Re: [gnugo-devel] XiStrat Go Variant

[kai brommann]

hi,
first of all thanx for your feedback

> don't know if you have a strong go engine as one of
> your goals; 

yes, beyond all the other goals :-)

> would suggest you consider the Tromp-Taylor ruleset
> for Go; it's much
> simpler than Japanese rules and has fewer
> exceptions.
> 

will look at it, I took the Japanese because the
standard Go teaching books seem to do as well, so it's
a trade off between standard folks and mathematical
approach

> If you wish to attempt a Go engine of any strength,
> I think you will find
> you need a lot of Go-specific code. 

will especially implement what can be considered game
domain-independent, or if Go-specific, then at least
usable for all graphs, in order to avoid over-tuning
and loosing the overall general mathematics out of
sight

> I suggest you consider implementing a
> GTP interface to the
> Go code; if you do so, then you can put GNU Go's
> regresssion tests to use
> as test cases, and also use the twogtp scripts to
> play games.  

I have a generic protocol in mind (applying to all
games), and for Go this could result in a GTP
superset, so perhaps finally one could play against
GnuGo on quadratic boards some time in the far future

> have an engine-side implementation of GTP in Java
> that is in large part
> complete; if you're interested I could clean it up
> and release it under
> the GPL.
> 

yes, this sounds interesting. if you can afford the
time, please do (seems then I would have to change
from LGPL to GPL), though I can't promise when the GTP
feature will be attacked within XiStrat (the ToDo list
is already quite long...) :-)

bye for now,
kbr

__________________________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online, calculators, forms, and more
http://tax.yahoo.com