Author: Vincent Diepeveen
Date: 19:04:51 09/29/00
Go up one level in this thread
On September 29, 2000 at 20:25:32, Bruce Moreland wrote: >On September 29, 2000 at 18:59:43, martin fierz wrote: > >>hi, >> >>i was wondering about the concept of singular extensions, which got my attention >>because the deep blue team used them and attributed some of their playing >>strength to these extensions. questions: is anybody of you using singular >>extensions in your programs? and if yes: how much does it help? and does anybody >>know a webpage with some pseudocode for singular extensions on? >> >>best regards >> martin > >I use singular extension in my program. In some cases I do a reduced depth >search before searching the first move. I don't always search all the moves, >and sometimes I abort after only searching a few. So there is some finesse >going on, it's not like you can just do it brute force without any thinking. > >If you do it brute force without any thinking, you'll probably kill yourself. > >If you constrain them, you can double effective tactical speed while still using >null move R=2. The cost is a ply of search depth. The rating increase, when >comparing versions in head to head blitz play, seems to be almost zero. These >results are a little confusing to me, and it's not clear that overall this >extension is a win or a loss. > >The version of my program that finished 2nd in Paderborn 1999 used this >extension. Honestly I don't remember if my 1998 Paris version did, but I don't >think it did. your 1997 paris version started regurarly 14 ply searches when playing me. that was at a 767Mhz alpha. 1999 version at quad xeon didn't even get near that, of course evaluation different, hardware different, but most likely also search different? note that shredder in wcc99 also searched way deeper as it did at over twice the same hardware in wmccc2000 >bruce
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.