Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why Can't My Program See This?

Author: Johan Melin

Date: 06:28:46 10/24/00

Go up one level in this thread


On October 19, 2000 at 22:07:42, Michael Neish wrote:

>
>Hi,
>
>I wonder whether anyone might be able to explain why my program cannot see the
>solution to BT2630 #3 below
>
>[D]5r1k/p1q2pp1/1pb4p/n3R1NQ/7P/3B1P2/2P3P1/7K w - -
>
>The solution is 1. Re6, leading to the following tactics:
>
>1. Re6   Kg8
>2. Rxh6  gxh6
>3. Qxh6  f5
>4. Qxf8+ Kxf8
>5. Ne6+ forking King and Queen and gaining material.

Hi,

I think it is a null move problem.

Consider the line: 1. Re6 Kg8 2. Rxh6 gxh6 3. Qxh6 null

Depending on your implementation of null move, this line might use the full 8
ply. If your program does not recognize the mate-in-1 threat, it will need an
extra ply.

I would suspect that HIARCS can see the mate-in-1 threat. This helps both by
avoiding the null-move oversight, and perhaps with extensions on such threats.

/Johan Melin

>My program does not see this if it searches 8-ply deep, even though it is doing
>check extensions, and should therefore see it (I think) because there are two
>checks in the sequence.
>
>If I make it do a 9-ply search it does find the right move (but well over the
>time limit), and funnily enough, the PV that it gives (the correct one) extends
>to eleven ply (because of the two checks).
>
>So we're going from not seeing far enough at 8-ply, even though it should, to
>seeing further than you need to at 9-ply.
>
>The only strong program I have got, HIARCS 7.0, finds the move at 4-ply.  Some
>heavy-duty extensions at work here I think ...
>
>Thanks for any feedback.
>
>Mike.



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.