Author: Daniel Clausen
Date: 09:23:38 02/10/03
Go up one level in this thread
On February 10, 2003 at 09:38:16, Sune Fischer wrote: [snip] >I think humans often like to setup some kind of minefield, get into positions >where the opponent is under pressure and needs to be careful not to make a weak >move. It doesn't work with computers, they will calmly find a way through the >maze, and possibly come out on the other side in a better position. [...] >I generally would trust the evaluation of GM's more than computers, but having a >positional advantage and profitting from it is two different things, particular >against computers. I like your example of a minefield. It's a good image for what I often think when I hear that "player X has a positional advantage". A position is either won, lost or drawn. That's it. A "better position" only exists in the area of "imperfect chess". It's not important how bad a position looks like or how many ways there are in order to "go wrong". A position is won, if there's at least _one_ forced way to win and that's it. And here's the analogy with the minefield: Problem: Cross a X*Y field from north to south (a) played by a human being Can't calculate every path thru the minefield and - based on the theory of probabilities - makes mistakes, since the human "only" uses heuristics in order to find a way thru the minefield. (using heuristics is better than random usually, but it's not perfect - if you don't know the difference, replay some KRBKNN games or similar) The human is successful under in the following cases though: -the route doesn't offer many forks, that is, the way is trivial (eg a KPK ending) -there are many forks but most of the ways lead to victory anyway (eg a KRK ending - even I can win this one :) (b) played perfectly Will find a way thru the minefield if there is at least one way. Whether there is more than one is not important _at all_ for perfect play. Now I'm not saying that computers play perfect chess, but it's known/common that in order to transform "a winning position" into "a won position" exact play is needed somewhere. That's typically the tactical blow which wins material. The computers are getting stronger and are almost perfect-playing opponents when it comes to these tactical blows mentioned above. And the stronger they get, the less likely it is for a human to be able to transform a winning position into a won position. So, basically a position X can be "a winning position" against another human, but at the same time can almost be "a lost position" against the computer, because _the path_ from "winning" to "won" is so incredibly small that it's unlikely the human will find it. Ok, I stop rambling here now. Just some food for thoughts maybe. :) Sargon
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.