Author: Dann Corbit
Date: 16:24:33 07/12/00
Go up one level in this thread
On July 12, 2000 at 18:57:13, Gareth McCaughan wrote: >On July 12, 2000 at 17:21:30, Dann Corbit wrote: > >> Consider a program that searches one million NPS and does no positional >> eval at all. It only looks at the material, period. It has no openings >> database, and no EGTB. It does not bother to order the moves at all >> and does (at least) use NegaScout to search efficiently. >> >> Consider a second program that searches at 1 NPS on identical hardware but >> considers all kinds of positional evaluation and chess knowledge implemented. >> It can do everything Yassar Sierwan talks about in his books (given enough >> time) as far as knowing the value of "pigs on the seventh" etc. It has >> opening (GM inspected) database and EGTB. It orders the moves perfectly >> 99.999% of the time and also uses NegaScout to search. >> >> Program 1 makes 13 plies and kicks the butt of program two which makes one ply >> at 60 seconds per move. >> >> Program 1 has more chess knowledge? If winning is the only measure then the >> answer is yes. > >If program 1's move ordering is sufficiently terrible, it won't get >to 13 ply at 60 seconds per move. > >If program 2 "orders the moves perfectly 99.999% of the time" then it >makes the right move 99.999% of the time and will beat the pants off >program 1. :-) If you are only seeing one ply deep, the move ordering is practically irrelevant for all intents and purposes. The move that looks best at one ply is tried first. You have not even looked two plies forward yet (IOW -- you don't even know what the 1st response is!)
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.