Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: If you like to solve a mate...

Author: leonid

Date: 17:24:31 03/08/01

Go up one level in this thread


On March 08, 2001 at 19:57:34, Dieter Buerssner wrote:

>On March 08, 2001 at 15:30:19, leonid wrote:
>
>>On March 07, 2001 at 22:34:41, Dieter Buerssner wrote:
>>
>>>On March 07, 2001 at 16:48:55, leonid wrote:
>>>
>>>>rkrR1q2/qqq1Q1Rq/nnQbN2Q/Qq2N3/3n2Q1/1Q1Q3b/2q3BB/Q6K w - -
>>>
>       nodes    time  score depth PV {mat. balance in centipawns at end of PV}
>>>   3735698  54.286  Mat07  2.  1.Qexc7+ Bxc7 2.Nd7+ Nxd7 3.Qaxc7+ Qxc7
>>>                               4.Qbxb5+ Nxb5 5.Qdxb5+ Nb6 6.Qbxb6+ Qxb6
>>>                               7.Qxb6# {350}
>
>>If you looked through this position by brute force, please say me what was
>>average nodes/second rate for this position.
>>
>>Recently, I speeded my mate solver and installed in the same time NPS counter.
>>Was surprised by finding that NPS is very low. It is usually somewhere between
>>35k and 120K on my AMD 400Mhz.
>
>I just want to confirm, that Uri's answers were correct. I included a line, that
>hopefully explains the output of Yace. Sorry, I was too lazy to include this in
>my first post. My hardware: AMD K6-2 475. I cannot remember the hash setting,
>but it probably was 20M. Also, I probably tested this position, while my
>soft-modem was running and eating some CPU cycles. Usually Yace gets about
>150000 nodes/s on my hardware in middlegame. In your "why only two queens?"
>positions, I only get half of this, or even less.

I even not looked into "usual" positions to find how many nodes/second my mate
solver reach. It should be much lower that for "heavy" positions that I usually
depose here.

Your NPS is very high compared with mine for sure. I was very surprised when I
found my numbers. You just never can say with certainty where you can expect
what numbers. Verification is the only secure way to find them. Later I still
expect to go with changes in my mate solver but I not expect that much change in
NPS figures.

>Without carefully analyzing this, I think, that most moves of the PV were found
>in quiescence search. This is also the reason, that the quiescence search tree
>of Yace can easily explode with many of the positions you posted. There are just
>too many checks and captures possible.

Usually I try to depose here positions that look like very heavy but have decent
branching factor. Mate also is never too far. Very seldom mate will be 9 moves
or beyond. Above position was mate in 8.

Cheers,
Leonid.


>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.