Re: [computer-go] Third KGS tournament: game-end protocol

[Tim Foden <tim@xxxxxxxx>]

Don Dailey wrote:

> What's the worst thing that could happen?  A program will think it's
> winning but fail to defend it's claim and lose the game.   But it's
> not like the program didn't have a choice.   During the actual game,
> the program is expected to defend itself,  why is it suddenly ok
> not to defend itself when there is disagreement?    
>
> So here is what you could easily have:
>
>  1. William Shuberts kgs-genmove_cleanup protocol
>  2. A programmers choice whether to implement the protocol or not.
>  3. A programmers choice HOW to implement it if he does implement it.
>  4. A server that can easily score all games without human intervention.
>
> Even if your program accepts the "kgs-genmove_cleanup" it is up to the
> program what it will do with the information (the knowledge that there
> is a disagreement) and this can't be enforced by the rules of GO so it
> can't properly be a requirement.
>
> But a good program will want accept this optional command and clear off
> the opponents dead groups when it realizes the opponent is unaware that
> his group is dead, otherwise it risks losing some points.   
>  
Hmm... how about this scenario..

There are two players, A and B.

Both A and B have passed.
All of A's groups are alive.
All of B's groups are alive.
A thinks all of it's own and B's groups are alive.
B thinks all of it's own groups are alive, but that one of A's is not.
A implements the protocol to play on... but doesn't need to play.
B doesn't implement the protocol to play on, so doesn't play.

I think this could result in the kind of problems we saw on Sunday.  
Because B isn't playing to kill the stones he thinks are dead, somehow 
the protocol gets in a loop, where both programs continually disagree, 
but play does not continue.

I think this is actually a worse scenario from a practical point of 
view.  :)

~~~

How can we fix this problem?

Maybe this would do it -->  In the scoring phase, we actually have more 
information than we have whilst playing... we know if the other player 
thinks differently about the status of a group.  So in the scoring 
phase, if there is a disagreement, the players should play it out until 
they agree.  Agreed?  So from the above example, I would say that the 
player B, that thinks a group is dead in disagreement to A, should move 
to try to prove his case.  A is allowed to pass, as he doesn't need to 
prove that a group is dead. If B actually passes, A's scoring should be 
used for the game. 
e.g.
A. PASS.
B. PASS.
* Score disagreement. A thinks alive, B thinks dead.
A. PASS.
B. Must make a move, or abide by A's scoring.

I think this would fix the problem for one prog that doesn't respond 
correctly.

~~~

If both programs act like B, then maybe we could say the rule is this.

If there is a dispute, then the prog that thinks stones are dead _must 
make a move_ or the game will be scored using the scoring of the other 
player.

e.g.
A PASS
B PASS
* Score disagreement, A thinks some of B's stones actually dead, B 
thinks some of A's actually dead.
A. Must make a move, or abide by B's scoring.
B. If A moved, {B can PASS}, else {B must make move, or abide by A's 
scoring.}
* play continues until 2 consecutive PASSes.
If moves have been made, we {rescore, and repeat disagreement protocol 
if necessary}, else {end of game.}

Thus we guarantee to end a game, in any case.

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