So, I'm at a little bit of a loss as to what to do. I guess I'll just make some decisions arbitrarily and accept any remaining mismatches as mismatches.

In the old days, before written rules, if a rules problem arose, you asked the best player you could to adjudicate. These days, the people who write the rules and the best players are rarely the same, although the rules writers are usually good players. Ing was an amateur. The J89 rules were not intended to match actual practice, and allowed a loophole for filling dame informally. That caused problems down the line. The AGA rules did not match actual practice, and actual practice differed, but the referees enforced the rules as written, for the most part. I doubt if the Chinese rules were intended to match actual practice, and it looks like they still don't. Not a problem if everyone agrees. The same thing happens with real life laws and regulations. Many are broken and the authorities wink or turn a blind eye.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

As far as the Chinese superko rule is concerned, look at it this way. You are the referee and your word is law.

Similarly with Japanese rules ambiguities. You are the judge, and you interpret the rules.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

If the J89 rules were intended to match actual practice, they would have left the 1949 rules in place. But the 1949 rules were criticized for being illogical and ad hoc. (Ad hoc, to the thing itself, as in matching actual practice. ) They ended up with rules that were legalistic, ambiguous, difficult to understand and apply, while the top players continued doing as they always had. Damn the rules, full speed ahead.

IMHO, rules are made for the players, not the other way around. Before go had written rules, everybody knew how to play, now we have written rules, and nobody does.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Why do you think that you must use official Japanese rules? Why not something simpler?
http://home.snafu.de/jasiek/sj.html

From https://lightvector.github.io/KataGo/rules.html

What kind of weirdness do you mean? Is it worse worse than in my example below?

Click Here To Show Diagram Code

[go]$$cB
$$ ----------------------
$$ | X X X X X X X O . . .
$$ | X O X . X . X O O . .
$$ | O . O X O X O X O . .
$$ | . O O O O O O X O . .
$$ | O O X X X X X X . . .
$$ | X X X . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]

Sorry to post so many times on the Chinese superko rule, but the internet is a wonderful thing. From the English version of the 2002 rules, approved by the Chinese Weiqi Association, section 20.


(Emphasis mine.)

The last clause is what gives the tournament organizers or the TD the right to rule differently from the text.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

https://senseis.xmp.net/?MolassesKo - in the position linked here, it seems to me that you want a player to be able to capture both stages without having to unblock ko (i.e. pass-for-ko) in between. Or am I wrong? Basically, the side that has to re-capture earlier captures in the multi-stage ko cannot keep up, so based which player ends up going first, one side or the other will win the fight unconditionally, rather than it being a seki or a no-result.

Additionally, your very own example indicates some difficulty if ko-blocking persists forever and is never cleared when a ko dissolves, at least as it is currently formulated. Maybe a different formulation of the "location associated to a ko-block" would work though.

Click Here To Show Diagram Code

[go]$$c
$$ ----------------------
$$ | . X O . O O X . . .
$$ | X O . O O X X . . .
$$ | . X O O X X . . . .
$$ | X O O X X . . . . .
$$ | O O X X . . . . . .
$$ | X X X X . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .
$$[/go]

Let me ask again why you are so obsessed with ko-pass-like rules. I have not checked the rare shapes in this thread but previously only two relatively rare shapes have been known for which ko-pass rules differ from the basic ko rule. (Infinite long cycles are allowed for detecting "no permanent stone".) IIRC, they are:

- perpetual ko + basic teire ko
- triple ko with external ko

For Japanese professionals, both are unacceptable. For computers, they are immaterial.

For AlphaZero-like self-play training purposes, neural net input/output representation, and compatibility with existing protocols for computer Go playing and game representation (GTP, SGF), I would like a ruleset that:

* Does not include any hypothetical play, branching of the game (i.e. copying the game to determine something) or any rollback of state upon determining the status of a chain, or a ko, or an area of the board. The game and all statuses are determined through alternating actions alone.

* All possible actions in any phase of the game are either a pass or are "naturally" representable as being associated with one or more locations on the board, such that these locations are disjoint between actions on a given turn (implications: any location has at most one possible action associated with it, the maximum number of possible legal actions on a 19x19 board is 362, there are no actions like "communicate a list of group and their proposed statues to be agreed upon" that cannot be naturally be encoded in such a way, etc.).

* Will still produce some result even with fairly badly-behaved players. Such as players that play completely at random, or do not understand when play are necessary or not, or that "mistakenly" take actions that result in the game ending "prematurely". Self-play rules must be able to handle such players.

* Can be efficiently implemented in an actual computer program without too much difficulty. I think the current draft is not too bad on this - only a small number of additional concepts are defined such "state" and "atari" beyond those that a computer implementation would need to implement anyways such as "region", and simple combinations of those concepts are enough to define the most complex concepts like "independent-life-region" without any more layers of definitions. And as for efficiency, everything is computable directly, for example no tree search over move sequences is necessary to implement any of the rules.

* Subject to the above technical requirements/restrictions, still does a reasonable job of matching most common situation results in the Japanese rules so long as the players ARE reasonably-well-behaved and act self-interestedly.

For most area-scoring rulesets, it is not hard to come up with a version that satisfies all the above technical requirements, in fact many of them pretty much do already - at least after omitting any human provisions for group status agreement. The challenge is to find one for Japanese rules. I'm not aware of mechanisms right now that do a better job of achieving such an approximation subject to such requirements than ko-pass-like rules. Are there any known?

