Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: USER Desirement: An Improvement in Fritz, et. al.

Author: Roger D Davis

Date: 12:10:07 05/26/02

Go up one level in this thread


One issue is that the program needs stable scores for a variety of moves in
order to select the best one. Presumably, it will require more time for the
scores for various moves to stabilize, so you might get a slower game than if
just one move is played. As I understand it, Alpha-beta does not return a true
score. Not that more time is a serious problem, it's just something to be aware
of, and on a fast system, it might not even be perceptible.

Another issue is that sometimes the best move is to avoid mate, and no other
move can do that. Or, the best move avoids the loss of a queen, and the second
best move, and all other moves, lose the queen. So, perhaps the second best move
could be chosen where it is not less than the best move by more than a constant,
and the user could select this constant.

So..."Of the 4 best moves, select the one which is just greater than the user
specified constant, and reject all others." Thus, if the user selected -3 as the
constant, the program would tend to fall into traps that would lose the knight
or bishop. If the constant were set to -1, the program would tend to lose a pawn
or make cumulative positional mistakes that would presumably present the user
with other opportunities in the long run.

Or maybe it's already being done like this, or in even a better way, and no one
has bothered to say so.

Roger



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.