Author: Fernando Villegas
Date: 09:19:51 01/15/01
Go up one level in this thread
The so called positional human assestment is something far to be a misterious thing. I have many times here made a more or less detailed examination of what happens when a human being proceeds to take a decision in the course of a game or whatever. Men just go thorough different stages oriented to different goals, so they are not compelled to calculate everything. I have given the example of thew tennis player: he is not under the pressure to calculate every posible trayectory since the very split of a second the racket of the opponent smash the ball; he just do a serie of discrete assestment thaty feed back in previous assestemnet. First he see the ball goes to his left and then he begin to go there; as he run he ssee the ball is in a flat trayectory and he prapares his arma to strike at low altitude and so and so. He get feedbacks of what is happenning and so each specific "calculation" take as a task a very limited scope of chances. In chess programming still prevail, on the contrary, the all-round-task code, where the program examines positions, in each ply level, with the very same full evaluation code. It does cut some branches, but still every node that is left pass over the entire code. What a chess player does? He divides the problem as the tennis player. Maybe his very first evaluation is just to grasp where the action is: queen or king side. Then and from the beginning the next analysis will put aside as irrelevant the pieces located that cannot intervene. Maybe then he will see if he is under attack or the rival. So, must he looking for defensive or attacking moves? And so and so. In each phase he just see and examine a few things. Only from time to time he calculates a good number of moves or lines. Probably just at the end, when he has a clear idea where the esence of the game is by now. I understand that this approach has not been essayed, at least not at full. At most special codes are used for endings. But of course the problem is not just divide the game in openning, middle game and ending. Fernando
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.