[Japanese Rules]

Newest Contents 2004-09-24 - Last Update 2008-11-04 - First Day 2004-09-13

INDEX | GO | RULES |
JAPANESE 2003 rules - COMMENTARY - INFORMAL
NEW AMATEUR-JAPANESE rules - COMMENTARY

Commentary on the Japanese 2003 Rules

General
Version
Acknowledgements
Exceptions
Legend
Application of the Rules to the Alternating-sequence
Application of the Rules to the Final-position
Reasons for some Particular Rules
Understanding the Japanese 2003 Rules
Best Model for Professional Japanese Rules
Japanese 1989 Rules
Further Research

General

The Japanese 2003 Rules

explain Japanese rules used by professionals,
are not official rules but invented by Robert Jasiek,
are logical, are complete, and have well defined terms.

The Japanese 2003 Rules are the result of a decade of experience as a Go rules researcher and of months of full time specific work on these rules. This is an indication of how difficult Japanese rules are.

Countless rulesets tried to explain Japanese style rules logically but all failed to explain their spirit completely. The Japanese 2003 Rules are the first ruleset that succeeds. To be more precise and honest, since the complexity of the spirit of Japanese style rules is literally infinite, one can never be 100% sure that a logical ruleset always behaves as that part of Japanese rules tradition expects that has not suffered from fashionable changes.

To approach 100% as closely as humanly possible, the author of the Japanese 2003 Rules has studied application to many examples, commented them thoroughly and extensively, and studied theoretical background. Since all this requires about a year of work, it will be made available commercially as a book or maybe as a PDF file.

The Japanese 2003 Rules are as logical and complete as rules can be that use natural language and not mathematical notation. The author intends to create also the latter as soon as possible. - All terms in the rules text are well defined because they are specified precisely and derived from other defined terms or from the axioms.

The Japanese 2003 Rules are core rules of play and do not include tournament rules, which may be treated in a tournament ruleset. E.g., tournament rules might specify how the score shall be counted or that a no-result rule shall override the result tie of the long cycle rules.

Version

The current version of the rules text is number 35a. Until also a mathematical version of the rules will be available, the version number counts since the start of development. The current version should be considered almost final.

The title of the rules contains the year 2003 because mainly they were developed between the beginning of July 2003 and the end of March 2004.

Acknowledgements

Bernd Gramlich: He has criticized the rules by making other suggestions and proof-read some drafts.
Bill Spight: He provided an original idea for a ko-pass rule that could be abused in a very useful, new manner.
Makrai Joszef: He provided a counter-example for an early draft of ko-pass restrictions during a hypothetical-sequence.
Robert Pauli: He has proof-read several drafts of the rule text. He provided a few crucial counter-examples for some of the early drafts. He has made many suggestions for improving clarity of the semantics of the text's wording. He has discovered a first kind of the concept "controls", which is used in the New Amateur-Japanese Rules.

Exceptions

Japanese rules are difficult to understand for various reasons. One of them is exceptions. It is only a slight exaggeration to say that professional Japanese rules consist of exceptions. Japanese rules are not as elegant as Territory Scoring rules could be but as ugly as tradition and careless professionals have made them.

Understanding the Japanese 2003 Rules becomes easier if one recognizes all the exceptions and distinguishes them from the rules that are necessary for such Territory Scoring rules that depend on Life and Death. Therefore the exceptions are listed even before a basic understanding of the rules is provided.

Ko Rules during the Alternating-sequence

The rule text about ko rules during the alternating-sequence is:

"The alternating-sequence is a sequence of moves that

[...]
{basic-ko} does not include a play that recreates the position just before the last opposing move,
{long-cycle-tie / long-cycle-win} exceptionally and immediately after a play lets the game end with the result tie / win of Black / win of White if the position just after the play is the same as a position just before an earlier move and if the number of white-stones minus the number of black-stones that have been removed during plays since the first occurrence of that position is equal to / greater than / smaller than zero, and
{direct-ko} does not include a pass that succeeds a pass if then a final-string were capturable-3.

[...]

The result is

[...]
given due to the rules about long-cycle-tie / long-cycle-win."

A simplification of the rule text could be:

"The alternating-sequence is a sequence of moves that

[...]
{superko} does not include a play that recreates a position that existed before an earlier move."

The superko rule is a general rule that works for recreation of a position after any number of intervening moves. The basic-ko rule is a specialized rule that applies to recreation of a position only after exactly two moves. Hence the basic-ko rule is only a special case of the superko rule. As a consequence, the basic-ko rule does not treat all other special cases. Those require additional, exceptional ko rules (the long-cycle-tie rule and the long-cycle-win rule) and exceptional result conditions.

Japanese rules use several ko rules instead of just one ko rule because of tradition. Strategically, the differences between the superko rule and the Japanese ko rules during the alternating-sequence are tiny. They occur only about once every 5,000th to 20,000th amateur game or 500th to 1,000th professional game. The exceptional rules have to be obeyed in each game while an exception becomes relevant only rarely.

The direct-ko rule is also included because of and only because of tradition. In scarce games it changes the score by about 1 point.

Definitions for Hypothetical-sequence

The rule text about definitions for hypothetical-sequence is:

"A ko consists of two adjacent intersections so that a play on one of them succeeded by a play on the other one would recreate the position.
A ko-stone is a string on one intersection of a ko.
A ko-capture is a play that is in a ko and removes exactly one stone.
A ko-pass by a player is announced as "ko-pass" if a ko-capture by the player in some ko is currently available but prohibited.
A hypothetical-move is either a play, a ko-pass, or a pass."

A simplification of the rule text could be:

[no text at all]

The definitions are used only for the exceptional ko rules during a hypothetical-sequence or a beginning of it (which is called left-part).

If during a hypothetical-sequence superko were used, then the additional definitions would be superfluous.

The definition of ko is not used in the basic-ko rule because it is written much shorter without defining the shape of a basic ko. However, the definition of ko is useful to make the exceptional ko rules during a hypothetical-sequence readable.The same can be said about the definitions of ko-stone and ko-capture.

Ko-pass is a third move type. It is required for the exceptional ko rules during a hypothetical-sequence. As a consequence, a hypothetical-sequence does not consist of moves but of hypothetical-moves. (Their nature is hypothetical in both cases.)

Why is ko-pass required as a third move type? The exceptional ko rules during a hypothetical-sequence use it. Why do they need to use it? Each of them (in the definitions of "ko-pass" and of "hypothetical-sequence", respectively) uses it, so the question becomes: Why must a hypothetical-sequence follow different ko rules than those of the alternating-sequence? The answer is: tradition. Tradition has created special desires to let particular kinds of shapes have particular outcomes. Without that tradition, a hypothetical-sequence could use the superko rule quite like the alternating-sequence could use it.

Ko Rules during a Hypothetical-sequence

Besides the aforementioned direct-ko rule, which refers also to hypothetical-sequence, and besides the ko rule integrated in the definition of "ko-pass", the rule text about ko rules during a hypothetical-sequence is:

"For a given player, a hypothetical-sequence is an imagined sequence of hypothetical-moves that

[...]
{hypothetical-ko:} between a player's ko-capture on one intersection of a ko and the next ko-pass by him or his opponent does not include an opponent's ko-capture on the other intersection."

A simplification of the rule text could be:

"{superko} [...] During a hypothetical-sequence, the positions of the alternating-sequence may also not be recreated."

The exceptional ko rules are used during a hypothetical-sequence or a beginning of it.

During a hypothetical-sequence the ko rules could consist of just one rule: the superko rule. Since the superko rule could also be used during the alternating-sequence, all that is needed for hypothetical-sequences is a clarification that the positions of the alternating-sequence may also not be recreated during a hypothetical-sequence.

Again tradition requires not to use the superko rule, independently of whether it would be useful.

Due to some traditional positions, it is not possible to use the same ko rules as during the alternating-sequence.

During a hypothetical-sequence, would it be possible to use only the basic-ko rule and moves but not ko-passes? In most practical positions tradition would like the outcomes. However, tradition by Japanese professionals rejects the outcomes of some rare positions, which they consider important regardless of rarity.

Since tradition rejects superko dogmatically, the hypothetical ko rules become complicated and ko-pass as a third move type becomes necessary. For tradition, it is immaterial that the superko principle was mentioned in both the Japanese 1949 Rules and the Japanese 1979 Rules (the World Amateur Go Championship Rules).

Infinite Hypothetical-sequences

The rule text about the end of infinite hypothetical-sequences is:

"For a given player, a hypothetical-sequence is an imagined sequence of hypothetical-moves that

[...]
either has a finite number of hypothetical-moves and ends with the pass succeeding a pass or has an infinite number of hypothetical-moves and does not have a pass succeeding a pass,
[...]"

A simplification of the rule text could be:

"For a given player, a hypothetical-sequence is an imagined sequence of hypothetical-moves that

[...]
ends with the pass succeeding a pass,
[...]"

