Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fritz5's learning is not productive against humans

Author: Robert Hyatt

Date: 06:52:44 11/18/98

Go up one level in this thread


On November 17, 1998 at 06:25:54, blass uri wrote:

>I found that fritz5 in the ssdf games repeat the same line it won again and
>again
>
>It is not productive against humans because after some games it is easy to
>predict fritz5's opening
>
>I think that it is better against human to avoid losing lines but not to repeat
>winning lines again and again
>
>Uri


Think about this:  a huge opening book.  It thrashes you in a particular
opening.  It can go on to another opening where it may win/lose/draw, or it
can play the winner again.  To maximize its performance rating, playing the
same opening again is a good idea.  If you find your mistake in the second
game, it doesn't pay off.  But if it takes you 10 games to finally "break"
that opening, your rating took a beating.  And after one loss, it will move
on to the next opening anyway...

IE I keep throwing you inside curve balls until you hit one, before I move to
my fastball, or slider, or etc...  Or I throw you a curve, you swing and miss,
I move to the fastball which you knock out of the park...  I go back to the
curve.  :)

Or in the case of Fritz, I stick with the curve until you prove you can hit it.




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.