Hash table question

[John Stanback]

In Zarkov I have found that I get slightly better results when
not storing nodes in which score < alpha (fail low nodes).
Perhaps this is because I have a small hash table (8K - 64K positions)
and fail low nodes are not that useful since the score probably won't
give any cutoffs. I figure it's better to use the available table space
for more "useful" nodes. Have you tried this on CB Bob?

John Stanback

[Robert Hyatt]

I haven't tried "not storing" them, but we do have lots of rules about
what overwrites what. Obviously good scores overwrite such junk windows
whenever they occur, and junk windown don't overwrite good scores unless
several strict criteria are met.

Unfortunately you are asking about the most "superstitious" part of
Cray Blitz. The golden rule, "if it ain't broke, don't fix it" has
been followed to the letter on the hashing code. It is probably worth
going over while I'm working on things here.

--
!Robert Hyatt Computer and Information Sciences !
!hy...@cis.uab.edu University of Alabama at Birmingham !

[Robert Hyatt]

In article <Cn4K9...@fc.hp.com> j...@fc.hp.com (John Stanback) writes:

Also, don't forget that with the zero-width window, *everything* will cut
off in the tree, either high or low. I can try a couple of normal searches
and "filter" these positions out to see what it does...