Without the simplification, infinite hypothetical-sequence need be considered in practice, in the semantics of all rules terms that refer to hypothetical-sequence, and explicitly in the definition of hypothetical-sequence.

Infinite hypothetical-sequence are a further consequence of the tradition to reject superko.

The simplification could be simplified further. The alternating-sequence and a hypothetical-sequence could use a common term "move-sequence". I.e. only once it would be necessey to specify how a move-sequence ends.

Strategy versus Hypothetical-strategy

During the alternating-sequence the objective of strategy for Black / White is to maximize / minimize the score of the final-position while during the hypothetical-analysis the objectives of hypothetical-strategy are to prove that each particular black-string or white-string is either alive or dead.

Therefore it is not valid if amateurs consider hypothetical-strategy as if it were strategy. Reading ahead may be a similar excercise and more often than not go skill might turn out to be helpful but otherwise strategy during the alternating-sequence and hypothetical-strategy during the hypothetical-analysis differ.

The different objectives are of so fundamental nature that every ruleset of Territory Scoring with life and death classification uses it. For this reason, one may excuse usage of exceptional objectives during the hypothetical-analysis.

Details and examples are beyond the scope of this page. However, one should note that Area Scoring rules do not need any of the following terms:

final-position
final-string
hypothetical-sequence
left-part
hypothetical-strategy
compatible
force

For (traditional) Japanese style rules, the contents of all these terms is essential. For Go rules in general, all these terms are exceptional.

Life and Death

Besides the various ko rules and the definitions related to hypothetical-strategy, the rule text about life and death needs the following terms:

uncapturable
permanent-stone
local-1
capturable-1
local-2
capturable-2
local-3
capturable-3
alive
dead

A simplification of the rule text would need only the following two terms of the New Amateur-Japanese Rules:

immortal
controls

As can be seen, filling dame during the alternating-sequence and doing without traditional life and death statuses in a few rare positions allows a great simplification of the rule text.

The term permanent-stone is defined not to introduce exceptions itself but help definitions of other terms.

Under the Japanese 2003 Rules, a basic exception of life and death is that during the alternating-sequence it depends on global interaction of all stones on the board while during the hypothetical-analysis it partially depends on interaction of subsets of all stones on the board. This very fundamental difference, however, is beyond the scope of this page. One might already be aware of this from the names of the local-x terms.

Score Difference Zero in Asymmetrical Sekis

Excluding so called sekis from the scores is a general exception of Japanese and Korean rules. It is not even used by all Territory Scoring rulesets.

In practice, sekis are not seen in every game and maybe about 95% of the sekis are symmetrical with respect to their internal territory difference. Only in a pretty scarce asymmetrical seki the internal territory difference matters at all.

Tradition of Japanese rules requires that even in asymmetrical sekis the internal territory difference is zero. The impact on the length of the rules text and the number of rules terms is considerable.

The rule text ensuring the score difference zero even in asymmetrical sekis needs the following terms

black-eye-string
white-eye-string
black-eye-point
white-eye-point
eye-point
dame
black-region
white-region
in-seki
black-territory
white-territory
score

A simplification of the rule text would need only the following term:

score

Once the hypothetical-analysis is done, there should not be any additional hypothetical-analysis. The terms specifying the players' scores must assess them statically, i.e. without further reading of hypothetical-sequences.

The simplification would be possible if life and death were replaced by the concepts immortal and controls. However, professional Japanese rules tradition does not allow this. To keep the tradition and replace all terms except the term score, which are used to ensure the score difference zero in asymmetrical sekis, the hypothetical-analysis of the Japanese 2003 Rules would have to use both all the life and death terms and the terms immortal and control. This coexistence of concepts would create problems. So all the terms ensuring the score difference zero in asymmetrical sekis remain in the rules.

Legend

General

Positive scores are favourable for Black. Negative scores are favourable for White.
For a final-position and unless stated otherwise, the prisoner-difference zero is assumed.

Diagrams

# = intersection with black stone
O = intersection with white stone
. = empty intersection
A blank space separates horizontally adjacent intersections.
A line break separates adjacent grid lines vertically.
a,b,c... = denoted intersections

`Sequences`

[] = move-sequence (or a part of it) of alternating moves of both players
#[] = Black moves first
O[] = White moves first
x = move on intersection x
p = pass
P = ko-pass (Japanese 2003 Rules)
X = ko-pass for intersection x (Japanese 1989 Rules)
t = tenuki (possibly on an imagined bigger board) or intersection t
[xy..] = obvious part omitted
[]* = repeated infinitely
[][] = move-sequence followed by another move-sequence
[]*[] = the first part is repeated several times
[..xy] = beginning of the alternating-sequence omitted; however, the obvious ".." is often omitted.
#[abcppp] = Black plays at a, then White plays at b, then Black plays at c, then White passes, then Black passes, then White passes.
O[abPcdP]* = White plays at a, then Black plays at b, then White ko-passes, then Black plays at c, then White plays at d, then Black ko-passes, then repeat.
O[abCDcdAB]* = White plays at a, then Black plays at b, then White ko-passes for the intersection c, then Black ko-passes for the intersection d, then White plays at c, then Black plays at d, then White ko-passes for the intersection a, then Black ko-passes for the intersection b, then repeat.

Application of the Rules to the Alternating-sequence

Simple parts of the Japanese 2003 Rules are not explained here at all. They and all details about difficult parts will be explained the most thoroughly in a book or PDF file.

Pass

(1)

Black to move during the alternating-sequence

. # . O .
# # # O O
. # O O .

intersections

. # a O .
# # # O O
. # O O .

#[app] is a possible sequence of moves. The second pass (a pass is indicated by the letter p) of the succession of passes ends the alternating-sequence.

final-position

. # # O .
# # # O O
. # O O .

(2)

White to move during the alternating-sequence

. # # O O .
# # . # O O
. # # O O .

intersections

. # # O O .
# # a b O O
. # # O O .

O[apbpp] is a possible sequence of moves. Since the basic-ko rule prohibits Black to play O[ab], the second move of the sequence is a pass. The second pass of the succession of two passes ends the alternating-sequence.

final-position

. # # O O .
# # O O O O
. # # O O .

Long-cycle-tie / long-cycle-win

(1)

Black to move during the alternating-sequence

# # # # #
. # O # O
# O . O .
O O O O O

intersections

# # # # #
a # b # c
a O b O c
O O O O O

The last play of the sequence of moves #[bacbac] invokes a long-cycle-tie because the cycle removes 3 black-stones and 3 white-stones, i.e. the difference of white-stones minus black-stones is zero. The result of the game is a "tie".

exceptional game end position

# # # # #
. # O # O
# O . O .
O O O O O

(2)

White to move during the alternating-sequence

. O O . # O
O # # # # O
. # O O O O
# # O . O .

intersections

a O O . # O
b # # # # O
c # O O O O
# # O . O .

The last play of the sequence of moves O[cab] invokes a long-cycle-win and the result "win of Black" because the cycle removes 1 black-stone and 2 white-stones, i.e. the difference of white-stones minus black-stones is greater than zero. (Of course, O[pp] would have been a strategically correct sequence of moves.)

exceptional game end position

. O O . # O
O # # # # O
. # O O O O
# # O . O .

(3)

Black to move during the alternating-sequence

. # # . O #
# O O O O #
. O # # # #
O O # . # .

intersections

a # # . O #
b O O O O #
c O # # # #
O O # . # .

The last play of the sequence of moves #[cab] invokes a long-cycle-win and the result "win of White" because the cycle removes 2 black-stones and 1 white-stone, i.e. the difference of white-stones minus black-stones is smaller than zero. (Of course, #[pp] would have been a strategically correct sequence of moves.)

exceptional game end position

. # # . O #
# O O O O #
. O # # # #
O O # . # .

Direct-ko

(1)

during the alternating-sequence

. O O # O . # .
O O # . # # # #
. O # # # # # .

The ko-stone is capturable-3, as is discussed later. Therefore the alternating-sequence may not continue and end with just two successive passes. This is prohibited by the direct-ko rule.

Application of the Rules to the Final-position

The real difficulty of the Japanese 2003 Rules (or any Japanese rules) is their application to the final-position. Details will be explained the most thoroughly in a book or PDF file.

Below we assume to have a final-position that must be analized.

Hypothetical

Probably your first surprise is the frequent use of the word hypothetical in rules terms. What is done with the final-position? Physically, nothing is done at all. All the determination of life and death is done - hypothetically. Move-sequences that prove life or death are not laid out on the board but they are all imagined. Analysis of the final-position is like strategic planning during the alternating-sequence: It is done only in the players' minds.

The move-sequence played as the alternating-sequence is executed physically with stones on the grid. Each move-sequence during the hypothetical-analysis is only imagined, it is hypothetical and therefore called hypothetical-sequence. - The objects that the alternating-sequence consists of are called moves while the objects that a hypothetical-sequence consists of are called hypothetical-moves. - Planning for the alternating-sequence is called strategy. Plannung for the hypothetical-analysis is called hypothetical-strategy. The rules do not need "strategy" as a term because for the rules it is sufficient to know of the alternating-sequence; only the players should use strategy to create the alternating-sequence.

