Recent rules debate on Baduk TV

[nagano]

There's a 125-post thread on suicide, in which you have posted so you must be aware of its existence, with plenty of arguments about why disallowing suicide is just as natural as allowing it. This really is an arbitrary decision.

I know, but with all due respect, I consider all of those arguments to be erroneous.

[Harleqin]

Robert, could you please give proper references? It would also help the Google pagerank if you gave actual links instead of "elsewhere".

[RobertJasiek]

I'm not sure I understand... A position can repeat during "normal" play in the middle of a game, no passing or "zugzwang" involved, and not lead to a cycle.

One of the problems created by you is your missing knowledge of what the difference between positional cycle and situational cycle is. You wrote "cycle", i.e., it might be either. When a position repeats, we have, by definition, a positional cycle. It might or might not be also a situational cycle. Situational cycles can (if allowed) recur. Positional cycles that are not situational cycles cannot recur by themselves but they together with more moves then building a longer situational cycle can recur. And this first let me think I could prove the following:

Code: Select all

You seem to assume
a) a maximal number of successive passes ending the game,
b) passes exist,
c) a next turn whenever the game end is not reached yet,
d) no ko rules are applied,
e) 

Proposition (under these assumptions):
Positional repetition allows an infinite game.

FAKE Proof:

Only sequences not ending the game are considered for the sake of possibly allowing an infinite game.

Let P be the initial sequence to a positional repetition.

Case 1: |P| is even: Repeat P infinitely. QED.

Case 2. |P| is odd: Repeat (P followed by a single pass) infinitely. QED.

...until I noticed that such a proposition requires another assumption (e), namely not to be one of the examples I gave for a position 1 move short of the game end. My fake proposition and fake proof also show why all the assumptions are needed, to answer some of your other questions. IOW, be more careful with exactly what you claim...! Not everything is as trivial as you hope!

Passing, game end conditions, ko, and suicide are irrelevant to this fact. Do you disagree?

I disagree, see above.

In fact, if suicide is allowed, you get the most simple (excluding passes) example of this: the suicide of one stone. It repeats the board position, so it would be prohibited by positional superko, but there's no reason why it would lead to non-termination, so there's no reason for it to be prohibited.

Here we see another carelessness of you: You did not consider your own made assumptions carefully! You presumed a repeated position, i.e., your assumptions did not allow for ko rules that would have prohibited repeated positions. In particular PSK is disallowed under your implied assumptions. Therefore single-stone suicide does not behave as you wish ALA we use your implied assumptions from your problem request message.

[RobertJasiek]

Robert, could you please give proper references?

No, sorry, I lack time to do a proper search myself. Since I have explained it a dozen of times before, it is fair enough that the interested reader invests the necessary time for searching.

The places to look for are:
- rec.games.go archives
- Sensei's
- godiscussions
- my webpages
- www.dgob.de (German)

The greatest advantages of PSK are (stating these again takes only little time for me; for the other advantages I would need to do searches myself):
- known by by far the most players (note: "players", not "rules experts") as "the" superko rule
- the space of a state can be perceived visually on only the board
- tactical reading is easier than for superko variants that also consider passes (take some difficult kos like moonshine life with extra liberties and compare the difficulties of reading yourself!)

[RobertJasiek]

I consider all of those arguments to be erroneous.

Start afresh. Define what a play is! Write it down and protocol all the design alternatives you have! Forget about rules - just be more precise and create that DEFINITION of a play. To say it again: The task is just to define play - not to consider external aims beyond this elementary task. So do your homework and report what you will have noticed!

[RobertJasiek]

I don't think its a question of opinion, if simplicity is the goal.

Before we can talk about simplicity, you first need to tell us WHICH simplicity you mean!
- simple text
- the least legal moves
- the dullest strategy
- the smallest number of principles used DURING rules DESIGN
- etc.

Then you need to tell us which other goals you also use and whether some have a greater priority!

One "rule" is a natural consequence of capture.

WTH don't you apply your goal of simplicity and OMIT the concept of capture entirely? It is superfluous! Simply define what a play is! (See Tromp-Taylor for some motivation.)

The other is an artificial construction.

Artificial is your usage of "capture", to start with.

It only makes sense to choose the simpler option

Therefore, to please your own goal, OMIT capture!

and not doing so is against the basic reasoning of Go in the first place.

What is "the basic reasoning of Go"? Write it down! (It can be done without referring to capture.)

I consider allowing suicide to be a component of basic essential rules

Why?

Actually I had not heard of [the Fixed-Ko-Rule] before, though I had thought of that option. But I think it has inherent problems as well, namely that it is arbitrary which side is considered the disturber.

Neither side is. The rule works without "disturber". Such a term is for interpretation only.

