Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To Bob, probably Crafty has an interesting problem

Author: Joachim Heuser

Date: 02:11:51 06/10/01

Go up one level in this thread


On June 09, 2001 at 17:23:56, Robert Hyatt wrote:

>On June 09, 2001 at 14:38:04, Joachim Heuser wrote:
>
>>On June 09, 2001 at 11:32:09, Robert Hyatt wrote:
>>
>>>On June 09, 2001 at 04:37:32, Joachim Heuser wrote:
>>>
>>>>On June 08, 2001 at 22:11:56, Jim Monaghan wrote:
>>>>
>>
>>>>I am quite sure that nearly every position from KBBKN is won for the two
>>>>bishops.
>>>
>>>Sorry, but you are quite wrong.  Go to my ftp site, cd to the TB/tbs
>>>directory, and download kbbkn*tbs and take a look.  More draws than
>>>wins.
>>>
>>
>>What i meant was: from nearly every position, where the two bishops are on
>>different colours and the knight cannot capture one of those bishops in the next
>>move, the two bishops can force mate or winning the knight, though this may take
>>more than 50 moves.
>>I set up some rather ugly positions with the two bishops not working together
>>and it took ~40-50 moves (according to cb-tablebases) to capture the knight.
>>
>>In the book german book from Kishon: "Schachcomputer" (1993), the author claims
>>that the machine "Alice" analysed this endgame to be always won (he doesn't
>>state when this analysis took place). There was a position, too, which should
>>take longest to capture the knight (66 moves):
>>K7/8/7B/8/8/5k2/6n1/7B w - - 0 1
>>wKa8,Bh1,h6/bKf3,Ng2
>>
>>I wasn't able to create a drawn position which matches the pattern mentioned
>>above.
>
>
>I haven't tried this, but the simple way to estimate how many wins vs draws
>there are in KBB vs KN is a Monte Carlo approach.  Generate quite a few random
>positions with bishops on opposite colors, and probe to see if it is won or
>drawn.  A few thousand should give an accurate percentage of what percentage
>is drawn vs won...

Hmm, well, i'll go for it in my next vacation ;-)

I looked at the kbbkn.tbs file - i cannot explain the contradiction between the
rate draws/wins and my statement above. With approx. 60% draws due to "wrong"
bishops it's still only a 6:5 win/draw-rate.

I tried to reach a drawn position from a won position by choosing bad moves for
the bishops. I came down to approx. 75 moves mating distance. In this region i
looked for white moves only to draw - every move lost or traded one bishop.

I'm at a loss.
Isn't there a program by Nalimov which creates statistics about each
tablebase-file?

Joachim




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.