Author: Sune Fischer
Date: 14:34:35 08/21/03
Go up one level in this thread
On August 21, 2003 at 17:25:46, Omid David Tabibi wrote: >>I think the competition is rather fierce at the top, so their decision is >>understandable if the choice is between gaining X Elo in real games, or solving >>artificial test positions. > >True, but the position in question was no artificial. It could very well arise >in actual games, and it can certainly arise many times in the huge search trees, >with one such horrible error wreaking havoc to the whole search tree. > >The probability of such an occurrence is very low, but if it occurs it's a total >disaster! The problem for us observers is that we only discover whenever it fails, we never realize that those 10 other games turned into a win because of the pruning. We thus falsely conclude it is hurting the program. >Imagine Junior playing aginst Kasparov in the 6th match, and this position >arises. Junior would draw the won match! Yes, but imagine the odds of that happening! I remember another famous computer that didn't use nullmove when playing Garry some years ago, probably rooted in the same fear. >As a real life example, take the frist game of Fritz vs Kramnik in Bahrain. Had >Fritz had blockage detection knowledge, it wouldn't have played Bg5 and could >have very well gone on to win... Right they can always get smarter, and just imagine how much they could prune if they understood blockages! :) -S.
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.