Hypothetical-sequence and Hypothetical-strategy

Maybe your second surprise is the usage of the terms hypothetical-sequence and hypothetical-strategy at all. Before Japanese rules did not use them. This was their major incompleteness.

The Japanese 1949 Rules and the World Amateur Go Championship 1979 Rules tried to hide the incompleteness. Partially, they tried to explain all shapes and classify life and death due to the shapes of the stones in the final-position. They failed. They had to fail because life and death of every shape only marginally depends on its visual appearance but mainly depends on the hypothetical-sequences possible in it. Visual appearance cannot while hypothetical-sequences can be described in general. Therefore in rules that depend on life and death it does not make sense to avoid reference to hypothetical-sequences.

The Japanese 1989 Rules hide the incompleteness behind the word "cannot" in phrases like "stones that cannot be captured by the opponent". Ask your Go teacher about a definition of strategy and receive the advice "I cannot know!". Ask the Japanese 1989 Rules about a definition of hypothetical-strategy and receive the advice "I cannot know!". The Japanese 1989 Rules do not know how to define hypothetical-strategy and therefore they hide it behind grammar.

It requires methods of mathematics to define hypothetical-strategy. In other words, the authors of precise Japanese rules should have studied some mathematics or informatics at university. Therefore it is little surprise that hypothetical-strategy could not be defined by Japanese professionals but by Bernd Gramlich, Robert Pauli, and Robert Jasiek, who have had some suitable education. (Is it a coincidence that all three are Germans or is mathematical education at German universities particularly good?) The final results can be seen in the so called Logical Japanese Rules of Go or in the Japanese 2003 Rules.

Hypothetical-sequence

It should be possible to understand the term hypothetical-sequence if one does not care too much about the details of the hypothetical-ko rules. Precise application of the rules requires also a very good understanding of the hypothetical-ko rules but in practice a pretty naive understanding of ko rules will do.

A hypothetical-sequence is not just given as such but it is always given for the final-position. Why? The hypothetical-moves that are plays of the hypothetical-sequence are played first in the final-position and then in descendant positions. The final-position for that the hypothetical-sequence is given is that position in that the first play of the hypothetical-sequence is played. This play creates a new position in that then the second play of the hypothetical-sequence will be played, etc. I.e. the position for that the hypothetical-sequence is given is the starting position; the hypothetical-sequence starts from it. In practice, the starting position for a hypothetical-sequence is the final-position, which is the position at the end of the alternating-sequence and therefore also the position in that life and death are analysed.

A hypothetical-sequence does not just exist in the air but it depends on a given player. Why? This is the player that has the first hypothetical-move in the hypothetical-sequence. E.g., #[abcdpp] is a hypothetical-sequence that starts with Black, O[efghpp] is a hypothetical-sequence that starts with White.

You have expected the players to alternate, haven't you? #[abcdpp] stands for "Black plays at a, then White plays at b, then Black plays at c, then White plays at d, then Black passes, then White passes.". Each play or pass in that is a hypothetical-move.

There can be finite hypothetical-sequences like #[abcdpp]. Each finite hypothetical-sequence ends with two passes. There can be infinite hypothetical-sequences like #[abcd]*, which stands for "Black plays at a, then White plays at b, then Black plays at c, then White plays at d, then repeat.". Each infinite hypothetical-sequence does not include two successive passes because otherwise it would already end after only a finite number of hypothetical-moves.

Just what does it mean that a hypothetical-sequence is infinite? It contains some recurring cycle. E.g., imagine a triple ko with a cycle of 6 plays; this can be repeated forever. Why? Because it is legal. (We are not in the alternating-sequence. During a hypothetical-sequence the ko rules are different and they allow infinite cycles.) Why are infinite hypothetical-sequences allowed? Because Japanese style rules hate superko! Ok, now that you accept that fate, how do we interpret an infinite hypothetical-sequence? Imagine our triple ko again. Is one of the big strings next to the ko mouths of the triple ko ever captured in a cycle of length 6 that repeats infinitely? No. Even though the hypothetical-sequence is infinite, neither of the big strings is captured. So although some hypothetical-sequences can be infinitely long, it is still possible to judge whether a particular string is captured or is not captured during a particular hypothetical-sequence.

Examples for the Terms Hypothetical-sequence and Left-part

(1)

final-position

. # # .
# # O #
O # O O
. O O .

intersections

. # # c
# # O d
b # O O
a O O .

Hypothetical-sequences:
#[pp]
#[acpp]
#[acPbdP]*
#[acPpdbPp]*
#[acpPpbdpPp]*
#[acpPpbdP]*
etc.

These are not hypothetical-sequences:
#[p]
O[p]
#[ppp]
#[p]*
#[ac]
#[acpppp]
#[acPPdbppPP]*
etc.

Left-parts of the hypothetical-sequence #[pp]:
#[p]
#[pp]

Left-parts of the hypothetical-sequence #[acpp]:
#[a]
#[ac]
#[acp]
#[acpp]

Left-parts of the hypothetical-sequence #[acPbdP]*:
#[a]
#]ac]
#[acP]
#[acPb]
#[acPbd]
#[acPbdP]
#[acPbdPa]
etc.
#[acPbdP]*

Left-parts of the hypothetical-sequence #[acPpdbPp]*:
#[a]
#]ac]
#[acP]
#[acPp]
#[acPpd]
#[acPpdb]
#[acPpdbP]
#[acPpdbPp]
#[acPpdbPpa]
etc.
#[acPpdbPp]*

Hypothetical-strategy

It is beyond the scope of this page to explain why the definitions of hypothetical-strategy and compatible make sense.

(Only very short hints shall be given for the mathematically interested reader: The definition of hypothetical-strategy introduces objects implicitly. With explicit definitions the text would be thrice as long. - You, the informatician, know how a tree looks like, don't you? So consider all the hypothetical-sequences that are supposed to form your hypothetical tree and construct its graphical representation from them: A node is a position and an arrow is a hypothetical-move. - Now consider the definition of hypothetical-strategy. Why does each left-part end by a hypothetical-move of the player? It is his hypothetical-strategy! He makes a decision only when it is his right to make a hypothetical-move and not when it is his opponent's right. What does the "not two are equal" condition mean? After some left-part it is the player's turn and he chooses one and only one follow-up hypothetical-move! Why is the maximal set condition necessary? It guarantees that the player makes one decision whenever it is his turn in the hypothetical tree, i.e. he never forgets to make a hypothetical-move at all when it is his turn.)

You are an amateur or professional Go player but not a mathematician. What do you do? Overlook the definition details of hypothetical-strategy. Compatible means simply that some hypothetical-sequence is an answer to some hypothetical-strategy; the opponent's decisions may not be ignored.

What is hypothetical-strategy of a player and hypothetical-strategy of his opponent? Simplifying, it works like your tactical reading ahead in a tsume-go problem. The term contains the word strategy instead of tactics because hypothetical-strategy is ultimately complete. It is as complete as if you start playing a game with the empty grid and already imagine ALL possible strategies as if you were omniscient. Hypothetical-strategy is not quite as tough because typically it is applied to the final-position, which normally is a pretty sincere strategic simplification compared to the empty grid's position. In practice, one has to make extreme further simplifications by means of informal assumptions like the following: "Do not self-atari a basic living string! Do not self-atari a big seki string! Do not consider obvious failures!"

Examples for the Terms Hypothetical-strategy, Compatible, and Force

We consider the double ko position again and make the usual simplifications.

(1)

final-position

. # # .
# # O #
O # O O
. O O .

intersections

. # # c
# # O d
b # O O
a O O .

Hypothetical-strategies

Black's hypothetical-strategy 1:
{
#[a],
#[acP],
#[acPpd],
#[acPpdbP],
#[acPpdbPpa],
etc.,
#[acPbd],
#[acPbdPa],
etc.,
#[app],
#[acPpdpp],
#[acPbdpp],
etc.
}

Black's hypothetical-strategy 2:
{
#[p],
#[pca],
#[pcaPd],
#[pcaPdbP],
#[pcaPdbPca],
etc.,
#[pcapp},
#[pcaPdbpPa],
etc.
}

Other hypothetical-strategies of Black are omitted.

Not hypothetical-strategies:

This is not a hypothetical-strategy of Black:
{
O[p],
etc.
}

Reason: O[p] does not start with a hypothetical-move of Black.

This is not a hypothetical-strategy of Black:
{
#[a],
#[p],
etc.
}

Reason: #[a] without its last hypothetical-move a and #[p] without its last hypothetical-move p are equal, even though #[] is not even a left-part. (The definition speaks of equality, not of equality of left-parts. #[] denotes the empty sequence, which implicitly can be assumed to be well-defined.)

This is not a hypothetical-strategy of Black:
{
#[a],
#[acP],
#[acp],
etc.
}

Reason: #[acP] without its last hypothetical-move P and #[acp] without its last hypothetical-move p are equal: #[ac].

