Computer Chess Club Archives




Subject: Re: If you want solve one mate...

Author: Heiner Marxen

Date: 15:31:55 01/21/01

Go up one level in this thread

On January 21, 2001 at 16:08:25, Uri Blass wrote:

>On January 21, 2001 at 15:22:30, Heiner Marxen wrote:
>>On January 21, 2001 at 14:13:30, leonid wrote:
>>>On January 21, 2001 at 13:44:00, Heiner Marxen wrote:
>>>>On January 21, 2001 at 13:16:22, leonid wrote:
>>>>>On January 21, 2001 at 12:43:34, Heiner Marxen wrote:
>>>>>>On January 21, 2001 at 11:56:09, leonid wrote:
>>>>>>>This mate position is very handy when you want to check the limits in your
>>>>>>>program. It is amusing also. Number of legal moves for both sides is slightly
>>>>>>>over one hundred.
>>>>>>> knq3q1/rq1q1qbQ/Qq2Q3/1Qn1Q2B/Q1qN1Q1r/4K1B1/1q4QR/1N1q3Q white to go.
>>>>>>A hint to you, Leonid:  if you like to see a diagram for the notation above:
>>>>>>(1) Stick a [D] directly before the FEN string,
>>>>>>(2) append (blank seperated) colour to move, i.e. b or w
>>>>>>(3) append (blank seperated) castling rights: - for nothing, KQkq or part
>>>>>>    of it for white/black king side/queen side
>>>>>>(4) append (blank seperated) e.p. info: - for "no e.p. possible"
>>>>>>You can look up the details in the FEN standard, part of the PGN standard.
>>>>>>Look at the line below, which I typed in, and see the diagram, provided not
>>>>>>by me, but by the CCC web server, which recognizes the [D] and FEN after it,
>>>>>>and inserts the graphics.  You can do that also.
>>>>>Thanks! Will try to use it the next time.
>>>>>>[D]knq3q1/rq1q1qbQ/Qq2Q3/1Qn1Q2B/Q1qN1Q1r/4K1B1/1q4QR/1N1q3Q w - -
>>>>>>Chest says "no mate in 9" (216 seconds, K7/600 335MB hash).
>>>>>>I will ask for more depth...
>>>>>Are there no mistake? You have really 335Mb hash?
>>>>Yes!  :-)
>>>>I bought 512 MB.  Was a bit expensive, but my monitor costs even more.
>>>>And on work I have seen what 512MB can do to big Chest jobs, so I really
>>>>wanted to have so much memory.  I like it :-)
>>>And what the monitor you have? 21 inch?
>>>The last time, around 3 months ago, I bought my last 17 inch. Very cheap and
>>>very good. My next dream is new computer. I need it for my programming.
>>It is a 15.1 inch TFT screen.  It had to fit into this tiny room, where
>>my computer is placed.  I am very satisfied with it, but it is not cheap.
>>>>>216 sec for 9 moves like very good. I went, by brute force, only up to 7 moves.
>>>>>It was already 2 min. 19 sec. But for this position, which look very heavy,
>>>>>branching factor was very normal. Like what you can expect from usual forced
>>>>>mate position.
>>>>>4 moves - 0.77 sec
>>>>>               branching factor 5.8
>>>>>5 moves - 4.39 sec
>>>>>                                5.3
>>>>>6 moves - 23.02 sec
>>>>>                                6.1
>>>>>7 moves - 2 min 19 sec
>>>>Yes, black has also 9 queens, and whites king is exposed.  If white does
>>>>anything not really forcing, black will start to force, or even mate.
>>>>Therefore the search trees can be small in many branches.  In 17% of all
>>>>considered cases, black mates white!  (yes, Chest tries that sometimes)
>>>>Here are my numbers:
>>>>#  1      0.00  0.87          1-         0
>>>>#  2      0.01  1.00          1-         0
>>>>#  3      0.03  0.96        107-         0
>>>>#  4      0.19  1.04        819-         0
>>>>#  5      0.87  1.23       4167-         0
>>>>#  6      3.73  1.48      17787-         0
>>>>#  7     14.90  1.79      72503-         0
>>>>#  8     56.43  2.17     290160-         0
>>>>#  9    213.00  2.57    1142162-         0
>>>># 10    831.85  2.90    4544295-     16385
>>>>(depth  seconds speed   node in   node out
>>>Pretty good branching factor!
>>Yes, and here is the next line:
>># 11   3686.85  2.90   20110217-  11363105
>>Chest does not find any mate in 11.
>>>>>>>If you will solve, it will be nice if you will indicate your result.
>>>>Is there any forced solution, at all?  Depth=11 will take around an hour,
>>>>and I will not go much further, if there is no real chance.
>>>For 11 moves I don't know. My solved it by selective only at 12. Actually, this
>>>position is more that strange. Solution was found by the most easy level in
>>>selective search. Because of very good for me branching factor, I expected it
>>>will be instant. It was not so. It took 2 seconds.
>>Fine, so I stop this, now.  Looks like there is no shorter mate,
>>which is not surprising for a position like this one.
>>I just asked crafty 17.7 for a depth=1 analysis, and even after 3 minutes
>>I still see no output ;-)  Obviously the quiescence search is exploding.
>>Can you give some PV for the forced #12 line?  Just curious.
>1.Qxa7+ Qxa7 2.Qxa7+ Kxa7 3.Qgxb7+ Qdxb7(to prevent the threat Qxb8+)
>4.Qexc5+ Q4xc5(Ka8 Qxb8+ and Qb7 is pinned) 5.Qxb8+ Qcxb8(Qbxb8 Bxb8+ mate in 2)
>6.Qxc5+ Q7b6(Q2b6 let Ra2+) and I guess that now chest has no problem to see
>that 7.Bxb8+ is a mate in 6.

I let Chest analyse after 1.Qxa7+ Qxa7 2.Qxa7+ Kxa7, and can confirm your line
upto but excluding 6.Qxc5+, where Chest choose to do 6.Bxb8+.  I suspect
both moves are ok.
BTW, after the first two moves besides 3.Qgxb7+ also 3.Qexc5+ and 3.Qeb6+
solve this.  I like the last one (it does not even capture):

3.Qeb6+ Ka8  4.Qgxb7+ Qdxb7  5.Qxb8+ Qxb8  6.Qxb8+ Qxb8  7.Q6a5+ Na6
8.Qaxa6+ Qa7  9.Qg8+ Bf8  10.Qxf8+ Qe8+  11.Qbxe8+ Qc8  12.Qxc8#

Crafty 17.7 after over 2 hours still has not completed depth=1  ;-)
I think Bob will not care much, since this position is not at all for real.


This page took 0.01 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.