Author: Fernando Villegas
Date: 06:17:07 12/29/97
Go up one level in this thread
Hi Don: I have heard before this issue about the difference between to bee good for solving problems and to be good for playing a game; what I have not heard before, neither thought in it, is the simple idea you made me see that a game is a kind of a series of problems that arise on the fly. Now, putting both things together, I have thought the following: a) A game in itself can be seen, also, not only as a series of problems, but as a problem in itself, a big one where each position or problem is just an element of it. b) As much in the resolution of a single problem the different aspects of the position must be carefully weighted (not only give all the score to material advantage or to put a knight in the center and so and so) or the solution will not be met, so the resolution of the big problem that is the game maybe needs a similar approach, but in different level. What I mean here is that perhaps the purpose to get the best tunning to solve a set of problems on the ground of a supposedly known best and second best solutions should be replaced by the purpose to get the bigger number of sound positions, even if sometimes that does not involves to pick up the very best move. My thought is that in a real game between human or computers what counts is not just to be capable of picking up a killer move; that’s can be decisive, but not very often. Usually the winner juts made a greater number of sound positional moves until that produced enought positional and finally material advantage by discrete and succesive steps. GM are not GM because they see all tactical shots -in the contrary, they miss a lot- but on the ground of better positional moves. They know what is better or at least good in a position and so they even does not calculate too much and so sometimes they miss sometimes tactical opportunities. So, I can imagine a test where the goal is not to pick up the very best moves, but to play the larger set of sound moves in minimal time. Can we call this a kind of fuzzy logic? Using this kind of approach, a program -or version- that has never pick up a killer move in a test could get a better score than another that did it, if the first did get more sound moves in less time for the complte set of problems. Have you teste that approach? Fernando, the guy with the great mouth
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.