Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: hash and checks

Author: Dieter Buerssner

Date: 06:50:52 01/21/01

Go up one level in this thread


On January 20, 2001 at 10:43:30, Robert Hyatt wrote:

>On January 20, 2001 at 10:10:30, Dieter Buerssner wrote:
>
>>On January 20, 2001 at 00:10:45, Robert Hyatt wrote:
>>
>>>Two things.  1.  You have to correct mate scores to be relative to the current
>>>ply when you store them.  2. You only do this if you are storing an EXACT score.

To clarify. I am speaking of adjusting mate values for exact scores,
lower bounds and upper bounds values, while storing and retrieving those values
to/from the hash table. From the above, I understand, that you think, that
adjusting upper/lower bound mate values is wrong.

>>I cannot see, why adjusting lower bound or upper bound scores is wrong.
>>If you have a lower bound score, that says, that this is a mate to 10 or better
>>from the root position, and you are 2 plies down in the search, this means,
>>that the score is a mate in 9 or better from this position. If you later
>>hit the same position 4 plies down from the root position, you know, that
>>this line yields a mate in 11 or better from the root.
>
>If the lower bound is mate in 30, and you correct this to mate in 20 because
>you are at ply=20, you just changed the bound, which is wrong.

I do not know, what you mean by "you changed the bound" here. I changed the
value, that will be stored in the hashtable. Yes, I will store the value
as mate in 20 for this position, just as if it were an exact value. When I later
retrieve it at the same ply, I will readjust it to mate in 30 from the root.
Nothing interesting here.

Interesting it gets, when you hit a transposition at a deeper ply. Say at
ply 30 (That the plys may be unreallistically high should not matter for the
argument). Say also, I have a beta of mate in 32. I will retrieve the
value mate in 20 from the hash table and adjust it to mate in 35 from the
root. Now this value is smaller than beta, and a beta cutoff will not
be allowed. This looks totally correct to me. All I know, is that from
this position, there is at least a mate in 20, and this is not enough
for a beta cutoff at this ply.

When you don't adjust the mate score for the lower bound value, you will
retrieve mate in 30, and get a wrong beta cutoff.

Regards,
Dieter



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.