KGS, bent four in japanese rules, need help.

[Xylol]

Hi,

this is the situation that occured in a rated KGS game (japanese rules) a few minutes ago.
Black passed, white passed and whilst counting we have diffrent opinions about the black group in the top left.

I think the c18 group is dead.

My question: do i have to play out the situation or is the group dead by default without playing the ko? This question will probably affect the outcome of the game.

A KGS Admin (Mef) permitted me to post the game, even if it is not finished yet. (My opponent had to go and the game is adjourned)

Any help is apreciated, thank you for your time.

http://files.gokgs.com/games/2013/2/10/Xylol-tavr.sgf

[EdLee]

Xylol, Black is dead in this case. No need to play out the ko.

(Unless there is some very special situation on the rest of the board, say, another bent four where White is dead, etc.)

One way to think about this -- for this board:
- Black can not play in the upper corner first (it is self-atari immediately and Black dies);
- White can get rid of ALL the ko threats on the rest of the board, one by one, without losing any points;
- Then, to capture Black, White first connects at A19. Black still cannot play there. Then White ataris Black (say, C19).
Black must take. Then White plays at 1-2 (B19). Black is forced to throw in at A19. White now takes the ko FIRST --
since there is zero ko threats anywhere else on the board, White wins the ko and captures Black.

[Xylol]

thank you, but i still have 3 questions:

1) how can i remove ko threats without losing points, at which state of the game i should do this and is it applied correctly in kgs?

2) how about unremovable ko threats like g14, because of the seki situation there?

3) If I'am going to capture black, I will lose 3 points:

w Connect: -1
w put b in atari: -1
black takes: +1 (for black stone) -4 (b capt. 4 stones) +4 (territory is larger by 4 now)
w put one on 2-1 : -1
b throw in: +1
w take ko: -1
w kill: -1

[jts]

Xylol, Black is dead in this case. No need to play out the ko.

(Unless there is some very special situation on the rest of the board, say, another bent four where White is dead, etc.)

One way to think about this -- for this board:
- Black can not play in the upper corner first (it is self-atari immediately and Black dies);
- White can get rid of ALL the ko threats on the rest of the board, one by one, without losing any points;
- Then, to capture Black, White first connects at A19. Black still cannot play there. Then White ataris Black (say, C19).
Black must take. Then White plays at 1-2 (B19). Black is forced to throw in at A19. White now takes the ko FIRST --
since there is zero ko threats anywhere else on the board, White wins the ko and captures Black.

Ed, you've given the explanation for Chinese rules - but note the seki which is in the same corner. That's not a ko threat that White is going to want to remove! This is a board where the rule sets give different results.

In Japanese rules the explanation of the bent four in the corner shape is much simpler; white is not obligated to capture dead stones during the game. After both players have passed, black can demand that W prove that he can capture the black stones, but no ko threats against external groups are allowed during this demonstration.

Note that Japanese rules don't call for removing ko threats, whereas Chinese rules don't call for the score to change when ko threats are removed.

[RobertJasiek]

Let me first correct mistakes in earlier opinions about Japanese style rules:

- Usually, life and death at the game end do not depend on numbers of (remote) ko threats.
- Ko threats need not be eliminated.
- External ko threats are allowed during sequences proving local status. Status proof play is global but status property is (usually) local. (Unless a group covers the whole board, so that local coincides with global.)
- For a player's stones to be status assessed, it is not the player to play first in hypothetical play.
- It is insufficient to prove death by capture, but one must exclude also other types of life (such as establishing a related "new" stone that is then permanent).

What are KGS-Japanese Rules?

Essentially they are undefined. In case of doubt, an administrator or moderator decides the outcome of a game. However, he should not make an arbitrary decision but stick to the fact that the rules are Japanese style (with the major exceptions that sekis can have territory and programming bugs exist, as you can see in the example SGF's markup). It is undefined which Japanese ruleset KGS-Japanese Rules refer to. Therefore, we can only assume that KGS-Japanese Rules use the intersection of all existing, currently sufficiently relevant written or verbal Japanese style rulesets, i.e., the intersection of

- verbal Japanese rulesets,
- the Japanese 1989 Rules and its later modifications,
- the World Amateur Go Championship 1980 Rules.

The following are not currently relevant:

- the Japanese 1949 Rules and its later modifications,
- the World Amateur Go Championship 1979 Rules.

Concerning verbal Japanese rulesets, there are, WRT to bent-4, two variants:

- no special bent-4 rule but the basic ko rule applies (since it always applies),
- special bent-4-in-the-corner-is-always-dead rule.

