RE: computer-go: agreement protocol

[Sai To Wang <saitowang_go@xxxxxxxxx>]

> -----Mensaje original-----
> De: owner-computer-go@xxxxxxxxxxxxxxxxx
> [mailto:owner-computer-go@xxxxxxxxxxxxxxxxx]En
nombre de Don Dailey
> Enviado el: Jueves, 05 de Julio de 2001 10:09 a.m.
> Para: computer-go@xxxxxxxxxxxxxxxxx
> CC: computer-go@xxxxxxxxxxxxxxxxx
> Asunto: Re: computer-go: gnu-go agreement protocol
> 
> 
> 
> Yes, the  Tromp/Taylor   rules are  very  much
> based on  all    your...
> 
> Don
> 

I agree that it is necessary to consider in the
protocol requirement for the human/computer and
human/human part. But this can be isolate as “human
extension” of the protocol.

The reason is, if a program can finally “agree” with
something, it means that the program already knew the
fact before the agreement phase. And, if the agreement
phase is somewhat need to repeat the:
agree? = No
play
...
pass
pass
agree?

steps, why not just play to the end at the first
place.

For human/computer and human/human extensions, the
situation is quite different.

For human/human the agreement phase as above is
already working in some variation in the Internet, and
the final scoring is almost always Tromp/Taylor like.

For human/computer if it is not a tournament, the
human side is somewhat dictating the final agreement
and the computer only can suggest the game status or
the computer will not agree in anything for the same
reason that if a program “agree” is because it already
knew the fact.

If it is a tournament, there is always high level
human (compare to a program) to dictate the out come,
so it can be just like the human/human part.

And after all, the “agreement” is never base on the
score, is nonsense to argue about the actual score,
because the final scoring rule is kind of flawless.
What “agreement” has to agree is the life/dead status
of every stone on the board, so the protocol only need
to send this type of information for the “agreement”
phase like we do in the Internet.

For the computer/computer protocol with few human
interventions, I’m still convinced that play to the
end is clear, simple, easy to implement and the most
important is, high accuracy regard the outcome of the
game.

Sai To


__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/