Re: [DISCUSSION] Game Tree vs. Pattern Matching

[Joan Pons Semelis <joan@xxxxxxxxxxxxxxxx>]

David wrote as answer to:

	At 02:26 AM 10/30/98 +0200, Hyoungsoo Yoon wrote:
	>Robert Jasiek wrote:
	>> 
	>> > An good evaluation function is almost all we need
	>> > to make a good go program.
I thuoght this 15 years ago and have not yet changed my mind. 
	>> 
	>> For this very reason my CG development has been done with paper
	>> and pencil only during the first year. This will continue for
	>> some time...
	>
	>For the very same reason, I only use my brains and nothing else.
	>I'm planning to do so during the first decacde of my CG development.
	>
	>Hyoungsoo
	>
	>
	I think designing on paper for a long time is a mistake.  You learn
	a lot about your evaluation ideas from trying them out in actual
	games.  I can't count how many great, general concepts I've had for
	go evaluations, and found through experiment that there were cases
	I had not considered.

	Davi

I can't count how many great, general concepts I've had for go
evaluations, and found by applying them to actual games to see how
would they behave that were bad. You only need pen and pencil,
or even less to do this. Of course everyone has his own system,
and also there are sometimes ideas that must be tested actually
programming them, so having a friendly framework to work in is
really usefull.

Joan Pons i Semelis joan@xxxxxxxxxxxxxxxxx