Re: [computer-go] Pattern matching - example play

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 14:13 2-12-2004 -0500, Don Dailey wrote:
>   X-Original-To: computer-go@xxxxxxxxxxxxxxxxx
>   DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com;
>
h=received:return-path:reply-to:from:to:subject:date:message-id:mime-version
:content-type:content-transfer-encoding:x-priority:x-msmail-priority:x-maile
r:x-mimeole:in-reply-to:importance;
>
b=FfHgK4ceM+ueOHdW0RJiFM2TPHtBkOfUmzRo8Vkmm96vYJJkLKDJxq/XCKvftzUY8jnxb4yN+p
V+f3O4kWVdb4ja0loySsw7YmdMy/pHAwuS7YXf7zHq54zeotGk1Z4iGuaSddtjd5PzixnGmOz75e
5x9W0gfpd5/yYuNmWRdvg=
>   From: "Mark Boon" <tesujisoftware@xxxxxxxxxxxxxxxxx>
>   Date: Thu, 2 Dec 2004 12:42:51 -0200
>   X-Priority: 3 (Normal)
>   X-MSMail-Priority: Normal
>   X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441
>   Importance: Normal
>   Reply-To: tesujisoftware@xxxxxxxxxxxxxxxxx, computer-go
<computer-go@xxxxxxxxxxxxxxxxx>
>   Sender: computer-go-bounces@xxxxxxxxxxxxxxxxx
>   X-Scanned-By: MIMEDefang 2.42
>
>   > -----Original Message-----
>   > From: computer-go-bounces@xxxxxxxxxxxxxxxxx
>   > [mailto:computer-go-bounces@xxxxxxxxxxxxxxxxx]On Behalf Of Erik van der
>   > Werf
>   > Sent: Wednesday, December 01, 2004 17:01
>   > To: computer-go
>   > Subject: Re: [computer-go] Pattern matching - example play
>   >
>   > Personally I think that at the current level of computer Go 9x9 is very
>   > interesting. After all beginners start on 9x9, not on 19x19. If we
>   > cannot even build a decent 9x9 player how can we ever hope to succeed on
>   > 19x19?
>
>   That may be true but still not relevant. If we're mainly interested in
>   making a Go program play 19x19 (which I think most of us are) then
'solving'
>   9x9 does not guarantee to bring us much closer to our goal of making a
19x19
>   program. The effort involved in making a program play well on 9x9 may
not be
>   a good investment when your goal is really 19x19.
>
>I'm not so sure.  The thing  that's really appealing about 9x9 is that
>your whole testing/learning cycle can  go much faster.  I think it's a
>really  nice way  to start  with, for  example, programs  that  try to
>learn.   I  really  believe  most  of the  lessons  learned  could  be
>tranfered more or less directly to any board size.    So the cycle
>might be:
>
>  1. Figure out what and how you need to build a strong 7x7/9x9 go program.
>  2. Build and test it.   
>  3. Reiterate with bigger boards.
>
>I'm not  sure anyone has even  created a really strong  7x7 GO program
>have they?  Can any of us create  a 7x7 that can hold it's own against
>any player?  I doubt it.  I think it's not so crazy to try to get that
>right first.

Here i agree 50% and disagree 50%. Commercially i disagree.

I just played my first go game after a year against smartgo at a 19x19 and
i can definitely positively again write here that 19x19 is *so much*
different from 9x9 go.

In 19x19 the corners are very important. In 9x9 overall board control is
more important. What works in 9x9 go doesn't work at all in 19x19.

However i do agree that from search viewpoint the lessons learned in 9x9
are very important.

Yet it is a fulltime job to develop a good program for 9x9 whereas such
effort can be better spent to 19x19 as the majority of the time goes to the
move selection and evaluation anyway.

It's clear that at todays fast hardware when puttin geffort in 9x9 you can
play it at a very strong level far stronger than coming 10 years will be
achieved in 19x19.

Again i'm not saying 'world champs level'. Current chessprograms despite
what the programmers all cry by giving a few GM's some money to play
against their products after signing a certain contract with certain
conditions, they aren't that level either.

>Probably  one would use  techniques that  remain applicable  to bigger
>boards.  So  things like  brute force search  on 7x7 boards  might not
>really be a quickly scalable approach.
>- Don

How do you find a strong go playing programmer who is willing to waste his
time onto a small board program?

The combination strong player, good programmer and good game programmer is
real hard to find on this planet. And usually not that rich that he can toy
the rest of his life without any worries.

Perhaps approach Amir Ban for that.

I read on the internet some wallstreet news that the company
(www.msystems.com) was sold for over half a billion and he's suing them now
for around a quarter of a billion or so :)

If Amir declines then there is little hope to offer to you :)

Vincent
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/