Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: One mate to solve.

Author: leonid

Date: 19:27:27 04/13/02

Go up one level in this thread


On April 13, 2002 at 20:04:10, Heiner Marxen wrote:

>On April 13, 2002 at 15:59:39, leonid wrote:
>
>>On April 13, 2002 at 15:15:07, John Merlino wrote:
>>
>>>On April 13, 2002 at 14:49:12, leonid wrote:
>>>
>>>>On April 13, 2002 at 14:36:31, John Merlino wrote:
>>>>
>>>>>On April 13, 2002 at 07:49:31, leonid wrote:
>>>>>
>>>>>>Hi!
>>>>>>
>>>>>>Recently I found that for many programs selective search for mate is pretty
>>>>>>different. At least, in few recent mate positions my default selective was
>>>>>>useless. I am curious to see how many other programs can solve this position by
>>>>>>selective. Mine was useless on this even if it look like to be ideal for
>>>>>>selective. Position in itself is not deep, or very difficult.
>>>>>>
>>>>>>[D]1qbqkbq1/QBRNBNQ1/1QnQpQn1/1q1RQ1p1/3rn3/2Q2Q2/Q6K/1r4r1 w - -
>>>>>>
>>>>>>Please indicate your result, never mind your way of solving this mate.
>>>>>>
>>>>>>Thanks,
>>>>>>Leonid.
>>>>>
>>>>>I don't know if it's the shortest mate, but it took Chessmaster a long time (on
>>>>>a PIII-733) to find Mate in 10 anyway (with some HUGE evals along the way), as
>>>>>the first seven moves are all non-checking moves:
>>>>
>>>>
>>>>So, actually Chessmaster took this position by selective and not that far from
>>>>shortest one. Pretty good!
>>>>
>>>>This position is mate in 9 moves.
>>>>
>>>>Cheers,
>>>>Leonid.
>>>
>>>I'm trying Chessmaster's Solve for Mate mode. It finished depth 8 in just a few
>>>minutes, but is taking quite a long time to finish depth 9. It still hasn't
>>>announced mate in 9, and it has been going for almost 40 minutes on my PIII-733
>>>(over five minutes longer than it took to announce Mate in 10 via selective
>>>search).
>>
>>
>>What is exactly "Mate mode"? Is this forced mate, in other words search that
>>look for every possible mate and miss no mates? And how you indicated before to
>>your program to look for mate?
>>
>>I remember Chessmater 4000 as very good mate solver but I don't remember exactly
>>if its mate look like to be forced mate, or selective, or just both in sequence.
>>Few solutions that were presented here were quick and were found on  Chessmater
>>8000. I am almost sure that they were found by selective search.
>>
>>My program also felt big jump in branching factor between 7 and 8 move. It went
>>to 13.09 after being just before this only 6.3. It could be that later, between
>>8 and 9 move, branching became even worst. What happened later I don't know. My
>>program stop just after first solution.
>
>Chest shows the same effect (K7/600, 350 MB hash, 3.9 hours):
>
>#  4      0.10s [  5.00]        5kN [  4.96]  1.02        866-         0
>#  5      0.41s [  4.10]       20kN [  3.71]  1.15       3165-         0
>#  6      1.49s [  3.63]       75kN [  3.80]  1.44      10739-         0
>#  7      9.59s [  6.44]      609kN [  8.16]  1.61      60240-         0
>#  8    210.98s [ 22.00]    15973kN [ 26.22]  1.59    1072278-         0
>#  9  13847.86s [ 65.64]  1122335kN [ 70.26]  1.69   58979212-  50231311


Hi, Heiner!

Our branching factors are really identical. My search by brute force (my
selective found nothing) was:

Moves       Time         Branching factor

4 moves     0.054 sec
                         5
5 moves     0.27 sec
                         4
6 moves     1.98 sec
                         6.3
7 moves     6.92 sec
                         13.09
8 moves     1 min 30 sec

Mate (first mate) was found 9 moves deep after 14 min and 25 sec.

LLchess. Celeron 600Mhz. No hash.

Cheers,
Leonid.

>For depth up to 7 black finds enough checks to hinder white from mating.
>Going deeper, this does not work any more, and the search trees become large.
>Chest find 2 solutions for the mate in 9:
>
>PV: Qdxe6 Rh1+ Qxh1 Rd2+ Qcxd2 Rxh1+ Kxh1 Nf2+ Qbxf2 Qh8+ Qxh8 Qxd5+ Qaxd5 Qxd7
>Nd6+ Qxd6 Qff7#
>PV: Qxb1 Rh1+ Qbxh1 Rd2+ Qbf2 Rxf2+ Qaxf2 Qh7+ Nh6 Qxh6+ Qxh6 Qxe7 Qhxg6+ Kd8
>Nxb8+ Bd7 Rc8#
>
>Most of the time black directs what happens.  I let Chest print the solution
>tree 7 plies deep (option -lll) and got over 13,000 lines of output.  Wow!
>
>Cheers,
>Heiner
>
>>Leonid.
>>
>>>jm



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.