Author: Robert Hyatt
Date: 21:16:09 08/30/01
Go up one level in this thread
On August 30, 2001 at 21:58:04, Pham Minh Tri wrote:
>Hi Bob,
>
>I am studying how Crafty closes the search window in the root search (file
>searchr.c). I have learned much from your code. Thanks.
>
>There is a point I am unclear: updating beta. I am afraid that the search of
>next moves may be faulty after updating it.
>
>Suppose a situation: original value of beta = 1000, a good move brings a value
>of 200, so Crafty changes beta to value+40 = 240. The next move gives a better
>value of 300. Crafty says the search is good and need not re-search with wider
>window and then stop searching because of value>beta. It turns out there is no
>bestmove. However, if program does not update beta (beta==1000), the search has
>to redo and may find a bestmove (say, a move gives exactly 400).
>
>What do I miss here?
>Many thanks,
>Pham
>
>PS: Here is pseudo code after searching with window [-alpha-1, -alpha]. I marked
>the unclear points by arrows.
>
> Search with window [-alpha-1, -alpha]
>
>==> if ((value > alpha) && (value < beta)) {
> Re-search with window [-beta, -alpha]
> }
> if (value > alpha) {
> root_value=alpha;
> if(value >= beta) {
> Update some thing
> return value;
> }
> alpha=value;
>==> beta=value+40;
> root_beta=beta;
> }
It is too late to look at that tonight with any reasonable level of
concentration. It looks wrong at first, and might well be something left
lying around from testing something else somewhere in the past. I see no
reason to tickle around with beta there. I will try to study it tomorrow
to see if I can remember what was going on here...
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.