Author: Christophe Theron
Date: 09:20:13 11/05/98
Go up one level in this thread
On November 05, 1998 at 11:54:00, Dann Corbit wrote:
>Are you using a neural net for the decision between tiger/rebel, or is it by
>temporal sequence [phase of the game], (or simply none of my business)?
>;-)
So far we have not decided, but there are a number of considerations we can take
into account.
We have first to determine in which "part" of the game each engine excels. Maybe
we can decide that by playing a number or Rebel against Tiger games and viewing
them manually.
We may end up with rules like "if closed middle game position, use X", "if in
time trouble in an open position use Y", "if playing queen opening at slow time
controls use X", "if opponent is likely to be human use Y"...
So there is a open territory to discover. We can even create specials modes
where the engines are choosen randomly at each move during the game.
I really don't know how this mode would play. The interesting thing about the
Rebel-Tiger association is that the engines have very different styles. When you
play Rebel against Tiger, the games are never boring. There is always a lot of
action, and you cannot say which is going to win until the real end of the game.
The general problem with computer chess engines, as it has been pointed out many
times, is that they have "holes". Suddenly, in a very well played game, the
engines looses track of the plan, plays one or 2 weak moves, and looses.
With 2 different engines you can cover these "holes" in such a way that the
resulting engine seems to have no hole at all. The condition is that the engines
are different enough so that they don't have the same "holes", which is already
the case, and that you are able to foresee in a position which engine is likely
to blunder, and we are going to work on this.
Christophe
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.