Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: INFINIITY question

Author: Tony Werten

Date: 07:55:57 11/29/01

Go up one level in this thread


On November 29, 2001 at 06:27:43, James Swafford wrote:

>On November 29, 2001 at 01:10:25, Tony Werten wrote:
>
>>On November 28, 2001 at 20:16:45, James Swafford wrote:
>>
>>>On November 28, 2001 at 13:41:26, Tony Werten wrote:
>>>
>>>>On November 28, 2001 at 13:05:23, Derek Mauro wrote:
>>>>
>>>>>When I call search(), I'm supposed to be using -INFINITY as alpha and INFINITY
>>>>>as beta.  Anyway, I've never used INFINITY in a program before.  How do I use
>>>>>INFINIITY?  Do I need to include a special library or just define it as the
>>>>>largest possible int value, or something else?
>>>>
>>>>Just use the value you use for checkmate (without depth correction ) That's
>>>>infinite enough.
>>>>
>>>>Tony
>>>
>>>The absolute value for INFINITY for me is a little bigger than
>>>CHECKMATE, so that the condition  if (score > alpha) is true
>>>even when the first move gets a score -CHECKMATE and alpha is
>>>-INFINITY.  Nothing gets backed up to the pv unless that condition
>>>is met.  (Of course you're in serious trouble in that case. :)
>>
>>You're already in trouble if a move scores -checkmate. If it happens at ply 0
>>than the game is finished, if it is ply 1 or more the score should be
>>-checkmate+1 (or more )
>
>That's true, assuming he will use depth correction.  (I do.)
>The point is to ensure the first move will get placed in the pv.
>If the range of valid scores is in the closed interval [-CM .. CM],
>why make assumptions when it's easier to move INFINITY outside
>that range?

True, but I got into the strange habit lately to do this anyway. The idea is
that if I think it shouldn't break anything, and I give it the chance there are
2 possibilities.

Either it doesn't break anything or I should rethink. It does seem to chance the
small bugs into big ones, wich are easier to find.

>
>A silly argument to be sure...

Yes, but then again, life tends to be rather silly. (with a few exceptions)

Tony

>
>--
>James
>
>
>
>
>
>
>>
>>Tony
>>
>>>
>>>
>>>--
>>>James



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.