Edit: Maybe not all of the above restrictions are necessary, but having them massively simplifies the process of making self-play training work and to make the bot compatible with existing tools, GUIs, etc.

I like to think that the current draft, despite the flaws, actually does a pretty good job of achieving the above objectives. If someone does see a simple change that would be a clear improvement, that would be great. Otherwise, I'm likely go just go ahead with it (after pondering for a little longer if any simple clear improvements are possible), unless someone comes up with something massively better.

I hope some of this work ends up being actually useful in the end. In the current state, it's pretty clear to me why no Go programmer wants to touch this stuff if they can help it.

Remains to be proved.



Sry. Better (i.e. more computer-friendly) at such an approximation: do not use ko-pass-like rules but use the basic ko rule (in its variant that allows recapture after regular pass).

This may easily be seen insufficient because of the moonshine problems.

I'm not sure whether these rules also aim at human usage besides bot usage, but in both cases they do look a bit too complicated. There are only 2-3 real problems in go rules design, all the rest is a consequence of poorly designed earlier rules that were created to solve something - kind of a snowball effect. Adding new rules is always dangerous and best avoided as far as possible.

For example, why do you need "second pass from the same wholeboard position"? You already have threefold repetition, wouldn't it be better to let that handle asymmetric cases as well?

Separate cleanup phases with temperature switch seems like a natural thing for bots (playing like this is not much different to TT cleanup, and can also be omitted in human play in case of agreement). But the ko pass rules again seem like an overkill. Once you had the first phase with normal ko rule, correct handling of real repetitions is ensured, and you should have a free hand to handle the rest. Even something very crude like switching to superko after the first phase may be viable (for bots at least).

You are wrong. It seems that unblocking each stage separately works better.

One possibility to solve this problem would be to define ko as a pair: the stone to be captured and the point where the new stone is to be played. A ko-block is assigned to this pair. In the position above the ko captured from different direction would be another ko without a ko block, and blck stones can be killed.

Anyway when creating new rules unexpected rules may occur. Often they will be discovered years afterwards, like it happened with Japanese 1989 rules.

Yes, rules beget rules. Matti has made the same point, and Ikeda did many years ago, warning against chasing anomalies. John Tromp made a set of elegant rules and said let the chips fall where they may. Go Seigen, while not a rules designer, also believed in accepting the results of play. Ing, OTOH, chased anomalies his whole life. I, a reluctant rules designer, adopt a principle or principles, and if the rules embody it or them, let the chips fall where they may. One principle is that a position has a score if and only if it has a temperature of -1 by territory scoring. This is a condition that you can test. Unlike questions of life and death, for instance.



That is easily explained from the principle that a final position has a score. If one player passes, modern rules typically take that as meaning that she has no way of improving the score for herself. If then the opponent initiates a sequence of play that returns to the same position, we may take that as meaning that the opponent cannot improve the score for himself, either. But how the position was reached may be different, such that the passer now has a play to improve her score. If not, she passes again. Which we take to mean that neither player can improve their score, so we stop play and score the game.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

For players that are really clueless you can use proximity scoring. A point belongs to the player whose stone it is closest to by manhattan distance. This allows territory in seki, OC, but players who leave nearly every group with dame can still get a score when they stop the game prematurely.



I think you are doing a great job!



My sympathies.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

I don't think this flies. Either one don't invent tricky new rules and let the chips fall where they do naturally, or it is one's responsibility to ensure the new addition works without causing problems, and describes the real game.

Yes, I understand where this came from. When you use this with superko where passes lift bans, it has an actual purpose. Without that, from a practical point having an n-fold repetition rule, a two-pass rule AND a separate pass-repetition rule still seems redundant.

A lot of people like Tromp-Taylor rules. IIUC, New Zealand uses them, and area scoring bots used to, maybe still do.





Then you can think of it as a fail-safe provision. No matter how you tinker with other rules, it is there to stop play in a reasonable spot.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Can you explain?

Yes, I'm well aware of wanting to limit complexity. Right now I'm strongly averse to adding any *more* complexity, I'm seeing if I can just make do with minor changes to the existing draft that neither make things more or less complex, just finding the right permutation of conditions.

These are aimed for bot usage, not human usage. There aren't even provisions for agreeing on dead stones or any such things. They are about as minimally complex as I could manage and still obtain the following:

* "No territory in seki"
* "No gain for filling one-sided seki dame during cleanup"
* "Bent 4 in corner, double ko death, are dead regardless of ko threats"
* "Basic moonshine life is not possible, the false eye group will actually die during cleanup play"
* "No obvious ways to infinitely prolong the game or no-resultify common positions that people do not expect to be no-resulted".
And maybe a few other basic things that I'm not thinking of right now but that I thought of when designing the ruleset.

I consider these above things pretty basic things that one should have if trying to have "Japanese-like" rules. Mis-fitting on more complex situations is tolerable and expected, but it seems worth spending some effort now to minimize the potential for that if it can be done by slight changes that do not add any further complexity.



Taking it out without adding anything else means that sending-two-returning-one is highly problematic under Area Scoring + Simple Ko - one does not want this position to be a no result. This rule is an effective way to bound that case, and also any future cases that I can't think of. One could also adopt a "count how many stones were captured of each color during this long cycle" rule, or something like that. That would be a different way to solve it. Either way, one actually does need a rule here rather than just having nothing.

They are so basic that the Japanese do not know how to do them well. Handling them is a real feather in your cap.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.