[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: Engineering (was: Most simple Go rules)



On 28 June 2001, Heikki Levanto wrote:

> All this takes is a small addition to a program: If it knows it has lost the
> game, it should resign. If its opponent passes, it should resign even
> easier, if it believes to be well behind. This kind of behaviour is not
> specified in any ruleset that I know, but is quite common among human
> players.
> 
> If 90% of the games end in resignation, the details of scoring become
> somewhat less important. Of course both programs should score by the same
> rules, so that they have at least a theoretical chance to reach an
> agreement!


I understand that human players do resign lost games out of respect
and not to waste another person's time.  But I am curious whether
you think such behavior in a program might not deprive the programmer
of valuable information about the program's performance.  I understand
the point of a tournament is to win, but still, regardless of whether
your program is on the losing or winning side, is it not still possible 
to obtain useful data about the program's engame evaluations/tactics
by playing the game out?

My thought would be that in program vs. program games, I wouldn't really 
mind "wasting" another program's time if some benefit could still be 
gained for either party.

Daniel Hallmark
P.S. Hope this doesn't sound argumentative.  I freely admit I don't have
     a working program and have never competed in a tournament so I really
     am interested in other's opinions.