Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: One mate to solve.

Author: Heiner Marxen

Date: 10:09:00 11/03/01

Go up one level in this thread


On November 03, 2001 at 08:21:01, leonid wrote:

>On November 02, 2001 at 22:02:32, Heiner Marxen wrote:
>
>>On November 02, 2001 at 18:29:21, leonid wrote:
>>
>>>On November 02, 2001 at 17:20:27, Paul wrote:
>>>
>>>>On November 02, 2001 at 17:05:48, Angrim wrote:
>>>>
>>>>>On November 02, 2001 at 10:12:03, leonid wrote:
>>>>>
>>>>>>Hi
>>>>>>
>>>>>>This one look like reflection of somebody free fantasy. In reality, it came from
>>>>>>absence of it. Was written after furtive glance on one Chinese character.
>>>>>>
>>>>>>[D]8/qkqqqqQ1/q2Q4/nPBNB3/q3B3/bNRQR3/q2P4/qrrbnQK1 w - -
>>>>>>
>>>>>>Please indicate your result.
>>>>>>
>>>>>>Thanks,
>>>>>>Leonid.
>>>>>
>>>>>Same system as usual:
>>>>>proved that move d6xc7 wins, 15 turns(roughly 30 ply)
>>>>>PN2:11172611 evals, 267628 expands, 42.24 seconds
>>>>>
>>>>>After manually making the move bxa6 instead, it finds that
>>>>>black has lost in 3 seconds.
>>>>>
>>>>>Angrim
>>>>
>>>>Hi guys,
>>>
>>>Hello to Everybody!
>>
>>Hello!
>
>Hi, Heiner!
>
>
>>>>Indeed, mine starts with bxa6+ (p3/933) ... but I know PNx search doesn't care
>>>>about the shortest line:
>>>>
>>>>00:09 WM9 07 bxa6+ Kb8 Qh8+ Qde8 Qxc7+ Qexc7 Qb5+ Qxb5 Qxb5+ Qb6 Qxb6+ Kc8 Qxc7+
>>>>Qxc7 Qxe8+ Qd8 Ne7#
>>>
>>>If I read correctly your response, it is mate in 9. It is really shortest mate
>>>and your program do very well.
>>>
>>>Mine was not very efficent in selective and find mate only in 10 moves. Five
>>>seconds. Brute force found mate in 9, after 52 min of search.
>>
>>Fine!  Chest confirms all this (2.5 hours K7/600 with 350 MB hash):
>>
>>PV: bxa6+ Kb8 Qh8+ Qc8 Qxc8+ Kxc8 Nxe7+ Qfxe7 Bf5+ Qad7 Bxd7+ Qcxd7 Bxa7+ Qc7
>>Qff5+ Qd7 Qfxd7#
>>
>>This is the only key move.
>>Leonid's time appears to be similar to that of Chest, this time:
>>
>>#  1      0.00s                 0kN           0.87          1-         0
>>#  2      0.01s                 0kN           1.00          1-         0
>>#  3      0.04s [  4.00]        2kN [ 17.74]  0.97         76-         0
>>#  4      0.28s [  7.00]       15kN [  9.66]  1.04       1353-         0
>>#  5      2.62s [  9.36]      117kN [  7.73]  1.16      13556-         0
>>#  6     18.96s [  7.24]      859kN [  7.37]  1.30     106311-         0
>>#  7    141.96s [  7.49]     6480kN [  7.54]  1.44     843816-         0
>>#  8   1099.97s [  7.75]    50015kN [  7.72]  1.64    6709202-    207220
>>#  9   8881.32s [  8.07]   394504kN [  7.89]  1.76   54835847-  46087946
>
>
>Since you indicated your complet time and mine goes very close to your, will
>write my time for brute force.
>
>Depth                Time            Branching factor        NPS
>
>4 moves              1.09 sec                                64k
>                                     8
>5 moves              8.79 sec                                60k
>                                     7.33
>6 moves              1 min 4 sec                             58k
>                                     7.42
>7 moves              7 min 55 sec                            53k
>                                     6.45
>8 moves              51 min 7 sec                            52k
>
>9 moves              51 min 21 sec for first solution.
>
>What I see the most remarkable in your search is that your specialized plys do
>much better that mine. It is visible when I look your and my time for 4 moves.

Yes.  It definitely looks like that.

>Heiner, can you remind (I am too lazy to read needed rescription) of how your
>kind of data can be seen. What must be said to your program. For now I can see
>total time and main winning line. Sometime it will be useful for me to see your
>time, mainly, for 4 and 5 moves. It is the most revealing of how much mine still
>miss something at low level.

You can give it option "-s" at the command line, i.e. instead of
    chest319 -b inputfile
you say
    chest319 -b -s inputfile

As a consequence there will appear statistics information (s = statistics)
in the output, before and after the result EPD line.
Just ignore the part you do not understand.
Some weeks ago I explained most parts of it to Paul (IIRC).  You may find
that posting in the archives (sorry, I did not save it).

Unfortunately, version 3.19 statistics do not include the node counts you
see me post here from the current development version.

Fortunately, there will be a version 4.1 to come, which will also include
Nalimov EGTBs and will work as a winboard engine.  I hope to be ready
around Christmas.

>>EBF was remarkably stable, just a bit below 8.

Your EBF was very similar, in fact.


>>>And how you people is impressed with the last Microsoft decision? I heard it
>>>today with big interest. Do you think that inner Windows structures and data
>>>will be published, and it will help to all of us in writing our new code?
>>
>>What decision?  Do you have some URL to read?
>
>I just listened to radio and for sure can go and see what is written on many
>pages that can be found in the Net. Actually, I will do this for sure but some
>opinion from those that I in writing is all the time good to hear.
>
>For me the most important is to know how much inner cooking of Windows system is
>now available to the public. Publication about Assembler writing for Windows was
>hammered by absence of direct and complet description of all Windows services
>and their constant change. Also I am eager to know if "Software Development Kit"
>will become free part of every Assembler, or even some Web accessible
>downloadable file. For now this package (that theoretically must be upgraded and
>bought after each version of Windows) is serious financial obstacle for many
>free lance writers like me. When I bought its version for Windows 3.1, its price
>was 500$ in Canadian money.

Aha, I see.
But I have not yet heard anything about it, except for your comments.
Thanks for explaining.

>I hope that now Windows will become something of half way between previous
>Windows and actual Linux. Would like to see this. I am very fond of Microsoft
>Assembler.

Hmmm, somehow I doubt it.  But those doubts may well be wrong.
Let's hope and see...

Cheers,
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.