Computer Chess Club Archives


Search

Terms

Messages

Subject: Scores in Gambit: Maybe Prospective Instead of Retrospective Evaluation?

Author: Fernando Villegas

Date: 09:47:28 11/13/00


Hi:
In a post someone compared the high positive score got by Gambit Tiger playing
e5 against Hiarcs 7.32 with the same move played by Crafty -if setted in that
position- but with a score of minus .23 or so.
Now an issue emerges, I think. What is "best"? How a program measure what is
best? Seems to me that maybe there are not only differences due to diferents
codes, but even two very different ways - or paradigms some one could say?- to
do it: one is trying to ponder each aspect of the position in an objetive way,
lastly in terms of how good the position currently is, NOW, when the calculation
stop. So an open column is well scored, of course, but not more than what
deserves in the current position. The other way could be to stimulate certain
behaviour giving to some aspects a clearly greater weight than what they coul
deserve IF seen only according to the actual position. In this case an open
column could be given a lot greater weight not because is so really good, but
just to give the program a reason to chose it and follow a path that the
programer dims preferable for knowledge considerations. We can call this
prospective scoring intead of clasical, retrospective scoring. Sorry for the
words. I cannot resist the temptation to add my voice to the production of new
fancy words to this nos so fancy world.
Anyway, If my memory serves me well, I believe that iit was precisely what
Kittinger did with his Zaphire handheld unit. An analisis by a french writer in
the now deceased chess computer magazin "La Puce Echiqueenne"  was very clear
about that. I remember that, according to the research of this guy, each time
Zaphire dims to be best, at once he gives a tremendous score to exchanging
everything so to simplify and increase his relative advantage. Maybe is the case
of Gambit?
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.