Kee Rules of Go / Feedback

Sub-page of KeeRulesOfGo

Discussion referring to earlier versions of Kee Rules

Robert Pauli: Dear Wilton, some humble suggestions and stupid questions :--)

  1. not adjacent to any intersection in the same status: Actually the closeness of a non-singleton group is missing too.
  2. if an intersection of the former is adjacent to that of the latter: Since order is of no matter, if an intersection of one is adjacent to that of the other might be better.
  3. emtpy group: Since this sounds like empty set, uncolored group or unoccupied group might be better.
  4. An area of a player is the union of her territories and groups adjacent to each other.: Why adjacent to each other? Couldn't that be dropped?
  5. A declared area is an area formed by capturing groups of all players declared dead.: Don't get it.
    • Area with no alive stones?
    • What's a capturing group?
  6. anticipated area: Is that an area with no own alive stones?
  7. two players playing with black and white stones: Better two players, one playing with black, the other with white stones or has two colors, black and white. Players play with a unique color.
  8. cannot place a stone to produce a captive group in her color: At the very moment of placement this very well is allowed, so maybe add . . . if no opposing groups become captive.
  9. all captive opponent groups are captured from: Better . . . are removed from.
  10. cannot place a stone to produce: Maybe better may not place . . .
  11. A player has to declare which of her groups are dead . . .:
    • Why the heck should she declare any stone of herself dead at all??
    • She herself passed for the first time in that position or nobody but she did?
  12. A player can anticipate . . .: Why shouldn't she boldly declare all opposing stones dead??
  13. to produce a board position ever passed by any player: Maybe to produce a board position at least one player already has passed in.
  14. A player can use her anticipated area to replace her declared area with release of all groups captured in the declared area after anticipation, if all intersections in the anticipated area also belong to the declared area.
    Not even I get it :--))
  15. ranked higher: Ranking is one's performance over many games. Maybe A player has won if each other player has less points might be better.
  16. Hint: Once I praised my rules on Rules of Go -- those with this and that -- but I purged that again . . .

Robert Pauli:

  • A2. ... Stones are conventionally black or white.
  • A7. Stone placement of a player is using her stone to occupy an unoccupied intersection Cut.
    "such that its status is changed to her occupied intersection" looks like a further condition, but actually is implied.
  • A9. A group is formed by connecting all adjacent intersections with the same status iteratively.
    "All"? Not really clear. Maybe:
    Groups form by repeatedly uniting two adjacent intersections that are in the same state.
  • B4. A player may not place a stone to produce a position that has ever occurred before unless its last occurrence was followed by a not yet expired renewal.

Discussion referring to the last version of Kee Rules

So far - still puzzled by the rest . . .

Robert Pauli: Kee, call'em fluid rules - they change at least once a week ;-)

Wilton Kee: Sorry about that. Hopefully it seems converging to a concrete one.


Robert Pauli:

  • B.4 Isn't that the same as:
    A player may not place a stone to produce a position which has ever occurred unless its last occurrence was ban deleted.
    I mean, if it's recovered later, it wasn't the last occurence . . . or do you mean that ban deletions on other positions also take the ban from this one?

Wilton Kee:

Ban deletion, despite being played on one position, has its effect of removing bans on every previous position.

Similarly, ban recovery, despite occurring on one position when such position is repeated, has its effect of recovering bans on every previous position.

Ban deletion may not be played on the same position twice. However, it does not mean that the ban effect to a position may not be activated / deactivated twice. This can happen when ban deletion is played on two different later positions.


Robert Pauli: Wilton, is the life cycle of a position with regard to B

     1 -> 2 -> 3 -> 4

with

  1. never occurred (yet)
  2. banned deletable
  3. ban deleted
  4. banned undeletable (recovered)

and

  • stone placement creating P moves P from 1 to 2 or from 3 to 4 (and P may not be in 2 or 4)
  • ban deletion moves all positions in 2 to 3, and
  • position affirmation changes no B aspect

?

Wilton Kee:

Recovered does not imply undeletable. Therefore, 3 may go to 2. Actually we may never know if a ban on a position is 2 or 4.


Robert Pauli:
Think my problem was confusing direct with indirect ban deletions, Wilton.

A direct ban deletion on one position indirectly deletes bans from all former positions. Any position may be subject to a direct ban deletion once only, but there is no limit on the number of indirect ban deletions it may take. As I get it now, B5 is only referring to direct ban deletions, whereas B3 refers to direct as well as indirect ones.

Let me try to put it in other words:

After cutting apart the sequence of positions at each point a ban deleting pass occurred,

  • none of the subsequences may contain any of its positions more than once (B4)
  • none of the subsequences may end with the same position as any other (B5)
  • and none of the subsequences may contain more than two positions that already occurred in former subsequences (B3+B4).
    ("two"? The first position in each subsequence necessarily repeats the last position of the preceding subsequence.)

Now right?

Wilton Kee:

You are right. However, I do not quite follow your third statement:

  • and none of the subsequences may contain more than two positions that already occurred in former subsequences (B3+B4).
    ("two"? The first position in each subsequence necessarily repeats the last position of the preceding subsequence.)

