Author: Robert Hyatt
Date: 19:27:05 10/18/00
Go up one level in this thread
On October 18, 2000 at 18:16:07, Brian Richardson wrote: >Most better programs seem to use SEE. I have not implemented this yet in >Tinker, but have some concerns. I realize the point of SEE is "static" (and >meant to be much faster than recursive make/unmake move searching), but is it >not a problem that the SEE is only concerned with attacks to one particuar >square at a time? I would think that a series of exchanges would open up other >significant move options that the SEE does not see (pardon the pun). I don't >recall reading anything about SEE in ICCA or elsewhere and was wondering what >others thought about this. Perhaps the move ordering (and pruning) benefits >simply outweigh the incomplete analysis, sort of like the crude history score >technique. See is definitely inaccurate... ie it doesn't usually understand things like overloaded pieces, pinned pieces, etc. But most of the time, things aren't overloaded or pinned, so it works perfectly well. And since it is only used for ordering captures in the basic search, it won't eliminate anything, just make the tree slightly bigger than optimal if it screws up. > >Actually, I was also wondering about a history question--what do people think is >better: One history table or history tables by side? > I do two, one for white, one for black. >Thanks >Brian
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.