Author: chandler yergin
Date: 11:44:37 04/22/05
Go up one level in this thread
On April 22, 2005 at 13:03:54, Eric Oldre wrote: >On April 22, 2005 at 06:10:40, chandler yergin wrote: > >>A hypothetical challenge: >> >>To get a 'fair' playing field, and to truly test the performance of a Program >>let's not use Opening Books DataBases or EGTB's. >> >>Get a little school girl age 10 -12 who has never played chess, >>teach her the rules of the game. >> >>Now, both the Computer & the girl know the Rules. >>Have a 10 game match under Tournament time controls.. >> >>Who do you think will win? > >All you taught the child was the rules of the game. Well some common sense would dictate she be showed how the pieces move, what Mate & Stalemate is, some elementary Mating positions and tactics.. But she would not have any Tournament experience. The program probably has >chess knowledge in it's eval that took years to figure out. Yes.. but it can't think or plan. With experience, the girl will learn quickly. The Computer won't learn anything new. Turn 'learning' off. > >It would only be fair to cut away some of the chess knowledge of the engine too. >Give it a material only eval function. And to be really fair, only give it a >value of 1 for each piece, since the child doesn't understand the relative >values of the pieces. > >If you did all that, you might have a interesting game to watch. No, I think making the Match longer... maybe 100 games as a real test of strength for Human and machine. I think the Computer will need all the knowledge that has been programmed in. Let them both play without books, databases, etc..
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.