Computer Chess Club Archives


Search

Terms

Messages

Subject: The Neverending Story Of Checks And QSearch

Author: Gian-Carlo Pascutto

Date: 11:20:04 10/23/00



Hi all,

I have been trying to tune my qsearch lately
and ran into the neverending story of how to
handle checks in the quiescence search.

I tried several possibilities...but none really
gave the desired results...I'd like to hear how
you handle this and why.

Also, considering that crazyhouse chess (lots of
checks, mating threats, branching factor=100+)
is important for me, how does this affect the
tradeoffs?


These are the options I considered:

A)
Fully handle checks and check escapes
-advantages: better tactical strength, better finding
 of mate combinations, more accuracy
-disadvantages: slower, less scalable search, easier
to explode

-->conclusion:
While the advantages sound nice, the disadvantages
are simply killing...my qsearch is already quite heavy,
doing more checks would probably be overkill.

B)
Partly handle checks in qsearch
-advantages: smaller, more scalable qsearch, faster
-disadvantages: possibly missed tactics

Also, what regards B...I have some implementation issues:

a) should the qsearch be allowed to start from a position
where the king is in check?

->at first I'd say no, but I found very little difference
when testing...

b) should standpat be allowed when the king is in check?

-> same as above, but I did 30 positions LESS on WAC
when enforcing this

More generally...which cases of check do I have to handle
if I do not fully do checks in the qsearch?

--
GCP



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.