Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: One mate to solve for fittest programs (solution)

Author: Heiner Marxen

Date: 09:08:53 05/05/01

Go up one level in this thread


On May 04, 2001 at 17:09:35, Heiner Marxen wrote:

>On May 04, 2001 at 16:05:36, leonid wrote:
>
>>On May 04, 2001 at 15:38:23, Angrim wrote:
>>
>>>On May 04, 2001 at 10:26:02, leonid wrote:
>>>
>>>>Hello!
>>>>
>>>>Had bad lack this midnight when one call came from my work. It took me 20
>>>>minutes to do it but I went to my bed only at 2 o'clock. One good result from
>>>>broken sleep was this position that you can try to solve. It probably will
>>>>demand one sleepless night from your program, if you will insist on shortest
>>>>mate.
>>>>
>>>>[D]RnqkqnR1/qBNbNBq1/QqQqQqQ1/BrQqQrB1/3q4/8/3Q4/3K4 w- -
>>>>
>>>>Please indicate your result.
>>>>
>>>>Thanks,
>>>>Leonid.
>>>
>>>this one IS tough,
>
>Chest agrees: there is no mate in 12, as found in 2.1 hours (K7/600, 350MB).
>
>>> wonder if having endgame tables would help a lot?
>>
>>Never!
>
>:-) :-)
>
>There are 16+15 = 31 men on the board.  A mate in 12 is 23 plies deep,
>so we are left with 31-23 = 8 men at least for a terminal position in
>such a search.  You have to go some moves deeper before there is a chance
>that 6-piece tables may help.
>
>
>>>I'll leave my proof number searcher running while I'm at work, maybe it
>>>will find an answer in a few hours.. currently its obsessing over Qxd7.
>>
>>Everything depend on each program branching factor. Two professional that I
>>tried until now (they are not specialized in solving mate nut never hung on you)
>>were slow. The every next program could have very good branching factor for this
>>position. I had the chance to see very good on mine, mainly for brute  force.
>>Selective was slow. It took this in 10 min 42 sec. Celeron 600Mhz. No hash.
>>
>>Leonid.

Chest has just completed depth=13.  There are two key moves for the mate
in 13: Qexd7+ and Qcxd7+.  Here are the PVs:

Qcxd7+ Qexd7 Rxf8+ Qxf8   Qxd7+  Qcxd7  Ne6+   Q5xe6  Bxb6+  Rxb6  Qaxb6+ Qaxb6
Rxb8+  Qxb8  Nc6+   Qbxc6 Qgxf6+ Rxf6 Qxb8+  Qcc8  Qxf8+  Qde8 Qxd4+  Qd7  Bxf6#

Qexd7+ Qcxd7 Ne6+  Kxe7   Nxd4+  Rxe5   Bxf6+  Qxf6   Qxf6+  Kxf6  Qh6+   Ke7
Qxd7+  Nfxd7 Rxe8+  Kxf7  Bxd5+  Rxd5 Re7+   Kxe7  Qg7+   Ke8  Qcc8+  Qd8  Qxd8#

On K7/600 (350 MB hash) it took 11.7 hours to find both key moves, and 15.3
hours total time to also print a solution tree containing the above PVs.


>Well, the effective branching factor is quite good for Chest.
>Here is the timing for the increasing depths:
>
>       seconds
>#  1      0.00  0.87          1-         0
>#  2      0.00  1.00          1-         0
>#  3      0.00  0.95         70-         0
>#  4      0.08  1.07        465-         0
>#  5      0.37  1.27       2085-         0
>#  6      1.46  1.57       7901-         0
>#  7      6.32  2.09      34402-         0
>#  8     25.78  2.57     141569-         0
>#  9     87.41  3.28     500658-         0
># 10    478.70  3.56    2712514-       478
># 11   1570.93  4.26    8976242-   1058845
># 12   7659.24  3.73   44489747-  35741846
 # 13  41979.42  3.21  233216096- 224468195

>depth  7-> 8: 4.079
>depth  8-> 9: 3.390
>depth  9->10: 5.476
>depth 10->11: 3.281
>depth 11->12: 4.875
 depth 12->13: 5.480

>It changes a bit up and down, but stays between 3 and 5.5 so far, which
>is not bad for such a crowded board and 69 initial legal moves.
>
>But no cigar, yet.
>
>>>proven that lots of the other moves lose already though.
>>>proved that 26 of the other moves lose after 10 minutes search..
>>>
>>>Angrim
>
>Yes, for (nearly) all partial mate-in-3 Chest tried to mate the attacker
>directly in 1 move, which succeeded in 16.2% of the cases.  A more
>sophisticated heuristic might succeed in many more cases.
>
>Now, Leonid, should I go on?  Depth 13 and 14 will take around 9 hours
>and 1.5 days!  What is your shortest (selective) solution?
>
>Cheers,
>Heiner

My estimates above were even a bit low.  This mate problem was definitely
a heavy one.  But I'm sure, you will not run out of such problems, Leonid :-)

Happy weekend!
Heiner



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.