Author: Dann Corbit
Date: 19:43:10 01/27/04
Go up one level in this thread
On January 27, 2004 at 22:36:45, Russell Reagan wrote: >On January 27, 2004 at 21:20:00, Dann Corbit wrote: > >>It's not the same thing. The aspiration search is in a loop. This only checks >>the root at two plies less than the main search. > >But the code you posted still has search instability, right? i.e. what happens >if the second IID search (-32500, alpha+1) fails high? Wouldn't it be better to >search with (-32500, 32500)? > > >>The entire purpose is to find a pretty good choice for the pv node if we don't >>have a guess yet. >> >>In other words: >>No data in the hash table? Then compute a shallower answer. > >Okay, so basically what you're doing is saying that if the re-search fails high, >you don't care, because whatever the IID pv move (while possibly different from >the move resulting from a (-32500, 32500) re-search), it is probably better than >whatever the random first move would have been. Is that about right? If it fails high, chances are that our choice of the pv node is a good one. If it fails low, then we better look again. We already know that our best guess stinks. If we don't improve it, what will we end up searching? IOW, if your 'best node' guess fails low -- it's bad. And if your 'best node' guess fails high -- it's good. We don't care what the real value is because we are going to search at greater depth anyway. We just want an idea of where to look first.
This page took 0.01 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.