This is not a hypothetical-strategy of Black:
{
#[a],
#[ac],
etc.
}

Reason: #[ac] does not end with a hypothetical-move of Black.

This is not a hypothetical-strategy of Black:
{
#[a],
#[acP],
#[acPbd],
#[acPbdPa]
}

Reason: There is at least one left-part of some hypothetical-sequence missing: the left-part #[pcp] of the hypothetical-sequence #[pcpp].

Compatible:

Consider the following hypothetical-strategy of Black:

{
#[a],
#[acP],
#[acPpd],
#[acPpdbP],
#[acPpdbPpa],
etc.,
#[acPbd],
#[acPbdPa],
etc.,
#[app],
#[acPpdpp],
#[acPbdpp],
etc.
}

This is a compatible hypothetical-sequence: #[app]. It is compatible because each of its left-parts that end with a hypothetical-move of Black is in the hypothetical-strategy: #[a], #[app].

This is not a compatible hypothetical-sequence: #[pp]. It is not compatible because at least one of its left-parts that end with a hypothetical-move of Black is not in the hypothetical-strategy: #[p].

Force:

Suppose Black shall force capture of the white-string b. Black tries his hypothetical-strategy {#[a], #[acp], etc.}. Each compatible sequence has to start by #[a..]. Therefore the aforementioned Black hypothetical-strategy can force capture of the white-string b.

Suppose Black shall force capture of the white-string b. Black tries his hypothetical-strategy {#[p], #[pca], etc.}. There is at least one compatible hypothetical-sequence without a capture of the white-string b: #[pp]. Therefore not each compatible hypothetical-sequence fulfils capture of the white-string b. Hence Black's try of his hypothetical-strategy {#[p], #[pca], etc.} is a bad choice. Black has to choose a better hypothetical-strategy like {#[a], #[acp], etc.}. As has been seen, this Black's hypothetical-strategy can force capture of the white-string b. It is sufficient that there is at least one suitable hypothetical-strategy of his.

Hypothetical-analysis

The hypothetical-analysis looks at the final-position and says to each black-string or white-string therein either "You are alive!" or "You are dead!". Simple, is it? It is not.

Each string of the final-position has a life and death status ("alive" or "dead") that has to be determined. This means that the hypothetical-analysis consists of many subanalyses, one for each string. In practice, one determines the status of one string, after that the status of another string, etc., until one will have determined each string's status.

So we are supposed to be determining the status of one particular string in the final-position. First we note that it belongs to one particular player (either Black or White). Then we ask the following questions about the string:

Is the string uncapturable? If yes, then we know the status: "alive". Otherwise we continue with the next question.
Is the string capturable-1? If yes, then we know the status: "alive". Otherwise we continue with the next question.
Is the string capturable-2? If yes, then we know the status: "alive". Otherwise we go to the next step.
The status of the string is: "dead".

Uncapturable

During the hypothetical-analysis of the final-position, we are considering a particular string of a particular player and the question "Is the string uncapturable?". How do we answer this question? We apply the definition:

"A player's final-string is uncapturable if the opponent cannot force capture of its stones."

To apply the definition, we have to understand it.

In the definition, there is the player and the opponent. The player is the player of the string. The opponent is his opponent.

There is a string at all. The string is a string in the final-position. There is the player of the string, i.e. its colour.

The really interesting term is "force". When the word is used informally to describe a strategic aspect, then intuitively the meaning is rather clear. However, here we do not have an informal word but the formal term. So in principle, the definition of "force" must be understood.

One thing should be clear for the careful reader: The opponent starts every hypothetical-sequence studied for the definition of uncapturable. Each starts from the final-position.

If both players want the same, then it is all too easy to avoid capture of the string. Do the players want the same? No. The definition says something else.

The opponent's desire is the most brutal: Unless he has already found a successful hypothetical-strategy, he wants to try EACH possible hypothetical-strategy! He wants to try very weak hypothetical-strategy and he also wants to try very strong hypothetical-strategy. The player must find an answer FOR EACH hypothetical-strategy of the opponent. Thereby it is ensured that the player answers also a possibly successful hypothetical-strategy of the opponent. For each hypothetical-strategy of the opponent, the player can find several answers but the player must find AT LEAST ONE answer to succeed. A successful answer is a hypothetical-sequence that is compatible to the currently considered hypothetical-strategy of the opponent and avoids capture of the string.

How does the opponent think? The opponent does not think at all. He does not need to think. He simply exercises brute-force. Thereby he finds the AT LEAST ONE successful hypothetical-strategy if some exists at all that is successful, i.e. captures the final-string and does not offer the player any successful compatible hypothetical-sequence.

Some weak hypothetical-strategies of the opponent will not even capture the string and the player has an easy time to find an answer. Some intermediate hypothetical-strategies of the opponent will capture the string if the player chooses weakly and will not capture the string if the player chooses strongly. If they exist, then some strong hypothetical-strategies of the opponent will capture the string and require the player to choose strong answers.

Since the opponent tries all possible hypothetical-strategies, necessarily he also tries the strong hypothetical-strategies. For them it matters if the player can find strong enough answers, i.e. whether they do exist. The string is uncapturable if the player always finds an answer that is so strong to avoid capture of the string.

Forget your typical problem book with just one answer variation per problem. To determine if the string is uncapturable, one has to study an infinite number of variations because the opponent has infinitely many weak attempts and a few strong attempts. Only in practice one has to make compromises by ignoring all (supposedly) weak attempts. The rules are well-defined but in practice one can apply them only incompletely because we do not live infinitely long and therefore cannot do what the rules ask us to do. This is a fundamental feature of Japanese style rules and their terrible design. Although the Japanese 2003 Rules explain Japanese style rules completely in theory, the Japanese 2003 Rules lack the power to prolong the lives of all human beings for an infinite time.

As you may have noticed, hypothetical-strategy in the definition of uncapturable has an objective that differs from strategy during the alternating-sequence, during which Black wants to maximize the score of and until the final-position while White wants to minimize it. During the hypothetical-analysis there are other objectives. In an analysis of a particular string that determines if it is uncapturable, the defender has the objective to avoid capture of the string while implicitly the attacker has the objective to achieve capture of the string. For the sake of completeness, also the means (play, pass versus play, ko-pass, pass) and the ko rules (basic-ko, long cycle ko rules, direct-ko rule versus hypothetical-ko rules) differ.

Examples for Uncapturable

(1)

final-position

# # # # # #
# . # . . #
# # # # # #

intersections

# # # # # #
# . # a b #
# # # # # #

analysis of the black-string

Is the black-string uncapturable?
We simplify by assuming that Black does not fill a one space eye.
White has the following hypothetical-strategies and for each Black has a compatible hypothetical-sequence that avoids capture of the black-string:

White: {O[p], O[pap], O[pbp]}.Black: O[pp].
White: {O[a], O[app], O[abp]}. Black: O[app].
White: {O[b], O[bpp], O[bap]}. Black: O[bapp].

So the black-string is uncapturable.
The example is so simple that (of course, only within the limits of the simplifying assumption) a complete listing of all White hypothetical-strategies has been possible.

(2)

final-position

. # # # # # #
# # . # # . #
. . . . . . .
O O O O O O O
. . . . . . .
. O . . . . .

intersections

. # # # # # #
# # a # # b #
c d e f w g h
O O O O O O O
i j k l x m n
q r s t y u v

analysis of the black-string

Is the black-string uncapturable?
We simplify extremely by assuming that Black does not fill a one space eye, White does not fill a one space eye, many hypothetical-sequences are obvious, etc.
There are still too many available hypothetical-strategies of White. Therefore we make extreme simplifications as follows: We assume "strong" White play, do not denote hypothetical-strategies explicitly, and consider only hypothetical-sequences that are "interesting" for both players:

O[egwhpp]
O[eghwpp]
O[gedfpp]
O[gefdpp]
etc.

Whatever White tries, Black ensures that the black-string avoids capture. It is uncapturable.
Note: The hypothetical-analysis does not care about strategic mistakes that might have been made during the alternating-sequence.

analysis of the big white-string

Is the big white-string uncapturable?
We make extreme simplifications as follows: We assume that Black does not fill a one space eye, White does not fill a one space eye, many hypothetical-sequences are obvious, etc.
There are too many available hypothetical-strategies of Black. Therefore we make further extreme simplificaitons and assume "strong" Black play, do not denote hypothetical-strategies explicitly, and consider only hypothetical-sequences that are "interesting" for both players:

#[jkqipp]
#[jksipp]
#[kjlmuyxvtupp]
#[kjmltspp]
etc.

Whatever Black tries, White ensures that the big white-string avoids capture. It is uncapturable.
Note: Of course, all stones removed or any empty spaces occupied or surrounded during some hypothetical-sequence are just hypothetical and do not have any impact on the score.

analysis of the small white-string

Is the small white-string uncapturable?
We simplify extremely by assuming that Black does not fill a one space eye, White does not fill a one space eye, many hypothetical-sequences are obvious, etc.
There are too many available hypothetical-strategies of Black. Therefore, simplifying extremely again, we assume "strong" Black play, do not denote hypothetical-strategies explicitly, and consider only hypothetical-sequences that are "interesting" for both players:

#[jkqipp]
#[jksipp]
#[kjlmuyxvtupp]
#[kjmltspp]
etc.

