.4: 1 (a4) 1 (a4) 1 (a4) 1 (a4) 1
(a4) 1 (a4) 0.014
[score > 5 after 23 s]
bote.5: 1 (a3) 1 (a3) 1 (a3) 1 (a3) 1
(a3) 1 (a3) 0.717
[score > 5 after 1.5 s]
bote.6: 0 (Kg1) 0 (f3) 0 (Kg1) 1 (Kh2) 1
(Kh2) 1 (Kh2) 18.088
[without TBs only 0.9 score after 1 min, but with correct PV. With TBs (and
bitbases) score +3 after 50 s, after 2 min > 8]
Summary: 5 5 5 6 6
6 19.158
Steve Maughan formula Sum(10.0*exp(-ln(2)*solution_time/60.0)) = 57.99
Regards,
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.