Concerning J1989 versus WAGC1980 Rules, there is the same variety of no versus some special rule for bent-4.

So, in order to get a proper interpretation of KGS-Japanese Rules, one must also find out whether both variants of bent-4 handling in Japanese style rulesets yield the same outcome.

The special bent-4 rule of the "dead independently of the rest of the board" type creates extra problems for "bent-4-adjacent-to-seki-shape". For such a special rule, it is undefined whether

- the black bent-4 string is dead but remains on the board, and there is no territory or
- the black bent-4 string is dead, is removed, there is territory and the seki shape breaks down, yielding more territory.

This undefined judgement cannot produce a common interpretation with J1989 Rules. Therefore, one referee judgement could be: the intersection of Japanese style rulesets is undefined for bent-4-adjacent-to-seki-shape. Hence, the game result is "undefined".

The other possible referee judgement ignores the "dead but remains on the board" possibility and finds out the remaining intersection judgement. However, due to the "local" ko threat at G/H14, the exact wording of the hypothetical ko rule and the life and death definitions can matter here. Possibly, this is the third shape where basic ko rule versus pass-for-ko rules matters. To find out whether the still considered rulesets have a common intersection here, each must be applied with great care. At the moment, I lack time for that; it requires about one to three hours of careful thinking. You can find out yourself by reading

http://home.snafu.de/jasiek/rules.html

and applying also every conceivable hypothetical ko rule(s) to Verbal Japanese Rules in comparison to J1989 and J2003. Maybe it is even relevant whether D19 and G15 are considered one group or two groups. The conclusion for the remaining intersection of rulesets could be "undefined" (if they disagree with each other) or "dead" (if they agree).

So the simple, no further study judgement is "undefined". The careful judgement could turn out to be "undefined" or "undefined or dead".

[shapenaji]

Xylol, Black is dead in this case. No need to play out the ko.

(Unless there is some very special situation on the rest of the board, say, another bent four where White is dead, etc.)

One way to think about this -- for this board:
- Black can not play in the upper corner first (it is self-atari immediately and Black dies);
- White can get rid of ALL the ko threats on the rest of the board, one by one, without losing any points;
- Then, to capture Black, White first connects at A19. Black still cannot play there. Then White ataris Black (say, C19).
Black must take. Then White plays at 1-2 (B19). Black is forced to throw in at A19. White now takes the ko FIRST --
since there is zero ko threats anywhere else on the board, White wins the ko and captures Black.

Ed, you've given the explanation for Chinese rules - but note the seki which is in the same corner. That's not a ko threat that White is going to want to remove! This is a board where the rule sets give different results.

In Japanese rules the explanation of the bent four in the corner shape is much simpler; white is not obligated to capture dead stones during the game. After both players have passed, black can demand that W prove that he can capture the black stones, but no ko threats against external groups are allowed during this demonstration.

Note that Japanese rules don't call for removing ko threats, whereas Chinese rules don't call for the score to change when ko threats are removed.

Actually, in this case the rules give the same, the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

[RobertJasiek]

the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

This kind of reasoning is insufficient for Japanese rules because there are Japanese style rulesets where life can be given by means of newly established then permanent stones. It is necessary to judge whether such permanent stones are locally related. (J1989 do not specify such a local relation, but it has been demonstrated that that was a mistake. J2003 specify a local relation.)

EDIT:

Under J2003, it is clear as follows: The string G15 is uncapturable. Therefore the local-2 of the black bent-4 string excludes the intersections of the string G15 and its adjacent empty intersections. Therefore the black bent-4 string is also not capturable-2, but dead.

So, under J2003, a ko threat on G/H14 is useless because it is outside the local-2. A ko threat in the nearer seki shape is useless because White can prevent any black permanent stone in the local-2 by filling it with a white two-eye-formation.

(All "local" white strings are alive. The nearer seki shape's black string is dead as before.)

J1989 interpreted by J2003 yields the same. J1989 interpreted only by J1989 itself yields "undefined". I favour the former.

Interpretation of Verbal Japanese Rules will be more difficult.

[jts]

Actually, in this case the rules give the same, the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

I think you're thinking of the temporary seki at A13, but to quote Yoda: "there is another"!

[quantumf]

Actually, in this case the rules give the same, the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

But then black can capture the large J15 group?

Wow, OP, this is a fascinating position. I can see why the admin asked you to post it the forum.

[shapenaji]

Actually, in this case the rules give the same, the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

I think you're thinking of the temporary seki at A13, but to quote Yoda: "there is another"!

oh, I just completely overlooked that one... yeah, the exchange is certainly better.