Then how is it resolved in Chinese rules?

The rules dispute that this topic was originally about.

For this, I would need to know the current Chinese rules. In the 1988 and presumably the 2002 Rules, the players may resume if they want to - until the game end is reached. In the dispute, the game end was reached and neiher player resumed. Therefore it becomes a matter of tournament rules, and I do not know the currently valid Chinese tournament rules, either.

[Harleqin]

Robert, could you please give proper references?

No, sorry, I lack time to do a proper search myself. Since I have explained it a dozen of times before, it is fair enough that the interested reader invests the necessary time for searching.

The places to look for are:
- rec.games.go archives
- Sensei's
- godiscussions
- my webpages
- http://www.dgob.de (German)

Well, no. There has been written too much to keep up with it, let alone find the latest version of whatever you are referring to. It is much easier for the author to look up the location of what he is referring to than for the reader to search all possible places. That is why in scientific articles, you always have a list of references. "Has been written elsewhere" without a proper reference is unscientific. I can only recommend to keep a list of such references on the topics you are interested in; I would have expected this from an expert in these matters.

Besides, if you always added links to works you cite, that would also drive their Google pagerank up, so that a Google search actually would turn up the relevant results.

The greatest advantages of PSK are (stating these again takes only little time for me; for the other advantages I would need to do searches myself):
- known by by far the most players (note: "players", not "rules experts") as "the" superko rule

I find this disputable. I think that players who do not know that more than one superko variant exists do not know whether the player to move is part of the restriction either.

- the space of a state can be perceived visually on only the board

I do not know what the "space of a state" is, but if you mean what I think you mean, then yes, this might be an advantage. However, situational superko needs to compare a given position only with half of the previous positions, so the perception argument is not completely one-sided.

- tactical reading is easier than for superko variants that also consider passes (take some difficult kos like moonshine life with extra liberties and compare the difficulties of reading yourself!)

I would be careful with claiming properties like "easier" when comparing with undefined sets of other variants. Passes are mistreated by many rule sets. For a sensible comparison, it is necessary to have a sensible "surrounding" rule set first. If there are multiple of such sensible rule set frameworks, a comparison might turn up different results depending on which you choose.

I do not see any difficulty in moonshine kos with extra liberties, but that might be due to the rule framework in my head that I usually assume.

[nagano]

First of all let me say that I realize this is a controversial issue and I hope I haven't offended anyone.

I don't think its a question of opinion, if simplicity is the goal.

Before we can talk about simplicity, you first need to tell us WHICH simplicity you mean!
- simple text
- the least legal moves
- the dullest strategy
- the smallest number of principles used DURING rules DESIGN
- etc.


Then you need to tell us which other goals you also use and whether some have a greater priority!

The smallest number of basic concepts required for the game to function properly.

One "rule" is a natural consequence of capture.

WTH don't you apply your goal of simplicity and OMIT the concept of capture entirely? It is superfluous! Simply define what a play is! (See Tromp-Taylor for some motivation.)

I sense some sarcasm here, but I actually do not care whether capture is defined as an individual principle or part of other rules, so long as the rules are clear. Capture has to exist as a concept or the entire game breaks down. Giving capture its own rule or not does not change that.

The other is an artificial construction.

Artificial is your usage of "capture", to start with.

See above. I feel like you're getting into semantics here.

I consider allowing suicide to be a component of basic essential rules

Why?

Because it is a natural consequence of the existence of capture, unless an arbitrary rule is developed to change it.

Actually I had not heard of [the Fixed-Ko-Rule] before, though I had thought of that option. But I think it has inherent problems as well, namely that it is arbitrary which side is considered the disturber.

Neither side is. The rule works without "disturber". Such a term is for interpretation only.

Then can you clarify how it would work in practice?

[RobertJasiek]

It is much easier for the author to look up the location of what he is referring to than for the reader to search all possible places.

Quite likely but my heap of not archived rules articles is roughly 30,000. The time for me to search everything somebody wants to see immediately is too high. You need to wait for Go Rules Encyclopedia to be written. For it, I will evaluate all my archives...

"Has been written elsewhere" without a proper reference is unscientific.

Of course. But this thread is not scientific-only but mostly casual discussion.

I can only recommend to keep a list of such references on the topics you are interested in; I would have expected this from an expert in these matters.

I keep as many references and articles and rulesets as I can manage within my available time, for everything beyond that see above.

Besides, if you always added links to works you cite, that would also drive their Google pagerank up, so that a Google search actually would turn up the relevant results.

I am not the slave of google politics.

I think that players who do not know that more than one superko variant exists do not know whether the player to move is part of the restriction either.

I have talked with some hundred players on the matter offline or online since 1995: Most understand PSK when they hear "superko". The exceptions are mostly those players having already read only some specific SSK / NSK rule.

