Author: Bruce Moreland
Date: 11:17:57 06/03/98
Go up one level in this thread
On June 03, 1998 at 13:58:02, Thorsten Czub wrote: >Sometimes - I guess - you have to make a weak move, that is NOT the best >move in the position - to realize a plan of yours. >Any grandmaster knows this. And of course any grandmaster understands >the point that there are many similar moves within the draw-range. >Chess-programmers often do not realize that you cannot really order the >moves for BEST criteria. Since you will never find out accurate enough. > >So chess players know by experience that many moves are ok. Since they >know this, they know that sometimes you can play a 2nd best or 3rd best >move to >realize a plan you need to win. >I ask myself how alpha-beta can ever catch this behaviour ! Alpha beta can attempt to drive toward trappy positions, if the evaluation function has some sense of what a trappy position looks like. What it can't understand are cases where it's the tree that is trappy, not the position. If there is easy way for the opponent to get to a position that you evaluate as +0.20 in your favor, you will choose it over a situation where there is a very very difficult means for the opponent to get to a position that is +0.18 in your favor. Obviously, you want to give them more chances to make mistakes, but alpha-beta won't do this itself. In order to get to the terminal nodes, each side must make a series of choices, and some of the choices are more difficult than others, and alpha beta absolutely does not pay attention to this. It might get a sniff, because if you are in trouble, you are busy trying to stay alive rather than maximizing minor positional features, so your score down a dangerous path might be less, but this isn't a certainty. bruce
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.