My bad! I fixated on that other seki.

[shapenaji]

the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

This kind of reasoning is insufficient for Japanese rules because there are Japanese style rulesets where life can be given by means of newly established then permanent stones. It is necessary to judge whether such permanent stones are locally related. (J1989 do not specify such a local relation, but it has been demonstrated that that was a mistake. J2003 specify a local relation.)

EDIT:

Under J2003, it is clear as follows: The string G15 is uncapturable. Therefore the local-2 of the black bent-4 string excludes the intersections of the string G15 and its adjacent empty intersections. Therefore the black bent-4 string is also not capturable-2, but dead.

So, under J2003, a ko threat on G/H14 is useless because it is outside the local-2. A ko threat in the nearer seki shape is useless because White can prevent any black permanent stone in the local-2 by filling it with a white two-eye-formation.

(All "local" white strings are alive. The nearer seki shape's black string is dead as before.)

J1989 interpreted by J2003 yields the same. J1989 interpreted only by J1989 itself yields "undefined". I favour the former.



Interpretation of Verbal Japanese Rules will be more difficult.

Yeah, I was more concerned with the result in chinese scoring. I figured that Japanese rules just held that it was dead.

I didn't see the unremovable non-local threat. (Embarassingly enough), so I was just suggesting that the resolution to OP's problem was simple.

It's obviously not, because of the ko threat, and because KGS doesn't actually implement Japanese rules, just japanese scoring if the players can agree on L+D. There is clearly a problem here, since KGS has no agreed-upon rules for resolving this (that I'm aware of).

[shapenaji]

But then black can capture the large J15 group?

Wow, OP, this is a fascinating position. I can see why the admin asked you to post it the forum.

yeah, fixated on the wrong threat

And yeah, it's a cool problem

[HermanHiddema]

the seki is not a viable ko threat in that ko, because white can finish the ko in answer to the threat

This kind of reasoning is insufficient for Japanese rules because there are Japanese style rulesets where life can be given by means of newly established then permanent stones. It is necessary to judge whether such permanent stones are locally related. (J1989 do not specify such a local relation, but it has been demonstrated that that was a mistake. J2003 specify a local relation.)

EDIT:

Under J2003, it is clear as follows: The string G15 is uncapturable. Therefore the local-2 of the black bent-4 string excludes the intersections of the string G15 and its adjacent empty intersections. Therefore the black bent-4 string is also not capturable-2, but dead.

So, under J2003, a ko threat on G/H14 is useless because it is outside the local-2. A ko threat in the nearer seki shape is useless because White can prevent any black permanent stone in the local-2 by filling it with a white two-eye-formation.

(All "local" white strings are alive. The nearer seki shape's black string is dead as before.)

J1989 interpreted by J2003 yields the same. J1989 interpreted only by J1989 itself yields "undefined". I favour the former.



Interpretation of Verbal Japanese Rules will be more difficult.

Yeah, I was more concerned with the result in chinese scoring. I figured that Japanese rules just held that it was dead.

I didn't see the unremovable non-local threat. (Embarassingly enough), so I was just suggesting that the resolution to OP's problem was simple.

It's obviously not, because of the ko threat, and because KGS doesn't actually implement Japanese rules, just japanese scoring if the players can agree on L+D. There is clearly a problem here, since KGS has no agreed-upon rules for resolving this (that I'm aware of).

OP problem is actually simple, under Japanese rules the group in the game is just dead.

[shapenaji]

OP problem is actually simple, under Japanese rules the group in the game is just dead.

I think Robert's point is valid, under which Japanese rules? KGS doesn't actually specify, "Japanese Rules" is just a way to score at the end once life and death is specified.

[HermanHiddema]

OP problem is actually simple, under Japanese rules the group in the game is just dead.

I think Robert's point is valid, under which Japanese rules? KGS doesn't actually specify, "Japanese Rules" is just a way to score at the end once life and death is specified.

The result is the same under any commonly accepted definition of "Japanese rules", Robert is just finding complications where there are none.

E.g. under Japanese 1989 rules (current Nihon Kiin rules), bent-four is dead because the only valid ko threat during life and death determination is a pass for that specific ko. So if the corner ko were played to determine life or death, white can respond to G14 with H14 and black is not allowed to retake the ko because he hasn't passed for it.

Under WAGC rules bent-four is specifically ruled dead by precedent.

Under any informal interpretation of "Japanese rules", bent-four is a well known dead group.

And there isn't really any other reasonable interpretation of "Japanese rules".