I do not know what the "space of a state" is,

The storage container for the information:
PSK: (intersection_1_color|..|intersection_361_color)
NSK: (intersection_1_color|..|intersection_361_color|created_by_play_of_player_of_color)
SSK: (intersection_1_color|..|intersection_361_color|created_by_move_of_player_of_color)

PSK has a visually perceivable 361-tuple. NSK/SSK need a 362-tuple, of which the last cell is not a visible information (values can be NEITHER, BLACK, WHITE, BLACK_AND_WHITE).

situational superko needs to compare a given position only with half of the previous positions, so the perception argument is not completely one-sided.

Players do not think like that (programs might). Players think "Have we just had a sequence of ko moves? If yes, then I need to compare their intersections only.".

I would be careful with claiming properties like "easier" when comparing with undefined sets of other variants.

The common superko variants are defined.

I do not see any difficulty in moonshine kos with extra liberties, but that might be due to the rule framework in my head that I usually assume.

How long did you need to verify that none of the possible single passes (like as the first move) alter strategy? Less than 0 seconds? If so, then SSK might be easier for you than PSK;)

[RobertJasiek]

Capture has to exist as a concept or the entire game breaks down.

Wrong. The essential concepts are "state / colour" assigned to each intersection and their posssible changes from turn to turn!

I feel like you're getting into semantics here.

Exactly. To minimize the number of concepts to please your definition of simple, semantics is essential.

Because it is a natural consequence of the existence of capture, unless an arbitrary rule is developed to change it.

The existence of capture violates your simplicity on the rules level. The existence of capture is not natural but only a practically implied concept from the necessary more fundmental concepts.

What we rather need as a fundamental concept is: After a play, each "string" is adjacent to at least one empty intersection.

A concept "libertyless strings are emptied" is not fundamental but a concept-level above because the concept does not work alone for itself but needs the quality of order: either "first current player('s strings are removed)" or "first opposing player".

It is a fundamental concept though that "first opposing 'strings' are emptied". This concept is chosen arbitrarily though, until we demand to model traditional Go. (Similarly: "Black starts the game.")

The next fundamental concept is the arbitrary but necessary decision for what to do with one's own still 'libertyless' strings.

Then can you clarify how it would work in practice?

Apply it!:) You ask for strategy though?:) Each basic ko is an eye. Long cycles are useless.

[nagano]

Clarification: every word I write here is not intended to be a part of a final rules text.

Capture has to exist as a concept or the entire game breaks down.

Wrong. The essential concepts are "state / colour" assigned to each intersection and their posssible changes from turn to turn!

I cannot agree with this. It seems you are confusing my usage of "concepts" with "rules" Concepts are the required ideas for the game to work on a functional level. Rules are simply the codification of those concepts into proper language. In order for you to develop the "rule" you must first have invented the "concept". The concept exists because it is impossible to make territory otherwise.

I feel like you're getting into semantics here.

Exactly. To minimize the number of concepts to please your definition of simple, semantics is essential.

Only if you're talking about rules in the terms of final language.

Because it is a natural consequence of the existence of capture, unless an arbitrary rule is developed to change it.

The existence of capture violates your simplicity on the rules level. The existence of capture is not natural but only a practically implied concept from the necessary more fundmental concepts.

Only if you equate "concepts" with "rules".

Then can you clarify how it would work in practice?

Apply it!:) You ask for strategy though?:) Each basic ko is an eye. Long cycles are useless.

I was asking for a clarification of the initial rule, not strategy. Your definition is ambiguous to me, but maybe you have defined some of the involved terms elsewhere that I do not know about.

[RobertJasiek]

It seems you are confusing my usage of "concepts" with "rules" Concepts are the required ideas for the game to work on a functional level.

I do not confuse concepts and rules but your imagination of concepts is not one related to your sense of simplicity but related to a different sense of it: "Fewest concepts on the informal talk level about what Go as a game shall be." This is about two conceptual levels above those fundamentals concepts I have been speaking of. I.e., while fundamental concepts approach rules from below, your informal concepts approach rules from above. (That it is informal you see, e.g., from the possible variation of speaking alternatively of removals instead of captures.)

Rules are simply the codification of those concepts into proper language.

Rules are more than that.

***

BTW, Japanese / Korean rules writers think like you but include also yet higher level strategic concepts (like "life" or "eye") in the prerequisites for what they want to be reflected in the rules.

In order for you to develop the "rule" you must first have invented the "concept".

For the sake of simplifying discussion, ok.

However, see above for different levels of concepts.

The concept exists because it is impossible to make territory otherwise.

Territory is also an informal concept (or in case of J / KOR rules: strategic concept). As a fundamental concept, we can consider empty 'strings' aka 'maximal' connected set of intersections each with the property empty.

