Re: [computer-go] I know we disagree,but I choose to do nothing about it.

[David G Doshay <ddoshay@xxxxxxx>]

On 26, Jul 2005, at 2:59 PM, Eric Boesch wrote:

> As one of the GnuGo team (Paul?  Gunnar?) said a while ago, and I 
> repeated.  I do think people have been too reluctant to admit their 
> selfish motivations in this discussion.  A poll of whose programs can 
> actually handle Tromp-Taylor would, I think, be more interesting than 
> doubtful arguments that there is something inherently absurd about 
> Tromp-Taylor computer tournaments.
SlugGo can play by these rules, but I still do not like it.

> I don't think anyone disputes that in a computer tournament using any 
> of the rules (Chinese, AGA, NZ) that have all-stones-on-the-board-live 
> as a last resort, it would also often end up as the first resort, just 
> like Tromp-Taylor.  Will this add dozens of moves to a typical game?  
> Yes.  Will it add hundreds, or give players a "strong incentive" to 
> play unchallenging and theoretically useless stones?  No.
On this point we simply disagree.

> The key protection is the same one as always:  whether programmers' 
> intents are commercial or not, they value their or their programs' 
> reputations, and repeatedly attempting very cheap, unsporting wins and 
> mostly failing anyhow wouldn't help in that regard.
Different programmers care about different things, but claiming a win 
in a tournament does not carry with it a description of how it was a 
win. It was a win under the rules of that tournament. It seldom ruins a 
reputation to say you won.

> Some programs, especially ones written from an area-rules point of 
> view, may plonk down dead stones as an unintended consequence of their 
> programming, but that's about it.  It's not profitable under any 
> ruleset to plonk down a stone that will just get captured again,
Under some rule sets there is a one point penalty for such moves if the 
opponent recognizes the lack of value in the move and passes.

> and while area versus territory scoring can make a difference in the 
> theoretical value of a move, the specific dispute resolution protocol 
> does not -- not unless the program's working theory is that its 
> opponent doesn't know how to finish the game, but I repeat that there 
> haven't been a rash of programs designed to play cheap ugly tricks.
Perhaps, But if this becomes the standard protocol then programs will 
stand to benefit from what you call a cheap ugly trick. If those are 
the rules then it cannot be called a cheap ugly trick, it is the rules. 
Even Don agrees with me that if there are points to be made that way 
then (some) programs will end up doing it.

But the simple fact is that we have exhausted the arguments and the 
tournament directors will decide how they will run their tournaments.

David

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