[computer-go] Re: Interesting problem

[David Doshay]

On 3, Jan 2007, at 1:32 PM, Sylvain Gelly wrote:

> Again sorry for this incredibly long game, I was expecting that  
> programs resign before the end. The politness by passing is enabled  
> only against human.

I do not think that any apology is needed. The length of the game was  
due only to a setting you have that is totally appropriate for a  
Chinese rules tournament game. And SlugGo is set not to resign, just  
to pass, which I think is also appropriate in a tournament game,  
especially a close one. I know that I would never resign a game I  
thought I had lost by less than the komi. I would pass, expect the  
opponent to pass, and then count it openly. It is also the case that  
I would respond to repeated play by my opponent exactly as SlugGo did  
(except that I probably would have made several simpler captures to  
make the situation obvious to the opponent).

My point is only that the consideration of the rules we use says  
something about what we expect our computers to do, and what we are  
willing to watch them do as a consequence of our rule set. There are  
often competing reasons, and often unexpected results. In this case I  
think the consequences are completely predictable with these rules,  
and with Tromp-Taylor rules even more so: very long extended endgames  
that humans 1) would never play, and 2) make derisive comments about,  
leading them to walk away with a very low opinion of the state of  
computer Go.

There are times and places where Tromp-Taylor rules are clearly best,  
such as cgos-type servers where a large number of games must be  
scored automatically and without human intervention. I just think  
that we will eventually will need to accept that open play in a  
public forum deserves a different set of considerations.

Cheers,
David