Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Buggy check extensions, back to square one =(.

Author: Robert Hyatt

Date: 20:33:35 03/20/03

Go up one level in this thread


On March 20, 2003 at 17:10:50, Uri Blass wrote:

>On March 20, 2003 at 14:49:41, Robert Hyatt wrote:
>
>>On March 20, 2003 at 12:17:23, Albert Bertilsson wrote:
>>
>>>Hi everybody!
>>>
>>>For those of you with a good memory you probably saw my posts about WAC, and
>>>that I implemented check extensions and got a really nice improvement.
>>>
>>>After playing some 150 games with about 10 lockups I also know that my
>>>extensions aren't bug free =(.
>>>
>>>In my stupidity I thought that all that needs to be done to get extensions work
>>>is to add this:
>>>
>>>First in Alpha-Beta:
>>>if (CurrentPly >= MAXPLY) return Eval();
>>
>>That might not be good enough if you have arrays dimensioned for xxx[MAXPLY].
>>
>>As when you get to MAXPLY you are already one beyond the array bounds and that
>>can wreck things badly...
>
>The question is what is MAXPLY
>
>If you choose MAXPLY=100 and have only one ply extension for checks then I do
>not see a practical case when you get close to the limit.
>
>My limit is 100 and inspite of the fact that I may do even more than one ply
>extension for checks I do not know about a position when I get close to that
>limit.

I hit 64 al the time.  So it's doable...


>
>The problem was a practical problem and the question is if
>increasing maxply solve it.
>
>I may consider limiting the ply based on the iteration(it is an idea that I did
>not try) but I see no reason to limit the depth by a small number like 30 or 40
>plies.
>
>Uri



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.