[Computer-go] kgs-genmove_cleanup again

Petr Baudis pasky at ucw.cz
Mon Mar 29 09:53:10 PDT 2010


On Mon, Mar 29, 2010 at 06:06:41PM +0200, Erik van der Werf wrote:
> On Mon, Mar 29, 2010 at 5:06 PM, Petr Baudis <pasky at ucw.cz> wrote:
> > The easiest way to fix this for me will be to just avoid passing if any
> > stones are considered dead.
> 
> Well, you can't remove your own dead stones ;-)

Sure, but if the opponent does not cooperate, I will just have to keep
playing until there are no valid moves left. Then I pass; if I can win
the game if the opponent removes my dead stones, I can certainly also
win if the opponent does not.

So actually, I might avoid passing just when all the _opponent_'s dead
stones are removed from the board.

> >  This will make the game much more tedious
> > for the human, but it's what wms wishes. :( I don't want to do ugly
> > second-order hacks like second-guessing final_status_list based on last
> > genmove command.
> 
> It is not guessing! The program should already know that the rules
> have changed for the cleanup phase. When final_status_list is called
> (for the second time) it can take this information into account
> perfectly.

final_status_list has a clear definition and no exceptions for that
definition are stated in kgsGtp docs; in fact, it even strongly implies
that final_status_list should never be called after kgs-genmove_cleanup
starts being issued. Also, kgs-genmove_cleanup says simply "do not pass
until all dead stones are removed", not "play as if all stones on board
are alive". Under the second definition, modifying final_status_list
output might be sensible, but certainly not under the first one.

So, given how my first attempt to freely interpret the kgsGtp interface
ended up, I will rather opt for following the specs to the letter this
time around. (Also, perhaps I feel a little stubborn about this right
now. ;-)

> If it couldn't, then what would happen if the human marked your living
> stones dead after the cleanup phase? We known that this happens
> frequently in unrated games, but in rated games it is supposed to be
> impossible. Given that on disagreement the server has no impartial
> knowledge to distinguish the living from the dead it should therefore
> force the human to do the actual removal of the bots dead stones as
> well (or disagree until eternity).

I fully agree with that. Well, it doesn't...

-- 
				Petr "Pasky" Baudis
http://pasky.or.cz/ | "Ars longa, vita brevis." -- Hippocrates



More information about the Computer-go mailing list