Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fail Soft question

Author: Chan Rasjid

Date: 11:31:29 02/18/06

Go up one level in this thread


On February 18, 2006 at 07:52:10, David B Weller wrote:

>Hi Vincent,
>
>Thanks.
>
>Yes. That is what I am doing. Will try and remove those.
>
>Let me make sure I understand perfectly:
>
>If a hash entry is LOWER or UPPER bound AND exceeds current search bounds, just
>return bounds. Do NOT return hashed value.
>
>Yes?
>
>-David

Hello David,

I am surprised you understand Vincent. I have a hard time as he gives only
subtle hints.

I have done TT and fail-soft many times over and still have no inkling if there
should be things to take special care of.

If I am right, fail-soft is always better. When a node FL, best may be < alpha
and we hash best and not the bound alpha. When it is FH, best is likely > beta
and we hash best and not beta.The lower/upper in TT will always be more useful
when the bounds are as "tight" as we get; ie smallest upper or greatest lower.

Following a strict coding usually prevents mistakes. eg.

v = hashProbe(&bound);//etc....

switch(bound){
case lower:
 if (v >= beta)
 return v;//FH
etc...
break;
case upper:
 if (v <= alpha)
 return v;//FL- search dont get better
etc...
break;
}


Best regards
Rasjid



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.