Author: Uri Blass
Date: 19:12:30 06/14/04
Go up one level in this thread
On June 14, 2004 at 17:35:30, Steve Glanzfeld wrote: >On June 14, 2004 at 17:18:33, Sandro Necchi wrote: > >[...] > >>Look this simple answer and think about: >> >>1. a !! move is not 100% a sure win; it only give better winning chances most >> of the time. >> >>2. a ! move gives only equality or a small edge in most cases. >> >>3. 2 ? moves are good enough to lose! >> >>4. a ?? moves is most of the time a loosing move. >> >>So we can conclude that: >> >>1. to be able to find the best moves in many positions not necessarely makes >> the program stronger. >> >>2. To make several mistakes or weak moves does make the program weaker! > >Sorry, I still don't get this. The above is clear to me, but: A good test will >check if for example how often the ! and !! moves are are found from the test >set, and how often ? and ?? moves are avoided. Let's imagine we have a program >which is able to do so very often, while another program can do that much less >often. > >What's wrong now when I say the first program must be stronger than the second? > >I don't assume you'd say it doesn't matter if a program finds !/!! move or not, >or if it avoids ?/?? moves or not :) I thought that's all what chess programming >is about: Finding the good moves, avoiding the bad moves, more often than the >competitor can. > >Steve The problem is that chess is a game and not list of single positions. I can mention 2 points: 1)An evaluation can be dependent on the history of the game or the history of search of the programs and test suites do not reveal this information. 2)imagine program that in 95% of the cases play tbe best move but in 5% of the cases play stupid blunders because of some bug. This program may do well in test suites that are not too easy but bad in games because the 5% of stupid blunders will usually decide the game against it. Uri
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.