Razoring
Joe Stella
want to ask "But What Have You Done For Me Lately?" :-)
I have been experimenting with the "razoring" idea:
if (depth == 1) /* one ply before the horizon */
{
score = quies_search();
if (score > beta) return score;
}
I am not sure if this makes my program better or worse.
The search is "faster" in the sense that it reaches deeper
depths more quickly, but then it seems to need the extra
depth just to find the same combinations. In the end,
I threw razoring out.
Has anyone else been experimenting with this? What were
the results?
Joe S.
Larry Craighead
writes:
How about doing this:
if (depth == 1) {
score = evaluate();
if (score > beta) return (quiesce());
}
Of course you can use the maximum positional value to cut down on the
number of evaluations.
> Joe S.
>
>
Matt Craighead
Joe Stella
plug...@ix.netcom.com (Larry Craighead) writes:
>How about doing this:
> if (depth == 1) {
> score = evaluate();
> if (score > beta) return (quiesce());
> }
Unless I am missing something, this doesn't make sense. If the evaluation
of a position is > beta, this could be because a queen is hanging and could
be captured on the very next move. Returning the quiesce() score would just
be reducing the search depth by one ply with no justification.
I wonder if the (relative) lack of answers to my original post means that
that no one is using razoring?
Joe S.
Larry Craighead
writes:
>In article <42588v$6...@ixnews2.ix.netcom.com>
>plug...@ix.netcom.com (Larry Craighead) writes:
>
>>How about doing this:
>
>> if (depth == 1) {
>> score = evaluate();
>> if (score > beta) return (quiesce());
>> }
>
>Unless I am missing something, this doesn't make sense. If the
evaluation
>of a position is > beta, this could be because a queen is hanging and
could
>be captured on the very next move. Returning the quiesce() score
would just
>be reducing the search depth by one ply with no justification.
Yes, I realized this soon after posting. You also need to verify that
the score from quiesce() is > beta. Of course, the idea is to cut back
on quiesce() calls if the current score is bad.
>I wonder if the (relative) lack of answers to my original post means
that
>that no one is using razoring?
Probably.
> Joe S.
>
Matt Craighead