Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: An hard test position

Author: leonid

Date: 13:35:31 08/07/00

Go up one level in this thread


On August 07, 2000 at 15:25:27, blass uri wrote:

>On August 07, 2000 at 15:02:24, leonid wrote:
>
>>On August 07, 2000 at 11:03:56, leonid wrote:
>>
>>>On August 07, 2000 at 08:30:35, blass uri wrote:
>>>
>>>>[D]6k1/6pp/1Q1pn1rr/p1p1p3/2P1P2q/1P3P1P/P2RN1PK/6R1 w - - 0 1
>>>>
>>>>How much time does your program need to find Rxd6?
>>>
>>>Four seconds.
>>>
>>>Probably this is not the best move but my program see only it in 8 and 10 ply
>>>brute force search. In 6 plys search it says Rook d2-d6.
>>>
>>>After every move by whites black likely to be sure to excute the mate.
>>>
>>>AMD 400Mhz. 8 ply - 3.5 second. 10 ply - 2 min. 45 second.
>>
>>
>>
>>My mistake! Had fixed idea that you indicated move by Queen.
>>
>>Exacte time and moves:
>>
>>6 plys - Rook d2-d6. 0.4 second.
>>
>>8 plys - Queen goes to d6. 3.5 second.
>>
>>10 plys - Queen - d6 2 min. 45 second.
>>
>>Search by brute force. Average NPS - 288k.
>
>Qxd6 is wrong.
>
>I believe that after Qxd6 Qf2 your program will do another mistake and play
>Qb8+(kh1 is the right move)
>
>after Qb8+ chessmaster(selevtive search=10) play Kf7 with evaluation of mate in
>15 moves
>
>Your program have no chance to see the mate because it needs to search 29 plies
>in order to see it(I assume no extensions)

Special part of my program do search for mate only 26 plys ahead. Can make it
see as much as I would like but I see this as highly unpractical.

When you demanded your mate search with Chessmaster, he was looking through
special rapid logic. By brute force it could go for ever. This is just this part
in my program (that search for mate) that is presumably the most efficent. I
remember that few positions against Chessmaster 4000 I tried around 5 years ago.
My program had very easy win in time. Don't know if now it is the same. What I
liked with Chessmaster (the same for Genius 2 that is more rapid) is that it
never lied to me and was able to search every position even the most strange
one.

>Chessmaster uses extensions so it can see it after 8 minutes on pentiumIII450
>
>I also believe that your program will not see that Qb8 is bad in a few minutes.

I just don't know. For this I must look into choice that was done on "second
stage". I remember that 8 plys deep program said that it make choice between 8
moves for 8 plys deep search, and 11 for 10 plys deep. When response is cristal
clear then on "second stage" no work is done. Nothing to choose from. Only one
move to execute.

>Part of the top programs need some minutes to see that Qb8+ is bad(chessmaster
>is an exception and has no problem to see that Kh1 is the best move after Qxd6
>Qf2.
>
>Your program is weaker because it does no extensions so it probably needs at
>least some hours.

Probably one kind of extensions I already have. This is for mate search. To be
sure that this is extension, I must to know how other programmers do it. Other
"extensions", or something like this, I will start probably very soon. As soon
as I will see that I can't speed any more basic part of my program. Extension
will only consume some basic speed and will never make it grow. This is why I it
is good to start with best speed before you will start spending it somewhere.
Recently I was ready to go into writing beyond basic speed but then I reached
some new speeding. This speeding was found on the place that I just was not able
to see before and for the reason that was indicated to me repeatedly (Tom, Bob)
and was rejected each time as baseless.

Leonid.

>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.