Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Unusual Mate in 3

Author: Thomas Mayer

Date: 12:27:17 03/02/06

Go up one level in this thread


Hi Bob,

On March 02, 2006 at 11:33:36, Robert Hyatt wrote:

>On March 02, 2006 at 07:29:46, Thomas Mayer wrote:
>
>>Hi Charles,
>>
>>>  Came across this position by accident browsing chess club web sites.
>>>  It comes from the Duke University Chess Club. The position is
>>>  1r5k/4NP1b/7K/8/6R1/8/8/8 w - - 0 1
>>>
>>>  Telepath found the mate in 4 but could not find the mate in 3.
>>>  After some debugging I found out why!!!
>>>
>>>  Just before the final ply in the mate in 3, Telepath announced insufficent
>>>  material which is true. But in this position you can mate with
>>>  insufficient material.
>>>  It put such a smile on my face, I had to share it.
>>>  Maybe some of you have the same logic flaw in your code.

>> well, somehow. With TBs Quark finds the mate, of course, but when I play to
>> the position it says draw for Bxg8...  When I play instead of just
>> analyzing it even ends the game in WinBoard with claiming draw. I wonder
>> how many more engines have this problem as well... :) There are of course
>> several other scenario thinkable where this problem might arise,
>> interesting and thanks for pointing that out, even when it is not really
>> strength related. Also it would be interesting to see if some engines / GUI
>> act here different when TB's are enabled or disabled.

> This is a danger of "interior node recognizers" if you use those.  It is
> safe to evaluate such positions as draw in your eval, because then your
> search will find the mate.  But if you try to recognize such draws at
> interior nodes, you potentially have this problem...

well, in fact it is not my "interior node recognizer". It is my draw function.
Whereas the INR-function is aware of that possibility, my draw function has no
idea - and it a) produces also cutoffs in the tree and b) is responsible for the
draw claim... funny... I did take so much care about my internal node
recognizers not to give back wrong results (even checking them with TBs) and
then the draw function is the problem... :)

Greets, Thomas



This page took 0.03 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.