Could you please explain a little bit more?

Robert Pauli:

Sorry for being cryptic, Wilton. It's: position, move, position, move, etc. - any action being a "move", even a pass. Now, since the position right before a (ban-deleting) pass is the same as the one after it, but this isn't the intentional repetition we want to enable, we have to allow a second.

You sidestep this by only banning repetitions made via stones, which is fine, but please remove the confusion between direct and indirect ban deletion, it's cruical.

Wilton Kee:

I choose not to use terms like "direct" or "indirect" ban deletion just because I believe things get even more complicated when more terms are defined.

This is not an easy sentence to digest at a first glance, I admit. But as long as it does not produce confusion, I think it is fine to keep it. Do you agree?

Robert Pauli:

No, sorry. Keep it as simple as possible, but not simpler - as Albert would put it.

Reading

A player may not delete ban on a position on which ban has ever been deleted.

strictly - and how else should one read rules? - prevents ban deletion in all positions that occurred prior to any ban deletion since the ban was removed from them as well.

If we have P1 m1 P2 m2 P3 and neither m1 nor m2 was a ban deletion, then P3 neither may repeat P1 nor P2 - each banned. If it continues with m3 P4 m4 P5 m5 and m3 is a ban deletion, implying P3 = P4, then P5 may repeat P1 - P4. If P5 = P3, the rule above prevents m5 to be a ban deletion, as intended. However, if P5 = P2 you want to allow ban deletion, but your rule does not: m3 deleted a ban from P2, not?

Wilton Kee:

For other readers' information, "P" stands for position and "m" stands for move in what written above.

First of all, if m3 is a ban deleting move (pass), it is correct that P3 = P4. However, P5 may not repeat P4 (= P3) because m3 only deletes P1 - P3.

If P5 = P3, then m5 must be a pass (as this implies P5 = P4). However, you are right that m5 cannot be a ban deleting pass because P3 (= P5) has been ban-deleted by m3.

If P5 = P2 and P2 <> P3, then m5 can be a ban deleting pass if there has been no ban deleting pass made on P2. m3 deleted a ban from P2 but not a chance to make a ban deleting pass on P2.

Robert Pauli:

You're right, P5 may only repeat P1 - P2. I slipped (ignoring my own "none of the subsequences may contain any of its positions more than once").

You're also right in that m5 necessarily has to be a pass - but that wasn't my intention, sorry. To be more general I should have added, say, P6 m6 P7 m7 P8 m8, with P5 - P7 being fresh positions, and asked about P8 and m8 instead, but since it doesn't matter, let's leave it as is.

With what I still can't agree, however, is that

m3 deleted a ban from P2 but not a chance to make a ban deleting pass on P2.

If "m3 deleted a ban from P2" and the rule is

A player may not delete ban on a position on which ban has ever been deleted.

anybody will conclude that no more ban deletion is possible on P2 - no matter what the rule designer's intention actually was.

Wilton, by no means do I want to push you to use "direct/indirect". This just was my quick and dirty way to make things clear.

The problem seems to come from being able to read "ban deletion" in two ways. First, as technical term, a special kind of pass, and, second, as a description of an effect on a position, no longer being banned. How about "ban-deleting pass" versus "ban deletion"? For instance:

A 5. A play can either be a stone placement, a ban-deleting pass, or a position-affirming pass.

B 4. A player may not place a stone to produce a position which has ever occurred unless its last occurrence was followed by a ban-deleting pass and such pass was not followed by a recovery.

B 5. A player may not pass ban-deleting in a position in which this ever happened.

So far the first usage. The second usage only appears in B3, as far I see. However, why isn't

B 3. A stone placement recovers ban when it produces a position on which ban has ever been deleted.

simply

B 3. A stone placement recovers ban when it produces a position which already occurred.

?

Wilton Kee:

The reason is simple - because it is just not the case.

Rule B 3 says ban recovered when a stone placement produces a position on which a ban-deleting pass has been made before. Therefore, this "ban-deletion" is still on its first usage.

To summarize:

1. Ban-deleting pass:

- always refers to a specific pass

- actively chosen by a player

2. Ban-recovering stone placement:

- always refers to a specific stone placement

- not actively chosen by a player

- a stone placement is ban-recovering when a ban-deleting pass has been made on the resulting position

3. To test whether a stone placement is valid:

- first check if the resulting position has occurred before.

- If no, valid. If yes, check if a ban-deleting pass has been made after its last occurrence.

- If no, invalid. If yes, check if it is followed by a ban-recovering stone placement.

- If yes, invalid. If no, valid.

4. To test whether a ban-deleting pass is valid:

- check if a ban-deleting pass has been made on the same position.

- If yes, invalid. If no, valid.


Robert Pauli:

So the usage in B3 also is in the first sense (ban-deleting pass) - I see:

B 3. A stone placement recovers ban(s) when it produces a position ever produced by a ban-deleting pass.

Let me still recommend my distinction to enable (meta-)talk of one ban-deleting pass deleting ban from several positions.

