Re: [computer-go] Computer Go tournament at EGF

[Don Dailey <drd@xxxxxxx>]

   From: David G Doshay <ddoshay@xxxxxxxxxxxxxxxxx>

   On 12, Feb 2005, at 9:40 AM, Don Dailey wrote:

   > The point he is making is that you can't isolate the execution time of
   > the algorithm from the playing strength.  It IS in fact easy to write
   > a perfect go program given no time constraints (just search until
   > solution.)

   Easy in theory. I again challenge you to do it. A 1 dan program does
   not exist. Please make one that will play a 1 dan level game in
   whatever time period you wish.

   >  The issue isn't how GOOD your program is, it is how good
   > it is in a given amount of time.

   Your issue, not mine, so you get to chose the time for the above
   challenge.

What I don't understand is how you have presumably built a scalable
system based on GnuGo, harnessed a lot of computing power to make it
work, and yet you don't understand the relationship between computing
power and playing strength, claiming it's only an issue for me.

If your source is open, then I could take it and modify it to run on a
serial machine.  Would it play better than just running GnuGo by
itself?  Would it take longer to make a move than GnuGo?   Would you
say there is a relationship here?    

Why does your system require a lot more time to make a move than GnuGo
does despite the network you have at your disposal?    Why does it play
better than GnuGo?    What am I missing here?

- Don


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