Re: [gnugo-devel] No connection better than weak connection (Newpatch)

[SP LEE]

> SP Lee wrote:
> > The similar connection is safe near the edge, so I have modified CC509 
> in
> > the following patch. At the same time I have modified the pattern 
> Shape83 to
> > prevent one of the regression fail. In the above mentioned game, gnugo 
> plays
> > F8 at move 117.
>
> After looking closer at this patch my conclusion is that this is not a
> proper solution to the problem of intransitive connections. Also the
> regression results are not all that good and it seems likely that the
> change will backfire in many common positions, wasting moves on
> unnecessary connections.
>
> The proper solution to the intransitivity problem will require some
> redesign of the entire dragon amalgamation procedure. The readconnect
> code is mostly good enough to detect such intransitivities, we just
> need to understand how best to use that information.
>
> /Gunnar

Maybe my solution is not a good one. However, my original thought was not 
only to solve the intransitivity problem. The eventual solution to 
intransitivity problem would be to check the connection between a weak group 
(not worm) to another group to make a strong dragon together. There are 
connections that make less aji and there are also connections that are more 
solid. We have to balance the effort to recognize all the aji's (threat to 
disconnect) after the connection and the effort to make a connection with 
minimum aji's. To my understanding, the latter is simpler.


SP Lee

_________________________________________________________________
The new MSN 8: smart spam protection and 2 months FREE*  
http://join.msn.com/?page=features/junkmail