Author: Stuart Cracraft
Date: 08:41:15 09/05/04
Go up one level in this thread
On September 04, 2004 at 06:56:31, Uri Blass wrote:
>On September 03, 2004 at 19:58:55, Ross Boyd wrote:
>
>>On September 03, 2004 at 10:25:27, Stuart Cracraft wrote:
>>
>>>On September 03, 2004 at 06:41:16, Ross Boyd wrote:
>>>
>>>>On September 03, 2004 at 00:43:27, Stuart Cracraft wrote:
>>>>TRACE solves this (now) in under a second on a PIII-450. Previously, it took
>>>>several minutes...
>>>>
>>>>I can tell you that I debugged my THREAT extension code and found Crafty's exact
>>>>method was not conducive to good results (for TRACE).
>>>>I extend when nullmove_score==-MATE+ply+1
>>>
>>>Please include the call to your search for the null move? I want to
>>>see your window.
>>
>>No problems....
>>
>>null_depth = remaining_depth - R; // R=2 for me, always
>>
>>if( null_depth > ONE_PLY)
>> nullmove_score = -search(-beta, -beta+1, null_depth - ONE_PLY);
>>else
>> nullmove_score = -quiesce(-beta,-beta+1);
>>
>>unmakenullmove();
>>
>>if (nullmove_score >= beta) {
>> return nullmove_score; // Fail soft
>>}
>>
>>if (nullmove_score == -CHECKMATE+ply+1) {
>> hashflags |= THREAT_FLAG; // Tag it as a position with mate threats
>>}
>>
>>// and then this code, (more or less)
>>if (hashflags & THREAT_FLAG)
>> remaining_depth += EXTEND_THREAT;
>>
>>
>>Actually, since changing my search to count nullmoves as a ply increment, it
>>works better with -CHECKMATE+ply+2....
>>
>>HTH,
>>
>>Ross
>
>How much is EXTEND_THREAT for you?
>
>At the moment I do not extend mate threats.
>
>I checked that my tree can explode if I extend one ply for mate threat because I
>also extend in most cases one ply or slightly more than it for positions when
>the king is in check(slightly more than it may happen often but not always in
>case of a single reply) and if a reply to check is a mate threat the program can
>crash.
>
>I can limit it to less than it but I am still afraid from explosion for the same
>reason because single reply to check can be extended by almost 2 plies in some
>cases that hopefully do not happen often.
>
>Uri
Just 1.
On my null move, I use -beta,-beta+1 for my window and do not increment
ply when calling search:
Mate threat then is
if value_returned_from_null_move equals -MATE+ply+2
I hope this is right -- I have tried settings of -MATE+ply+0 through
-MATE+ply+2 and none of them help my test suite results, although when
I print the board if the above is true, it's clear it is finding
mate threat positions, so I don't know why mate threat is not at least
more helpful.
My test suite result is approximately the same whether or not I use
mate threat.
Note: all of this is with the new program that returns value for everything
and never beta. The suite stayed at the current 241/300 on WAC.
Somehow I think mate threat is not helping me.
Stuart
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.