Whatever Black tries, White ensures that the small white-string avoids capture. It is uncapturable.
Note: Do you confuse this analysis with that one for the big white-string? These two analyses are independent from each other! Consider it sheer coincidence if some hypothetical-sequences look familiar, however, here they serve a fundamentally different purpose: to prove uncapturable of the small white-string.

Permanent-stone

While an uncapturable string is a string that is already present in the final-position and for that the opponent cannot force its capture, a permament-stone is a stone that is not already present in the final-position but is PLAYED during a hypothetical-sequence and that is then not removed during the rest of the hypothetical-sequence. So although the terms uncapturable and permanent-stone are about avoiding capture, they are otherwise different.

Permanent-stone is a helping term used by the terms capturable-1 and capturable-2.

(1)

final-position

# # # # # #
# . # . . #
# # # # # #

intersections

# # # # # #
# . # a b #
# # # # # #

Consider the hypothetical-sequence O[abpp]. In it, the stone played as O[ab] is a permanent-stone because it is not removed during the rest of the hypothetical-sequence.

(2)

final-position

. # # .
# # O #
O # O O
. O O .

intersections

. # # c
# # O d
b # O O
a O O .

Consider the hypothetical-sequence #[app]. In it, the stone played as #[a] is a permanent-stone because it is not removed during the rest of the hypothetical-sequence. The stone is a permanent-stone despite the possible hypothetical-sequence #[acPpdbPp]*. What makes the stone a permanent-stone is its relation to the considered hypothetical-sequence #[app]. The stone does not exist for itself but exists due to this hypothetical-sequence.

Local-1, Local-2, Local-3

Each local-x environment is given for a particular final-string. Each final-string has its own local-x environments, although it can happen that environments of different final-strings are identical.

Local-1 is the same as "under the final-string". Only those intersections of the final-string are included that are intersections of the final-string as it appears in the final-position. If during a hypothetical-sequence the string should grow, this does not enlarge the local-1 of the final-string in the final-position.

Local-2 means roughly "under the final-string or nearby as far as not reaching uncapturable or capturable-1 final-strings of the final-string's player".

Local-3 means roughly "under the final-string or nearby as far as not reaching uncapturable or capturable-1 final-strings of either player".

We shall already consider examples, although capturable-1 is explained a little later.

(1)

final-position

O . . .
# # # #
O O O O
. . . .

local-1 of the black-string

O . . .
1 1 1 1
O O O O
. . . .

local-2 of the black-string

2 2 2 2
2 2 2 2
2 2 2 2
2 2 2 2

local-3 of the black-string

3 3 3 3
3 3 3 3
O O O O
. . . .

local-1 of the big white-string

O . . .
# # # #
1 1 1 1
. . . .

local-2 of the big white-string

2 2 2 2
2 2 2 2
2 2 2 2
2 2 2 2

local-3 of the big white-string

O . . .
# # # #
3 3 3 3
3 3 3 3

local-1 of the small white-string

1 . . .
# # # #
O O O O
. . . .

local-2 of the small white-string

2 2 2 2
2 2 2 2
O O O O
. . . .

local-3 of the small white-string

3 3 3 3
# # # #
O O O O
. . . .

(2)

final-position

. O # # .
# # O # .
O O O # .
. . O # O

local-1 of the small black-string

. O # # .
1 1 O # .
O O O # .
. . O # O

local-2 of the small black-string

2 2 # # .
2 2 2 # .
2 2 2 # .
2 2 2 # O

local-3 of the small black-string

3 O # # .
3 3 O # .
O O O # .
. . O # O

local-1 of the upper white-string

. 1 # # .
# # O # .
O O O # .
. . O # O

local-2 of the upper white-string

2 2 2 2 2
2 2 O 2 2
O O O 2 2
. . O 2 2

local-3 of the upper white-string

3 3 # # .
3 3 O # .
O O O # .
. . O # O

local-1 of the right white-string

. O # # .
# # O # .
O O O # .
. . O # 1

local-2 of the right white-string

. O 2 2 2
# # O 2 2
O O O 2 2
. . O 2 2

local-3 of the right white-string

. O # # 3
# # O # 3
O O O # 3
. . O # 3

(3)

final-position

. O O # O . # .
O O # . # # # #
. O # # # # # .

local-1 of the small black-string

. O O 1 O . # .
O O # . # # # #
. O # # # # # .

local-2 of the small black-string

2 2 2 2 2 2 # .
2 2 # 2 # # # #
2 2 # # # # # .

local-3 of the small black-string

. O O 3 3 3 # .
O O # 3 # # # #
. O # # # # # .

local-1 of the small white-string

. O O # 1 . # .
O O # . # # # #
. O # # # # # .

local-2 of the small white-string

. O O 2 2 2 2 2
O O 2 2 2 2 2 2
. O 2 2 2 2 2 2

local-3 of the small white-string

. O O 3 3 3 # .
O O # 3 # # # #
. O # # # # # .

Capturable-1, Capturable-2, Capturable-3

An uncapturable final-string is neither capturable-1, capturable-2, nor capturable-3. A final-string might be capturable-2 only if it is not capturable-1. Capturable-3 is a special type of capturable-2 final-strings and only a ko-stone might be capturable-3. Therefore one first determines all uncapturable final-strings, then all capturable-1 final-strings, then all capturable-2 final-strings, then all capturable-3 final-strings.

(1)

final-position

O . . .
# # # #
O O O O
. . . .

intersections

a b c d
# # # #
O O O O
. . . .

uncapturable final-strings U

O . . .
U U U U
U U U U
. . . .

determination why the white-string a is not capturable-1
#[bpp]
etc.

Black forces both capture of the white-string a and that a white permanent-stone does not appear under the white-string a (i.e. local-1).

capturable-1 final-strings: none

O . . .
# # # #
O O O O
. . . .

determination why the white-string a is not capturable-2

To recall, White needs to try to play a permanent-stone on one of the intersections 2:

2 2 2 2
2 2 2 2
O O O O
. . . .

hypothetical-sequences
#[bpp]
#[bcdpp]
#[bdcpp]
etc.

White fails. In other words, Black forces both capture of the white-string a and that a white permanent-stone does not appear on any of the intersections a, b, c, d (or of the black-string's intersections) that are the local-2 of the white-string a.

capturable-2 final-strings: none

O . . .
# # # #
O O O O
. . . .

(2)

final-position

. O # # .
# # O # .
O O O # .
. . O # O

intersections

a b # # .
c d O # g
O O O # h
e f O # i

uncapturable final-strings U

. O U U .
# # U U .
U U U U .
. . U U O

determination why the black-string c-d is not capturable-1
O[app]
O[acdpp]
O[adcpp]
etc.

White forces both capture of the black-string c-d and that a black local-1 permanent-stone does not appear.

determination why the white-string b is capturable-1
#[abpp]
#[abadpp]
#[eacdpp]
etc.

Black cannot force both capture of the white-string b and prevent a white local-1 permanent-stone.

determination why the white-string i is not capturable-1
#[hpp]
#[hgpp]
#[hgpapp]
etc.

Black both forces capture of the white-string i and prevents a white permanent-stone on the intersection i.

capturable-1 final-strings 1

. 1 # # .
# # O # .
O O O # .
. . O # O

determination why the black-string c-d is not capturable-2

To recall, Black may try to establish a permanent-stone on any of the local-2 intersections:

2 2 # # .
2 2 2 # .
2 2 2 # .
2 2 2 # O

hypothetical-sequences
O[app]
O[acdpp]
O[adcpp]
O[aefgpp]
etc.

White can both force capture of the black-string c-d and prevent any black local-2 permanent-stone.

determination why the white-string i is not capturable-2

To recall, White may try to establish a permanent-stone on any of the local-2 intersections:

. O 2 2 2
# # O 2 2
O O O 2 2
. . O 2 2

hypothetical-sequences:
#[hpp]
#[hepp]
etc.

Black can both force capture of the white-string i and prevent any white local-2 permanent-stone.

capturable-2 final-strings: none

. O # # .
# # O # .
O O O # .
. . O # O

(3)

final-position

. O O # O . # .
O O # . # # # #
. O # # # # # .

intersections

. O O b c d # .
O O # a # # # #
. O # # # # # .

uncapturable final-strings U

. U U # O . U .
U U U . U U U U
. U U U U U U .

determination why the white-string c is not capturable-1
#[dpp]
etc.

THe white-string c is captured while White cannot play a permanent-stone on the intersection c.

determination why the white-string c is not capturable-2

To recall, White may try to establish a permanent-stone on any of the local-2 intersections:

. O O 2 2 2 2 2
O O 2 2 2 2 2 2
. O 2 2 2 2 2 2

hypothetical-sequences
#[dpp]
etc.

