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

[Robert Jasiek <jasiek@xxxxxxxx>]

Mark Boon wrote:
> In any ruleset there must be an arbiter who gives the final score.

If we speak of tournament rules, yes, more or less. However,
some tournament rulesets allow the players to create a final,
wrong result agreement that even an arbiter may not change.

> I see a lot of people saying that playing to the end is no problem for
> humans. I agree it's usually not a huge amount of work, but it shows little
> understanding of human play in reality.

What do you call "human play in reality?" Do you observe what
happens due to some rules or do you collect data what players
would prefer if they were allowed to choose? Surely, different
players have different preferences. Furthermore different 
periods of time have had different preferences.

> The Japanese rules have come into
> existence exactly because playing out to the end is troublesome and a
> quicker way of getting to the end-result has evolved.

This is a reasonable theory. However, inhowfar did they notice
the effects of their changes? Did they even notice that they
added a group gift?

> We all know that the
> Japanese rules have theoretical problems, but they are very practical

Could you please explain this? I consider them to be very
impractical because
- scoring includes more than the board
- prisoners must be kept correctly
- cheating with prisoners is possible
- beginners do not understand what is scored
- players do not understand exactly why something is scored
- referees often fail to explain exactly players why something 
is scored
- the game procedures are by far too complex
- there are too many exceptions and gaps
- counting destroys the position
- counting allows cheating
- counting easily allows accidental mistakes
- counting is slower than point by point half counting for
Chinese scoring

> and
> fullfill a human need to speed up the end of the game.

Since there are so many possible sources for accidental
or intentional mistakes, I have to be so careful in
practice that Japanese counting is very slow. You think
I am too anxious? Not so, I have already experienced too 
many mistakes! E.g. Guo Juan reconfirms every close result
after counting by replaying the entire game in a tournament.

> No matter what you guys argue, I'm absolutely convinced that if people have
> the option of playing to the bitter end,

"the bitter end" should be called "a clear end".

> or they get the (same) result
> several dozen meaningless

If they were meaningless, then they would not change the score.

> moves before that, a vast majority of people is
> going to opt for ending the game early.

Why would you not simply allow either option - playing out
or not playing out? With the impractical Japanese rules you
do not even have this option.

> Since the title was 'Most simple Go rules' I'm going to argue here that most
> simple Go rules for humans are the Japanese rules.

For what concept of simplicity and how do you justify it then?

> - Chinese rules are cumbersome in counting.

Indeed.

> - Ing rules are much better in that respect.

Yes, but they are still in the same class of rulesets that all 
destroy the final position.

> Whatever people may think about
> the Ing rules, at least they are an honorable attempt to make the process of
> ending the game and counting as simple as possible.

"honourable attempt" is justified, even though the result has
been a failure.

> - The Tromp/Taylor rules will have similar counting problems as the Chinese
> ones.

Inhowfar?! Tromp-Taylor do not even specify any counting method!
Any problems would come from assigning a bad rather than a good
counting method to them.

> And it forces to make many

On average 30 is a reasonable guess. Calling this "many" is
an exaggeration.

> more moves than usually necessary.

It depends on what you count. If you count stones put on the
board _before_ scoring, then you are right. If you count the
game tree necessary to prove stones' status, the you are the
most wrong since a hypothetical game tree consists of many
lines of playing out instead of only one line.

***

If you are really so bothered about numbers of stones put
on the board, then why don't you count prisoners twice since
you have to put them on the board twice?

> - The Japanese rules work very well in real life play,

Apparently, this depends on the players and what they want
to perceive.

> as we all have experienced.

I have not, see above.

> So now you may understand better what I'm trying to accomplish, which is
> adding a simple protocol

I do appreciate efforts of finding protocols that properly 
deal with all rulesets.

> allowing for a
> simple dispute resolving mechanism.

Simple dispute mechanisms are deceiptive. You might be able
to check illegal moves under logical rules, however, you will
have greater problems with hardware failures, unsportsmanlike
behaviour of the players or of kibitzes, and a general ethical
system of disputes that reasonably combines rules of play,
tournament rules, and their combination.

> The whole argument about ko is just academic.

If you compare superko with basic ko plus a long cycle rule,
then you are more or less right, however, the practical
consequences of the pass-for-ko-rule (or a substituting set
of precedental rules) are practical problems that are not
just academic.

--
robert jasiek