Author: KarinsDad
Date: 15:45:38 06/11/99
Go up one level in this thread
On June 11, 1999 at 18:10:31, Dave Gomboc wrote: [snip] > >Keep doubting. :-) Chess has a lot of slack in it. You can make a mistake and >still save the game. I have read that it takes two or three "GM-level" errors >for a position to become lost. > >(I think that) you think GMs play substantially less than perfect, because this >is implied by the belief that if two perfect players could start from a position >that GMs rate as "slight edge for white", the result would be a win for white. >I do not share this belief. (Though perhaps I am overstating your belief... >feel free to clarify. :-) > Yes, I guess I do think that. There is no doubt about it. GMs (and especially superGMs) play a great game of chess. But we do have proof that they only understand chess to a limited level. Examples of the proof: 1) Two GMs look at a position for several hours and cannot come up with the proper continuation. Capablaca looks at the position for several seconds and finds it. 2) GMs analyze an endgame position to death. They come up with the conclusion that it is a draw. Later on, tablebases prove that it is really a win for one side. 3) Chess programs play 1200 level chess at ply 4 (approximately). They play 1400 level chess at ply 6, etc., etc. Everytime a program adds one more ply to it's ability to search, it gets a little stronger. Programs have been able to beat GMs for quite some time now. That means that the GMs are not consistently able to outperform a program for an entire game. When it goes from ply 14 to ply 15 to ply 20 to ply 40 to ply 200 (obviously, the latter set of plys are not achievable within the foreseeable future of technology), it will continue to play better and better chess. The reason that GMs and superGMs play SO WELL is that they have experience and intuition on their side to prune down the number of positions they must analyze. However, most GMs only look down 10 ply or so when analyzing (depending on the position). There experience translates this into a 12, 14, or 16 ply implied search depth. So, it is inevitable that moves that today are considered best in certain positions, will one day be considered inferior (just like positions 50 years ago are considered inferior today). The difference is that programs of the future will determine this very quickly as opposed to via years of playing and analyzing. So, my conclusion has to be that ANY mistake is enough to lose the game. However, since GMs are human, they do not have the ability to take advantage of a 1/10th pawn mistake. Computers someday will probably have that ability (it may still take them 100 moves to capitalize on it, even against another similar strength computer). We do not yet have proof of it, but it seems likely. It seems likely that as long as you add one more ply (all other things being equal), that you will continue to improve. And with improvement comes the ability to capitalize on smaller and smaller mistakes. KarinsDad :)
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.