Author: Dieter Buerssner
Date: 10:39:41 03/22/05
Go up one level in this thread
On March 22, 2005 at 09:54:25, Roy Brunjes wrote:
>I get a different result and am a bit puzzled as to why.
Don't worry. It is a bit hash luck involved to get this
>I used Yace (version
>unknown but downloaded directly from the Yace website as most current Windows
>version).
It showed the version, see the comment further down.
>I have loaded all of the 3-4-5 piece TBs from Bob Hyatt's ftp site.
I don't think so, see again a comment further down.
>Engine: Yace Paderborn (512 MB)
here is the version
> 37/49 27:27 +4.66 1.h7+ Kxh7 2.Kxa7 Kg6 3.Ka6 Kg5 4.Ka5 Kg6 5.Kb4 Kg5 6.Kc5
>Kg6 7.Kd4 Kh5 8.Kxe3 Kh4 9.Kxe4 Kg4 10.Ke5 Kg5 11.e4 Kg4 12.Kd6 Kg5 13.e5 Kf4
>14.e6 (2.182.112.503) 1324
>interior recognizer probes 329271, found 116529 egtb probes 1234829, found
^^^^^^^^^^^^^^^^^^^^^^^^^
>957080 max_depth 49^
^^^^^^^
With all 3,4,5 men TBs installed, these two numbers will be identical.
>A couple of things are strange here:
>
>1. Look at how much slower things progress on my system. I am using a Centrino
>1.8 GHz laptop (certainly not a whole lot slower than a 3 GHz P4). Perhaps the
>results above had much of the TBs already in RAM from previous accesses to them?
Yes, possible. Actually, on your computer less nodes were used to show a
significant positive score. But the search started very slow due to the many TB
accesses.
>2. At one point, Yace sees an eval of 11.46 and with additional searching it
>drops its eval to 4.66 . That was not seen in the post above either.
Yes, this is due to some hash luck. Yace searched rather deep, but it really is
not deep enough yet, to eval the pos as +10. Look at this PV I got:
Mat50 30. 1.h7+ Kxh7 2.Kxa7 Kg6 3.Ka6 Kg5 4.Ka5 Kg4 5.Kb6
Kf5 6.Kb5 Ke6 7.Kc4 Ke5 8.Kc5 Ke6 9.Kd4 Kf5
10.Kd5 g6 11.Kc5 Kg5 12.Kc6 Kg4 13.Kd6 Kf5
14.Kd5 g5 15.Kc5 Ke5 16.Kc4 Kf4 17.Kd4 Kf5
18.Kd5 g4 19.Kc5 Ke5 20.Kc4 Ke6 21.Kd4 Kf5
22.Kd5 Kf4 23.Ke6 Kg5 24.Ke5 Kg6 25.Kxe4 Kg5
26.Ke5 Kh5 27.Ke4 Kh4 28.Kxe3 Kg5 29.Ke4 Kh5
30.Kf4 Kh4 31.e3 Kh5 32.Kxg3 {EGTB} 32...Kg5!
33.e4! Kh5 34.Kf4! Kg6! 35.Kxg4 Kf6! 36.Kf4!
Ke6 37.g4 Kf6 38.g5+ Kg6 39.e5! Kg7 40.e6 Kg6!
41.Ke5! Kg7! 42.Kd6! Kg8 43.e7 Kf7! 44.Kd7!
Kg6! 45.e8=Q+! Kxg5 46.Qe4! Kh5 47.Qg2 Kh6
48.Ke6! Kh7 49.Kf6 Kh8 50.Qg7#! {1001}
After a7, the next pawn is captured at ply 55! Still no chess engine could eval
this as +11, I guess. Transition to TB is only after 63 plies.
You can try http://www.moq.dk/download/yace9987.zip (Not linked from the web
page). It is quite a bit better in most pawn endgames. Set up with all bitbases
(see readme.txt from your current version - it is not included here, IIRC) and
for aggressive TB-usage, it shows a mate score on my computer in 20 seconds. It
does not show the total PV yet. The above PV was spit out after 28 minutes. GUIs
I know will not be able to display it, however.
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.