Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: node explosion

Author: Jan K.

Date: 15:21:00 08/22/05

Go up one level in this thread


It's imho a bit hard to give any intelligent advice here without more data or
seeing the source ;) But if you want some general hints..... 1) such things
simply happen when changing pv and they're worse if your move ordering is
bad-ffh of 72% seems to be bad, so if you worry about move ordering, implement
hash tables first, _then_ do some tests and you will see if it really doesn't
work. I don't think that you can get good move ordering without hash. Btw if you
only consider pv move from previous ply and you search it first at the root,
which moves you search next? Random order or such a sort you speak of can fail
badly here(especially in positions with quiet moves and that is the start
position), this can very easily cause such node explosion. You don't say
anything about ordering of quiet moves, so I can guess nothing. 2)do usual
actions to see what happened-for example check for extension explosions, qsearch
explosions, check the order of moves at the root and how many nodes each move
consumed,check eval for asymetry that could introduce search instability and so
on...



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.