Author: Stuart Cracraft
Date: 09:36:13 07/28/04
Go up one level in this thread
On July 28, 2004 at 12:25:32, Tord Romstad wrote: >On July 28, 2004 at 12:04:41, Stuart Cracraft wrote: > >>So in looking at ETC and implementing it (without much benefit >>I might add, in a PVS/NEGASCOUT search), I thought a little more >>and perhaps have a complete misunderstanding of it, hence the >>implementation. Am reconsidering it for a recent MTD(f) implementation >>that is at parity with the PVS/NEGASCOUT with very little work. >> >>From what I have read, ETC is a check prior to search() in the >>main search of all child positions hashkeys after each child >>makemove(). If any exist in the hash table and have length >= >>current depth and provide a cutoff, then skip searching all >>the children. >> >>Is that about it? > >Yes. Two additional points worth mentioning: > >If your makemove() is expensive, you might want to write a special version of >makemove() which updates the hash key and nothing else. > >It is probably not a good idea to use ETC close to the leaves. I use it >only when the remaining depth is at least 4 plies. > >Tord Thanks -- I have heard 2 instead of 4 but will experiment. My makemove is cheap so no problem there. (I wish it were expensive since then my program would be smarter.) Do you use MTD(f) with ETC? If so, what is your observation? Stuart
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.