Thank you for the step-by-step summary. Very good! However, shouldn't point 3 strictly be

3. To test whether a stone placement is valid:
- first check if the resulting position has occurred before.
- If no, valid. If yes, check if a ban-deleting pass has been made after its last occurrence without being followed by a ban-recovering stone placement.
- If no, invalid. If yes, valid.

?

Wilton Kee: What is the difference?

Another question: If only one ban-deleting pass ever has occurred and nobody performs a second one, no ban-recovering stone placement actually can happen. Right?

Wilton Kee: Right.

Guess I got part B now (:-). Let's move on.

I'd prefer

G2. The area of a player consists of all her stones and territories in the final position.

since anything of herself without two eyes was kicked off during testifying and we want to refer to her complete area, no matter if it comes in one part or not.

Wilton Kee: It is a good suggestion.

Let me also (humbly) correct some English:

  • A2 . . . representing herself.
  • G4a . . . if it belongs to her . . .
  • G4b . . . if it does not belong to her . . .
  • G4c . . . unless it was removed in the . . .
  • Use "testifying phase" instead of "testified phase".

Wilton Kee: Thank you for your valuable check.

Regarding the last point and "consolidating phase" / "consoldated position", you might consider to use

  • "testified position" instead "final position" and
  • "final position" instead "chilled position".

Wilton Kee: I understand "chilled position" may not be a good term and "final position" does cause some confusion. I just think of using "scoring position" to refer to the final position in normal phase and "testified position" to refer to the position after testifying phase. What do you think?

It's not quite clear how the extra phases are restricted by positions that occurred outside of them. Guess they don't matter. Right?

Oh! it seems there's a bug in G4a. Shouldn't it be

A player gets one point from every not by herself occupied intersection in the chilled position if it belongs to her area in all final positions.

Otherwise you're ignoring territory beneath dead stones, which still are around in the chilled position.

Wilton Kee: It is not a bug. Actually instead of the player getting one point, it's the one who owns the dead stone loses one point.

It seems that your "Japanese" territory scoring doesn't treat bent four correctly. Example:

[Diagram]

Seki plus Bent Four (3x12 board)



Ignoring when exactly White has to capture in the seki to earn points, in Japan this will end as B+9
(= 12-3 = White takes two, Black throws in one, and White takes it - finished, with four white stones dead in bent four.)

Under your rules, however, White will hold back the capture to keep a ko threat. This will prevent Black from starting the ko in any consolidating phase: only B+2. Right?

Wilton Kee: I admit. We actually cannot use positional superko in testifying phase if we are going to strictly follow Japanese Rule. I will work on it. Any suggestion?

Robert Pauli:

  • What's the difference? Well, if the last occurrence of the position in question is follwed by a ban-deleting pass and a ban-recovering placement follows that one, you conclude that the position may not be repeated by placement, but since another ban-deleting pass could follow, this may indeed be legal.
  • Game ends in scoring position, testifying phase ends in testified position: fine. An alternative for testifying/testified might be purging/purged.
  • You gave no answer how position repetition is treated during consolidating and testifying. Please do (especially if history starts afresh).
  • Wilton, G4a must be a (technical) bug. Simple example:
[Diagram]

3x3, no captives

Position affirmed by all, but no agreement on declaration: chilled, nothing removed. No matter if Black is tested or White, Black "seals off" a 3x3 area in either case. According to G4a, Black receives 3 points; according to G4b, White receives -3 points: B+6. But, of course, this is B+9. G4a missed to count the intersections covered by the dead white stones.

  • Don't admit - I was wrong. As I see it now, your rules actually give B+9 (best play and fixed G4a assumed): If Black is tested, Black starts the ko and trades the seki. However, this is of no use for White, since when White is tested, Black does not start the ko but simply waits until White gets paralized, then takes all. Since Black gets the right half in either case, the four white stones are dead. Since each player gets the left half in one line, it doesn't count: B+12. Cosidering this, White squeezes three captives out of the seki before chilling: B+9.
  • Let me try again:
[Diagram]

21x3, no captives: W+12, not W+8



If the left end is seki, as usual, this is W+12. Under your rules, however, we have to check two lines:

Black starts
No need to start the ko since it goes to Black as soon White gets paralized.
White starts
White removes ko threats (a and b) while Black sets up the ko. After White retook it, Black threatens with c. White ignores that, trading center for left end.

So, left end and white center belong to nobody due to changing ownership. The rest is W+8. Right?


Discussion referring to the June 2005 version of Kee Rules

Wilton Kee: Robert, I have amended quite a lot in this version. I am sure the amended version does answer all your questions above, despite the fact that new questions may arise. :)

The left side of the above diagram would always be a seki under the amended rule because of the created-group superko rule (stronger than situational superko rule). For normal ko, it is equivalent to the philosophy in Japanese Rule that only passes can lift the ko ban. Therefore, the amended version also gives W+12 in this case.


This is a copy of the living page "Kee Rules of Go / Feedback" at Sensei's Library.
(OC) 2011 the Authors, published under the OpenContent License V1.0.
[Welcome to Sensei's Library!]
StartingPoints
ReferenceSection
About