Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: mate threat extension/null move

Author: Bas Hamstra

Date: 06:29:12 09/26/04

Go up one level in this thread


On September 26, 2004 at 00:13:20, Stuart Cracraft wrote:

>On September 25, 2004 at 23:59:27, Rick Bischoff wrote:
>
>>Hi,
>>
>>I only do checks if a capturing move results in a check, then I just call
>>alphabeta with a depth of zero.
>
>This is what I do -- apparently it is not very popular amongst people
>with the better programs.
>
>From what I have heard, most of them call the quiescence routine rather
>than do a back and forth between quiescence and main search when
>quiescence is in check.
>
>Perhaps they are concerned about too many extensions in the quiescence
>beyond incheck or checking extensions creeping in.
>
>To them I suggest avoiding the extra extensions by noting where the
>call to main search comes from and disable all but the extensions they
>want.
>
>Stuart

To you I suggest a comparison between your method and mine for wac141. Here is
mine, it needs 0.2 seconds (43k nodes) to find c1f4 and 0.6 seconds (150k nodes)
to see a mate in 6.

 2.   -184        50         107  g2f1 e8e2
 3    -178        60         603  g2f1 f4d5 f6g5
 3.   -178        60         695  g2f1 f4d5 f6g5
 4    -186        60        1478  g2f1 f4d5 c1g5 d5f6 g5f6
 4.   -186        60        1695  g2f1 f4d5 c1g5 d5f6 g5f6
 5    -171       121       15418  g2f1 e8e2 h1h2 g8f8 h2h3
 5.   -171       121       16257  g2f1 e8e2 h1h2 g8f8 h2h3
 6    -216       181       32290  g2f1 e8e2 b3c2 d6b4 f6e5 c7a5 c1f4 e2c2
 6    -216       231       46055  c1f4!!
 6     318       361       80343  c1f4 d6f4 h4h5 g6h5 h1h5 f4h6 h5h6 c7h2 h6h2
 6.    318       361       80738  c1f4 d6f4 h4h5 g6h5 h1h5 f4h6 h5h6 c7h2 h6h2
 7   29989       591      149497  c1f4 d6f4 h4h5 g6h5 h1h5 f4h6 h5h6 c7h2 h6h2
                                  g8f8 h2h8
 7.  29989       601      149497  c1f4 d6f4 h4h5 g6h5 h1h5 f4h6 h5h6 c7h2 h6h2
                                  g8f8 h2h8

The trick is to generare check-giving moves in the first plies of quisce, so
nullmove will not overlook simple mate in one's. This is not a hard position,
it's an easy one. *Unless* you make your search blind for mates in the last 3
plies, by doing a capture-only qsearch combined with nullmove.

Bas.



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.