[computer-go] IEEE Spectrum article by Deep Blue creator

[terry mcintyre]

----- Original Message ----
From: Don Dailey <drdailey at cox.net>

>I think [Hsu] is betting on null move proving - but I'm real skeptical that
> it will be effective in Computer Go.   It will indeed reduce the tree
> significantly, but this comes at a qualitative price that is not so bad
> in Chess but is a lot in Go.



Hsu also discusses the gains from caching life-and-death analysis of groups. I suspect that
this will greatly reduce computational effort, once an efficient mechanism is implemented.
Existing monte carlo programs cache information about playable/non playable points; when
augmented with knowledge about life and death, search should more quickly home in on crucial 
lines of play.

I've been playing against Mogo the last few weeks. It has a very interesting style of play, and it 
often does quite well in tactical analysis, but sometimes it misses a key move and fails to kill or
fails to preserve a large group - game over! A good life-and-death cache would be a definite improvement.

Caching parts of trees works better in Go, since well-defined sections of the board can sometimes be 
partitioned from the rest of the board. Where such partitions leak, analysis is likely to be critical; 
for example, ladders and ladder breakers can extend across the board; invasions often depend on 
cutting points halfway across the board. 
 



       
____________________________________________________________________________________
Moody friends. Drama queens. Your life? Nope! - their life, your story. Play Sims Stories at Yahoo! Games.
http://sims.yahoo.com/  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20071002/42139a4d/attachment.htm