THe white-string c is captured while White cannot play a permanent-stone on any local-2 intersection.

determination why the black-string b is not capturable-1
O[aPbpp]
etc.

determination why the black-string b is capturable-2

To recall, Black may try to establish a permanent-stone on any of the local-2 intersections:

2 2 2 2 2 2 # .
2 2 # 2 # # # #
2 2 # # # # # .

hypothetical-sequences
O[adpp]
O[adbpp]
etc.

White can capture the black-string b but Black can play a local-2 permanent-stone on the intersection d.

capturable-2 final-strings 2

. O O 2 O . # .
O O # . # # # #
. O # # # # # .

The capturable-2 final-string b is a ko-stone. Therefore we have to verify whether it is also capturable-3.

determination why the black-string b is capturable-3

To recall, local-3 of the black-string b is:

. O O 3 3 3 # .
O O # 3 # # # #
. O # # # # # .

Local-2 and local-3 of the capturable-2 ko-stone b are unequal. Therefore it is capturable-3. Thus the direct-ko rule prohibits the final-position from occurring.

capturable-3 final-strings 3

. O O 3 O . # .
O O # . # # # #
. O # # # # # .

Scoring

For a final-position, the hypothetical-analysis has been performed, either alive or dead is assigned to each final-string, and the final-position can be scored.

Of course, you are going to complain that scoring uses so many terms. They are necessary to distinguish scoring intersections outside, what the rules consider, sekis from not scoring intersections in sekis and to have a precise method for determination of the score in the general case of an arbitrary final-position. If you do not like the many terms, then oppose traditional Japanese rules. They need them to maintain tradition. If you want Japanese rules and want them to maintain tradition, then you must bear all those terms.

Compared to the Japanese 1989 Rules, the Japanese 2003 Rules introduce a few more terms:

black-eye-string / white-eye-string: These terms enable us to have a definition text for the terms "black-eye-point" / "white-eye-point" that the reader can understand without too great difficulties. Without the terms "black-eye-string" / "white-eye-string", understanding the definition text for "black-eye-point" / "white-eye-point" would require exceptional mathematical education.
black-region / white-region: The Japanese 1989 simply failed to introduce and define these terms. The Japanese 2003 Rules fill this gap.

Maybe it is easier to see the terms for scoring in action, so we enter examples immediately:

(1)

final-position

. # . .
# # # #
# # # #
O O O O
. O . O

(Each string is alive.)

black-eye-strings B

B # B B
# # # #
# # # #
O O O O
B O B O

white-eye-strings W

W # W W
# # # #
# # # #
O O O O
W O W O

black-eye-points B

B # B B
# # # #
# # # #
O O O O
. O . O

white-eye-points W

. # . .
# # # #
# # # #
O O O O
W O W O

eye-points E and dame D (none)

E # E E
# # # #
# # # #
O O O O
E O E O

black-regions B and white-regions W

B B B B
B B B B
B B B B
W W W W
W W W W

regions not in-seki R and regions in-seki S (none)

R R R R
R R R R
R R R R
R R R R
R R R R

black-territory B and white-territory W

B # B B
# # # #
# # # #
O O O O
W O W O

scoring intersections for Black

1 # 1 1
# # # #
# # # #
O O O O
. O . O

scoring intersections for White

. # . .
# # # #
# # # #
O O O O
1 O 1 O

prisoner-difference = 0

score = 1

result = Black win

(2)

final-position

# # # # # #
# . # . O #
# # # # # #

hypothetical-analysis: A = alive, D = dead

(details omitted)

A A A A A A
A . A . D A
A A A A A A

black-eye-strings B

# # # # # #
# B # B B #
# # # # # #

white-eye-strings W

# # # # # #
# W # W O #
# # # # # #

black-eye-points B

# # # # # #
# B # B B #
# # # # # #

white-eye-points W (none)

# # # # # #
# . # . O #
# # # # # #

eye-points E and dame D (none)

# # # # # #
# E # E E #
# # # # # #

black-regions B

B B B B B B
B B B B B B
B B B B B B

white-regions W (none)

# # # # # #
# . # . O #
# # # # # #

black-regions not in-seki B and black-regions in-seki S (none)

B B B B B B
B B B B B B
B B B B B B

white-regions not in-seki W (none) and white-regions in-seki S (none)

# # # # # #
# . # . O #
# # # # # #

black-territory B and white-territory W (none)

# # # # # #
# B # B B #
# # # # # #

scoring intersections for Black

# # # # # #
# 1 # 1 2 #
# # # # # #

scoring intersections for White (none)

# # # # # #
# . # . O #
# # # # # #

prisoner-difference = 0

score = 4

result = Black win

(3)

final-position

. O O O .
# # # # #

(Each string is alive.)

black-eye-strings B

B O O O B
# # # # #

white-eye-strings W

W O O O W
# # # # #

black-eye-points B (none)

. O O O .
# # # # #

white-eye-points W (none)

. O O O .
# # # # #

eye-points E (none) and dame D

D O O O D
# # # # #

black-regions B

. O O O .
B B B B B

white-regions W

. W W W .
# # # # #

black-regions not in-seki B (none) and black-regions in-seki S

. O O O .
S S S S S

white-regions not in-seki W (none) and white-regions in-seki S

. S S S .
# # # # #

black-territory B (none) and white-territory W (none)

. O O O .
# # # # #

scoring intersections for Black (none)

. O O O .
# # # # #

scoring intersections for White (none)

. O O O .
# # # # #

prisoner-difference = 0

score = 0

result = tie

(4)

final-position

. # O # O
# . O # O
O O O . O
O O O O O

hypothetical-analysis: A = alive, D = dead

(details omitted)

. A A D A
A . A D A
A A A . A
A A A A A

black-eye-strings B

B # O # O
# B O # O
O O O B O
O O O O O

white-eye-strings W

W # O W O
# W O W O
O O O W O
O O O O O

black-eye-points B

B # O # O
# . O # O
O O O . O
O O O O O

white-eye-points W

. # O W O
# . O W O
O O O W O
O O O O O

eye-points E and dame D

E # O E O
# D O E O
O O O E O
O O O O O

black-regions B

B B O # O
B . O # O
O O O . O
O O O O O

white-regions W

. # W W W
# . W W W
W W W W W
W W W W W

black-regions not in-seki B (none) and black-regions in-seki S

S S O # O
S . O # O
O O O . O
O O O O O

white-regions not in-seki W (none) and white-regions in-seki S

. # S S S
# . S S S
S S S S S
S S S S S

black-territory B (none) and white-territory W (none)

. # O # O
# . O # O
O O O . O
O O O O O

scoring intersections for Black (none)

. # O # O
# . O # O
O O O . O
O O O O O

scoring intersections for White (none)

. # O # O
# . O # O
O O O . O
O O O O O

prisoner-difference = 0

score = 0

result = tie

Note: White should not have passed too early during the alternating-sequence, but he may have made that strategic mistake.

Reasons for Some Particular Rules

Alternating-sequence without Reference to an Infinite Number of Moves

The definition of the alternating-sequence does not need to refer to an infinite number of moves because there is the long-cycle rule.

Finite and Infinite Hypothetical-sequences

Japanese Go rules tradition failed to reflect which hypothetical-sequences should be finite or infinite. Therefore interpretations have some freedom. In particular, the following designs are possible:

Finite and infinite hypothetical-sequences exist.
Robert Pauli: All hypothetical-sequences are infinite. Passes do not have an ending function.
The Fixed Ko Rule makes all hypothetical-sequences finite.

For the Japanese 2003 Rules with its intention to model professional tradition as closely as possible, the Fixed Ko Rule is not really an option because its method resembles superko, which professional Japanese Go rules tradition rejects. Letting all hypothetical-sequences be infinite is elegant (because the end conditions of the definition of hypothetical-sequence can be omitted) but the reading ahead does not resemble a typical Go player's reading any longer; he expects sequences that are non-cyclical on the board to end.

A hypothetical-sequence can be infinite at all because a superko principle is not used and the ko-pass restrictions do not prohibit cute or dull long cycles.

Basic Terms Used for the Hypothetical-analysis

Everything related to analysis of the final-position is called "hypothetical" because the final-position is not disturbed.

"capture" and "final-position" are defined so that reading terms about life and death and for determination of the result becomes easier.

In terms about life and death, the Japanese 2003 Rules use the well-defined terms "hypothetical-strategy" and "hypothetical-sequence" where the Japanese 1989 Rules use the undefined "cannot".

The terms "compatible" and "force" are not called "hypothetical-compatible" and "hypothetical-force" because it is believed that that would not encourage reading the rules text. One should note, however, that "force", as used for a Go player's strategy during the alternating-sequence, is not the same "force" that is used in the Japanese 2003 Rules, although there is a close functional relation.

Ko Restrictions during the Alternating-sequence

As the wording of the basic-ko rule shows, superko could easily replace both it and the long-cycle rule. With superko it would be more easy to do without further restriction rules for the hypothetical-analysis. However, tradition demands that superko be not used.

