Computer Chess Club Archives




Subject: Re: Forced mate so solve...(proof number search)

Author: leonid

Date: 18:11:50 02/07/01

Go up one level in this thread

On February 06, 2001 at 12:12:58, Angrim wrote:

>On February 06, 2001 at 06:28:19, leonid wrote:
>>On February 06, 2001 at 02:57:32, Angrim wrote:
>>>On February 05, 2001 at 07:23:58, leonid wrote:
>>>>If you like to solve forced mate position full of "depleted uranium"...
>>>>[D]NKBN4/QRQ3rR/2Qb2r1/N1b1nb2/B2nbbbB/1RNqqqq1/8/2N2k2 b - -
>>>>Please, indicate your result.
>>>Useing proof number search, with no chess knowlege other than which
>>>moves are legal and the definition of checkmate, on an Athlon 650mhz:
>>>:proved that move c5xa7 wins, 15 turns
>>>:PN:402888 evals, 9241 expands, 56 max ply,  1.96 seconds
>>>not the shortest win, but at least its fairly quick.
>>Can you give more detailed description of your search program? I am curious
>>since its base is probably identical to my mate solver. Mine use nothing more
>>but two parameters (not special value for each piece) that are legality of moves
>>and presence of check. It is everything that program need in order to find mate
>>or draw.
>It uses a fairly generic proof number search, you might try searching
>for either "proof number search" or "pn-search" on the web, there are several
>papers on the topic available.  PN-search is a general tree proving algorithm
>that is domain independant, although it seems to do better for domains
>such as game tree searches where there is some concept of mobility being good.
>If you didn't recognise the name of the search method, it is fairly
>unlikely that you are useing it.

Since you wrote about "pn-search" and "proof number search", I found already few
times those expressions used in few places. If you could in few words indicate
what those terms signify it will be of real help, at least me.


This page took 0.03 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.