Author: Ed Schröder
Date: 14:04:34 09/08/04
Go up one level in this thread
On September 08, 2004 at 16:49:33, zaid magi wrote: >taken from: http://members.home.nl/matador/chess840.htm#EXTENSIONS >[quote] >The check extension in REBEL is rewarded in MakeMove(); when it is recognized >that the side to move moves out of check, the depth is extended with 1 ply, the >common procedure in most chess programs. > >An exception is made for the first check, it is not extended, also the fourth >check is extended with 2 plies being in sync again with common procedure to >extend every out_of_check situation with one ply. The idea behind: when there is >only one check in the tree, it is probably not so important, thus skip it. >However when you have 4 checks in the Search, the chance is big that checks play >an important role, better get in sync, thus extend 2 plies. > >Skipping the "first check" is very time sensitive, it speed-up REBEL with 25%, >however its elo gain is very small in comparison with the common procedure to >extend every out_of_check situation, for REBEL the gain is about +5 elo, you >must find out yourself if the idea works in your own engine. >[/quote] > >what does "first check" actually mean ?? >does he mean the first check that appear in that chess tree ?! >or the first check in the plies previous to the position under analysis ?!?! > >thanks for any explanation :D [d]k7/8/8/8/8/8/8/K6R w - - 1.Rh8+ Ka7 -> first check -> no extension 2.Rh7+ Kc6 -> second check -> extend one ply 3.Rh6+ Kd5 -> third check -> extend one ply 4.Rh5+ Ke4 -> fourth check -> extend two plies Contrary what's stated on the page I now believe this system is more worth than +5 elo in comparison with the system to extend every check with one ply. I estimate it now to be +10 elo. My best, Ed
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.