Only if you equate "concepts" with "rules".

No, see above for levels of concepts.

I was asking for a clarification of the initial rule

Let A be the position before a play. Let B be the position after that play. The play is defined by the pair (A|B). Fixed-ko-rule: In a game, each particular (A|B) may occur at most once during a play.

http://senseis.xmp.net/?FixedKoRule

[nagano]

It seems you are confusing my usage of "concepts" with "rules" Concepts are the required ideas for the game to work on a functional level.

I do not confuse concepts and rules but your imagination of concepts is not one related to your sense of simplicity but related to a different sense of it: "Fewest concepts on the informal talk level about what Go as a game shall be." This is about two conceptual levels above those fundamentals concepts I have been speaking of. I.e., while fundamental concepts approach rules from below, your informal concepts approach rules from above. (That it is informal you see, e.g., from the possible variation of speaking alternatively of removals instead of captures.)

Tromp-Taylor rules allow suicide, and never mention capture. I actually have nothing against the Tromp-Taylor rules as they are written. They do not decide to have the rule or not it is just one of the consequences of the underlying rule. This much I agree with, and it actually confirms my argument.

Rules are simply the codification of those concepts into proper language.

Rules are more than that.

What are they, then?

BTW, Japanese / Korean rules writers think like you but include also yet higher level strategic concepts (like "life" or "eye") in the prerequisites for what they want to be reflected in the rules.

Um, no, they don't. I would never develop arbitrary rules about life or eyes, which are conditional and depend on the whole board position. For the last time: capture is required. It does not matter what you call it. Yes, I understand there are mathematical considerations and Tromp-Taylor rules do address those, and I actually think that's probably the best way to deal with it from a rules perspective. I am not arguing for or against having a separate capture rule. I believe your "removal" and my "capture" are actually the same thing.

The concept exists because it is impossible to make territory otherwise.

Territory is also an informal concept (or in case of J / KOR rules: strategic concept). As a fundamental concept, we can consider empty 'strings' aka 'maximal' connected set of intersections each with the property empty.

Again, I am not arguing against a mathematically constructed ruleset. Yes, I was using the word territory informally. After all, I prefer area rules. Perhaps I should have said: "Whatever you call the method of scoring you are using, removal must be possible if the game is ever to be scored."

I was asking for a clarification of the initial rule

Let A be the position before a play. Let B be the position after that play. The play is defined by the pair (A|B). Fixed-ko-rule: In a game, each particular (A|B) may occur at most once during a play.

http://senseis.xmp.net/?FixedKoRule

Ok, then. So the rule does mean what I originally thought it did. In that case, I still disagree. Although it does not actually have the word "disturber", the concept is there. The basic effect of the rule is that the attacking side (black in the linked example) always wins, and that white is the disurber. The alternative would be that the position is allowed to repeat once, allowing white to defend. This rule would imply, stated or not, that black is the disturber. So a choice of one or the other is arbitrary, and from a subjective perspective, I think a decision in favor of either is unfair in most cases.

[palapiku]

Here we see another carelessness of you: You did not consider your own made assumptions carefully! You presumed a repeated position, i.e., your assumptions did not allow for ko rules that would have prohibited repeated positions. In particular PSK is disallowed under your implied assumptions. Therefore single-stone suicide does not behave as you wish ALA we use your implied assumptions from your problem request message.

Sorry, I still don't understand, and I think you don't understand me either.

First of all I'm not interested in cycles, either positional or situational. I'm only interested in non-terminating games. Theoretically, cycles of any kind are harmless as long as the game ends. It's non-termination that superko is meant to address. Do you agree?

The suicide of one stone is obviously prohibited by PSK. That was my point. The problem with PSK is precisely that it prohibits situations which don't lead to non-terminating games, such as the suicide of one stone. One could argue about whether the suicide of one stone is a good or bad idea to allow, etc, but it doesn't lead to non-termination and therefore should not fall under the jurisdiction of a superko rule. (Unless both players do it, which could lead to non-termination and is correctly disallowed by SSK.)

[RobertJasiek]

Rules are simply the codification of those concepts into proper language.

Rules are more than that.

What are they, then?

Rules are also a rough approximation of what - on the fundamental level - the most formally becomes axioms and definitions.

Um, no, they don't. I would never develop arbitrary rules about life or eyes, which are conditional and depend on the whole board position. For the last time: capture is required. It does not matter what you call it. Yes, I understand there are mathematical considerations and Tromp-Taylor rules do address those, and I actually think that's probably the best way to deal with it from a rules perspective. I am not arguing for or against having a separate capture rule. I believe your "removal" and my "capture" are actually the same thing. [...]

I lack time to continue this discussion now.