Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Using too-shallow mate scores from the hash table

Author: Robert Hyatt

Date: 17:40:01 07/06/98

Go up one level in this thread


On July 06, 1998 at 14:11:03, David Eppstein wrote:

>On July 05, 1998 at 14:14:36, Ernst A. Heinz wrote:
>>>No, not stupid at all -- just obvious and straightforward. I have used it
>>>with good success in "DarkThought" for a long time. :-)
>
>Straightforward, I agree.  Obvious?  It took me a year to notice, maybe I am
>slow, but how many other similar little tricks does everyone have to rediscover
>for himself (or worse not rediscover)?
>
>I think we really need a good text that will cover all these details.
>Bob was talking about doing something like that a while ago; what happened?
>
>On July 06, 1998 at 13:04:17, Robert Hyatt wrote:
>>In Cray Blitz (and in crafty, unless I have broken it) I store mates with a
>>"draft" of infinity, because a mate is independent of any draft, and is >*always* correct.
>
>You might find a mate in fewer moves if you searched longer; there's no
>game-theoretic value to doing so (assuming your time management code isn't
>stupid) but there is a certain aesthetic value.
>
>As I said, I looked at crafty and didn't see it, which was what led me to think
>someone other than myself might not already know about it.  StoreBest() and its
>relatives certainly detect whether a score is mate, but only use that to modify
>word1; the draft lives in word2 and doesn't seem to be conditional on whether
>the score is mate. LookUp also has some mate-score-conditional code, but only
>after it's discarded too-shallow positions.  So if it isn't in these two places,
>where are you hiding this piece of code?


Good eyes... where it lives is in the version of StoreBest() and
StoreRefutation() that uses multiple probes into a single hash table, rather
than the current "Belle" implementation.  Not sure why it was removed, probably
at the request of someone that insisted on finding shorter mates.  Once a mate
score hits the hash table in CB or the other Store() code, that's it...




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.