The long-cycle rule also defines long-cycle wins so that a player cannot refuse to accept a loss by continuing to move forever.

The purpose of the direct-ko rule (and the terms "local-3" and "capturable-3") is to maintain tradition in some particular classes of endgame positions.

The terms "ko" and "ko-capture" are useful only for terms related to hypothetical-analysis. The terms are not needed for the basic-ko rule.

Ko-pass and Ko-pass Rules

Ko-passes and ko-pass rules are necessary because there is at least one position that would contradict tradition if instead the basic-ko rule were used as the one and only ko restriction during hypothetical-sequences:

(1)

final-position

. # # # # #
# # . # O #
O O # O . O
. O O O O O
O # # # # #
# # . # . #
# # # # # #
O O O O O O
O O . O . O

intersections

z # # # # #
# # c # e #
O O d O f O
a O O O O O
b # # # # #
# # . # . #
# # # # # #
O O O O O O
O O . O . O

Tradition, which wants to justify the feeling that z shall be a so called eye while the big upper white-string is deemed to have no so called eye, demands that all the upper white-strings are dead. However, the basic-ko rule as the only ko restriction lets the big upper white-string be uncapturable:
#[acfbpp]
#[acfbde]*
#[fcpp]
#[fcaepp]
#[fcaedb]*
#[fcaebzpp]
etc.

Under the ko-pass rules of the Japanese 2003 Rules, the big upper white-string is not uncapturable (nor capturable-1, nor capturable-2, but dead):
#[aPf..]
#[acfPd..]
etc.

The move types "ko-pass" and "pass" differ mainly in the restriction rules for them: A move-sequence is ended by successive passes. Hypothetical-moves that are ko-captures are restricted due to ko-passes and rules for them.

Japanese rules authorities tried hard to describe the hypothetical ko rules precisely but they have failed thrice:

The Japanese 1949 Rules spoke of "kos where immediate means are available" versus "kos where immediate means are not available".
The World Amateur Go Championship 1979 Rules speak of "direct kos" versus "indirect kos".
The Japanese 1989 Rules speak of "a pass for a particular ko before recapturing in it is allowed again".

The hypothetical ko rules of the Japanese 2003 Rules succeed in all test positions of the author. These include any official positions of the three written Japanese rulesets and many further essential positions.

Relaxed Ko-pass Restriction for Both Players instead of Only One Player

The ko-pass rules use a restriction concept for all kos together and not for each ko separately so that some examples of the second part of the official commentary on the Japanese 1989 Rules are explained correctly. Otherwise it would also have been possible to consider each ko separately. Probably the authors of the Japanese 1989 rules did not realize the adverse effect of that concept on those examples. If they had noticed it, then the spirit of the Japanese 1989 Rules suggests that they would have chosen a rule that describes them better.

(1)

This is example 16 of the second part of the official commentary on the Japanese 1989 Rules.

final-position

. O O . O . O # . # O
O # # O O O O # # O O
# # . # # O # # O . O
. # # # O # # # # O O
# # O O O O O # # . O
O O O . O . O # # # O

intersections

b c d e O f O # g h O
a # # O O O O # # O O
# # . # # O # # i j O
. # # # O # # # # O O
# # O O O O O # # k O
O O O . O . O # # # O

Under the ko-pass rules of the Japanese 1989 Rules, the right big white string is uncapturable:
#[jgHIhiJG]*
#[b][gjIHihGJ]*
#[bgjIedciHkpp]
etc.

Under the ko-pass rules of the Japanese 2003 Rules, the right big white string is not uncapturable (nor capturable-1, nor capturable-2, but dead):
#[bpedcpePf..]
#[bPedcpePf..]
#[bgjPedcpePf..]
#[bgjPedcihPegf..]
#[bgjPedcihPePf..]
#[bgjPeif..]
etc.

Summarizing this example, the Japanese 1989 Rules contradict the official commentary on the Japanese 1989 Rules while the Japanese 2003 Rules agree to what the official commentary on the Japanese 1989 Rules says about life and death in the example!

Note: For other examples, this failure of the Japanese 1989 Rules and this success of the Japanese 2003 rules can be observed as well.

Types of Alive

There are three types of "alive" in the rules: 1) "uncapturable", 2) "capturable-1", 3) "capturable-2". Why? Because tradition demands it. Only from a view of elegant rules design, one could declare anything that is not "[uncapturable-]alive" to be "dead".

One might also be tempted to omit "capturable-2". However, professional tradition demands a particular behaviour of example 2 of the second part of the official commentary on the Japanese 1989 Rules. Among the official examples this one has the most frequent shape, which John Fairbairn estimates as once every 800 games. Considering that the long-cycle rule is never used even once in most amateur go players' lives, a low frequency is not sufficient reason for omitting a rule. When the Japanese professionals want a certain rules behaviour for a certain shape, then the rules must acknowledge that. Of course, in principle it would be possible to become more reasonable in future like using the concept of "control" as in the New Amateur-Japanese Rules. - Until recently everybody overlooked a slightly more frequent - because it can occur everywhere on the board and not just in the corner - application of "capturable-2":

final-position

O O O O O O O O .
O # # # # # # # O
O # . # . # # # O
O # # # O # # # O
O # . O # O . # O
O # # # . # # # O
O # # # . # # # O
O # # # # # # # O
. O O O O O O O O

Seki

Under the Japanese 1989 Rules or the Japanese 2003 Rules, one must not confuse the strategic term "seki" with the rules term "in-seki". Most in-seki black-regions or white-regions are not sekis but the consequence of a player's oversight not to fill some intersections during the alternating-sequence that will be dame in the final-position. The rules do not define "seki" because a definition for that is even more complicated and longer than the rules related to "in-seki". Hence only strategically one can speak of "no territory in sekis" while "seki" has been undefined so far in such contexts. However, precise rules do not include undefined terms.

As an aside, some strategic terms like "seki" and "independently-alive" have been defined formally and generally by the author elsewhere.

Tournament Rules

Tournament rules are not included in the Japanese 2003 Rules.

A counting method for practically determining the score is considered to be a tournament rule and therefore omitted.

Game stop, confirmation phase, resumption, effective moves, loss of both players are not included in the rules because dame and teire can be occupied formally and unequivocally during the alternating-sequence. So far some Japanese professionals have preferred to be informal and equivocal about dame and teire. The Japanese 2003 Rules intend to be logical and therefore leave it to tournament rules to be unreasonable. The same has to be said about formal instead of informal passes and about none instead of countless shapes with undefined scores.

Understanding the Japanese 2003 Rules

The rules text does not consist of rules but of a few axioms and many definitions. Axioms and definitions are the precise means to specify rules while rules would be imprecise. Why? A definition carefully relies on other definitions (or axioms). Thereby things are explained hierarchically and each term can be reduced to the axioms (and to the most common words of natural language). Contrarily, although rules would also rely on definitions, rules would do so only partially and fail to create a complete hierarchy of terms.
Mathematical rules prefer to introduce "intersection" as an axiom and define "line" while the Japanese 2003 Rules introduce "lines" as an axiom and define "intersection". For each rules text, the reason is a significantly shorter text length.
The Japanese 2003 Rules are praised to be "logical" while mathematical rules can be even more logical. The Japanese 2003 Rules are just about as logical as rules that rely on terms defined by natural language can be. Logic is one of the fundamental fields of mathematics but the purpose of Go rules is not to reinvent mathematics. Therefore as a compromise still quite some natural language is tolerated in the Japanese 2003 Rules. On the other hand, so little undefined natural language is used that the rules contents is unequivocal for every careful reader.
The Japanese 2003 Rules are praised to be "complete". What does this mean? Each used term is defined, where "defined" indicates a possible reduction to axioms and the most common words of natural language. There is no term-like phrase that is not defined. The rules create only such objects (like, e.g., "positions") to that the rules can be applied. I.e., the rules can always be applied. E.g., each "final-position" has a "score" and it has one and only one particular "score".
For an understanding of the rules, there are the bottum-up approach from the axioms, via the basic definitions, to the more and more difficult definitions and the top-down approach from the most difficult definitions, via the definitions of its used terms, to the axioms. In practice one is probably well advised to use both approaches.
The rules have a structure. Terms are defined for the purpose of their usage. It helps to recognize every term's usage. Some terms have a rather restricted, specialized usage while others occur everywhere.
Many terms are supposed to have names that assist one's intuitive understanding as a Go player. However, quite some terms have "hypothetical" in their names, what prolongs them. The maybe not so apparent reason is that they are used for the "hypothetical-analysis" but not for the "alternating-sequence", where similar terms could be used but would require slightly different definitions. E.g., one might want to speak of "strategy" in case of the "alternating-sequence" while one speaks of "hypothetical-strategy" in case of the "hypothetical-analysis".
One of the greatest difficulties of understanding the Japanese 2003 Rules is to distinguish terms for essential or exceptional Go concepts. Japanese rules are by far more difficult than they could be (if they shall model tradition) since they treat many exceptional concepts.
"player" as used in one term and "player" as used in another term might be the same or might be different players. E.g., they might be the player and the opponent. E.g., in the term "capturable-1" the text "the opponent" in "the opponent cannot force [...]", when applied to the term "force", becomes "a player". I.e., inserting "the opponent" as "a player", one gets: "An opponent can force something if there is at least one hypothetical-strategy of his so that each compatible hypothetical-sequence fulfils that something." Then applying "cannot" instead of "can" one gets: "An opponent cannot force something if there is not one hypothetical-strategy of his so that each compatible hypothetical-sequence fulfils that something." Why? Existence of some hypothetical-strategies is the major qualification of the conditional clause. "not one" is the opposite of "at least one" here. We speak of the opposite since we use "cannot" instead of "can". Next one would replace the "something".
[Further explanations are beyond this webpage.]

Best Model for Professional Japanese Rules

Failure of Various Models

The Japanese 1949 Rules have in particular the following weaknesses:

They tried to define life and death more by referring to shapes while clear definitions of hypothetical-sequence and hypothetical-strategy would have been necessary.
They listed some precedental shapes and asked the reader to generalize while the rules should have provided the generalization.
Everything is ambiguous or very ambiguous.

The World Amateur Go Championship 1979 Rules are very close to the Japanese 1949 Rules. The weaknesses are essentially the same.
The Japanese 1989 Rules have the following weaknesses:

They use only the word "cannot" where they should have explained hypothetical-sequence and hypothetical-strategy.
They fail in many other aspects: see here.

Amateurs pretend to use Japanese rules verbally. However, understanding is weak. To hide their weak understanding, many amateurs refuse even to discuss rules or their application. When amateurs pretend to explain Japanese rules to beginners, they do not explain rules about life and death but they pretend that strategy would be the rules. However, strategy is not the rules themselves but the consequences of their application. Amateurs do not understand the rules of life and death but only common strategic consequences. Not understanding the rules themselves is not a good explanation of them.
A lot of logical model rulesets (like, e.g., the Logical Japanese Rules of Go by Robert Pauli) have tried to approach Japanese rules. While they do approach them, they all fail to do so sufficiently closely: In quite some practical positions they fail to treat them as everybody playing Japanese style rules expects that they should be treated.
The New Amateur-Japanese Rules explain Japanese rules tradition pretty well but not in some rare classes of positions. There they differ from professional Japanese rules tradition.

Success of the Japanese 2003 Rules

The rules are as logical, unequivocal, complete, and precise as semi-formal rules could be. - It has been said many times elsewhere why these characteristics are very important for a ruleset. Therefore it need not be repreated here.

The Japanese 2003 Rules describe professional Japanese rules tradition so well that one might explain it by the rules. However, how do we know anything about the quality of that description? With what do we compare the rules to evaluate that "what"?

For all test positions published together with official professional Japanese rules texts, the Japanese 2003 Rules assess the same score as professional Japanese rules tradition expects. Where the latter varied, the Japanese 2003 Rules agree to the newest official commentary or rules. Besides the Japanese 2003 Rules agree to professional Japanese rules tradition (as far as it has ever been clear enough) in all (!) other test positions provided by rules experts. The scope of test positions is very broad and deep. The Japanese 2003 Rules are the first ruleset ever that explains professional Japanese rules tradition correctly.

Japanese 1989 Rules

The Japanese 2003 Rules should replace the Japanese 1989 Rules, which have badly defined terms, are incomplete, are illogical, and are used in particular by the Japanese Go Associations. Until then, the Japanese 2003 Rules can serve as the authoritative commentary on the Japanese 1989 Rules.

The Japanese 2003 Rules are not designed to explain only the Japanese 1989 Rules but all rulesets that have been used or influenced by Japanese professionals.

The Japanese 1989 Rules have many flaws. The following 28 threads on news:rec.games.go, which might be retrieved via www.google.com, advanced groups search, provide a pretty good discussion:

Flaws of the Japanese 1989 Rules (000)
Flaws of the Japanese 1989 Rules (001)
Flaws of the Japanese 1989 Rules (002)
Flaws of the Japanese 1989 Rules (003)
Flaws of the Japanese 1989 Rules (004)
Flaws of the Japanese 1989 Rules (005)
Flaws of the Japanese 1989 Rules (006)
Flaws of the Japanese 1989 Rules (007)
Flaws of the Japanese 1989 Rules (008)
Flaws of the Japanese 1989 Rules (009)
Flaws of the Japanese 1989 Rules (010)
Flaws of the Japanese 1989 Rules (011)
Flaws of the Japanese 1989 Rules (012)
Flaws of the Japanese 1989 Rules (013)
Flaws of the Japanese 1989 Rules (014)
Flaws of the Japanese 1989 Rules (015)
Flaws of the Japanese 1989 Rules (016)
Flaws of the Japanese 1989 Rules (017)
Flaws of the Japanese 1989 Rules (018)
Flaws of the Japanese 1989 Rules (019)
Flaws of the Japanese 1989 Rules (020)
Flaws of the Japanese 1989 Rules (021)
Flaws of the Japanese 1989 Rules (022)
Flaws of the Japanese 1989 Rules (023)
Flaws of the Japanese 1989 Rules (024)
Flaws of the Japanese 1989 Rules (025)
Flaws of the Japanese 1989 Rules (026)
Flaws of the Japanese 1989 Rules (027)

Here is a citation from the last thread:

Newsgroups: rec.games.go
Subject: Flaws of the Japanese 1989 Rules (027)
From: jasiek@snafu.de (Robert Jasiek)
Date: Sat, 01 May 2004 09:13:08 GMT

This is a summary of the flaws.

Abbreviations:
P. = preamble
* = and other §§
-- = not a flaw but a translation problem
- = not exactly a flaw but unnecessary confusion
o = flaw
+ = very important flaw
++ = extremely important flaw

No. +- § Contents

001   o P. ambiguous "spirit of good sense and mutual trust"
002   o 1   ambiguous "compete in skill"
003   o 1   incomplete aim "more territory"
004   o 1   ambiguous "game"
005   o 2   unspecified starting player of alternating play
006   o 2* ambiguous "move"
007   - 3   confusing selective reference to exceptions
008 -- 6   imprecise translation: for "ko", "can" should be "could"
009   o 6   ambiguous scope of application of the basic ko rule
010   o 7.1 ambiguous "stones"
011   o 7.1 unspecified starting player of hypothetical play
012   o 7.1 unspecified hypothetical nature
013   o 7.1 unspecified principle of alternation
014   o 7.1 unspecified nature of hypothetical moves
015   o 7.1 unspecified available types of hypothetical moves
016   o 7.1 unspecified effect of successive hypothetical passes
017   o 7.1 unspecified meaning of infinite hypothetical sequences
018   o 7.1 ambiguous nature of non- or recursive "alive" construction
019   + 7.1 ambiguous "would enable"
020   o 7.1 ambiguous conditional construction "if capturing [...]"
021 ++ 7.1 missing definitions of hypothetical-sequence, -strategy
022   o 7.2 ambiguous scope of application of the ko-pass rule
023   o 7.2 ambiguous difference between pass and ko-pass
024   o 7.2 unspecified successions of pass / ko-pass
025   o 7.2 ambiguous consequences of alternating sequence on ko-passes
026   o 7.2 unspecified presupposition for early ko-pass
027   o 7.2 unspecified validity of double ko-pass
028   o 7.2 ambiguous "temporarily disappearing" ko
029   o 8   ambiguous "surrounded"
030   o 8   ambiguous "[in] seki"
031   + 9.2 ambiguous purpose of confirmation phase
032   o 9.2 ambiguous determination of "end of the game"
033   - 9.3 superfluous resumption
034   o 9.3 ambiguous simultaneous resumptions
035   o 9.3 unspecified repetitive resumptions
036   + 10.1 ambiguous nature of application of life-death definitions
037   o 10.1 ambiguous "opposing dead stones from his territory"
038   o 10.2 unspecified treatment of excess prisoners
039   o 12   ambiguous "position"
040   + 12   ambiguous strategy due to uncomparable "no result"
041   o 12   unspecified treatment of infinite alternation
042   o 13.1 ambiguous "effective move" and "affect"
043   - 13.1 superfluous loss of both

--
robert jasiek

Furthermore, as pure interpretations of the rules text show, the Japanese 1989 Rules contradict Japanese Go rules tradition when the rules are applied to some kinds of examples. In particular, the official commentary on the Japanese 1989 Rules violates and contradicts those rules for several examples.

Further Research

The rules can be modified to explain Korean rules or the referee ko rules of the Chinese 1988 Rules. Probably a rules modification can also be used to classify Fighting Kos and Disturbing Kos of the Ing 1991 Rules more clearly.
The rules can be used by computer go programmers.
The rules can be used to explain strategy or shapes.
During development of the rules, the author has discovered new kinds of shapes like, e.g., stable anti-sekis.