Do Japanese Rules ( or their KGS implementation ) suck ?

ethanb · **#21**

Bill Spight wrote:

RobertJasiek wrote:

Right, never trust a programmer about non-trivial implementations.

***

What does false eyes mean? Is living with false-eyes no territory?

I think that the term false eye is ambiguous in English, that different writers use it differently. Some would say that the eyes in the Two Headed Dragon ( http://senseis.xmp.net/?TwoHeadedDragon ) are false, some would not. I accept both usages, but I would prefer to call those eyes defective, to remove the ambiguity.

Weird, I've never seen those type of eyes called false (except in that Sensei's article, and even there they just say that they LOOK false.) I would feel it was odd if I heard the term "defective" in that context too - those are real eyes. But it's better than overloading "false eye" if you must have a term for the shape, that's for sure.

The only way I've seen "false eye" used is to mean "an open internal liberty of which one (or more) surrounding chain(s) of stones may be placed in atari if all external liberties are taken."

What books use it otherwise (so I can avoid them and warn others?)

kirkmc · **#22**

If they were truly false, you could force the player to connect, proving that they are false. Calling them false is showing a misunderstanding of the concept of eyes - the fact that two of them connect makes them real eyes and unkillable, hence they're not false.

topazg · **#23**

At the risk of opening a can of worms, perhaps real eyes should be "independent" eyes (i.e. an eye independent of the status of other connected chains)?

kirkmc · **#24**

Why? The example above shows two eyes that, because of their shapes, are real. Why do they need to be independent?

You often see situations (sorry, I don't know how to make diagrams easily), where a group has one eye at each end - what you could call "independent" eyes - and a false eye connecting two strings. Yet because of that situation the false eye is not false and counts as a point. It doesn't contribute to live, since if it were the second (and not third) eye, it would be truly false, but in its context it is real.

RobertJasiek · **#25**

Still I do not know whether KGS-Japanese Rules score points in fake false eyes of a) independent life or b) sekis.

kirkmc · **#26**

RobertJasiek wrote:

Still I do not know whether KGS-Japanese Rules score points in fake false eyes of a) independent life or b) sekis.

In most cases, sekis are clearly marked as not being points. But I recall seeing once or twice a seki that KGS though was either a living or dead group, giving points to one side or the other. For fake false eyes - if you mean what I said above, false eyes connecting two groups each with one eye - it counts the point, in my experience.

Woodstock · **#27**

amnal wrote:

Whilst this is techincally true, I've never ever seen it happen and be a problem. And I can't envisage a situation (well...I could maybe make a really contrived one) where it ever would be. You aren't losing time if the game is in the scoring phase, so that isn't a problem.

Disputes over the status of groups are a frequent occurrence with beginners ,
however. Also , during scoring phase a player can press undo and restart the game
along with the timer without consent of their opponent while he / she is talking
to an admin. This can be a problem.

Of course this particular issue doesn't depend on rulesets. But take a rather
complex , yet common dispute : Bent 4 in the corner.
Lots of people have never heard of it , thus insist on living in seki. Under Japa-
nese rules the group is dead automatically without having to play it out. All this
takes explaining and convincing and is pretty much unfeasible in a blitz game.

Afaik , with other rulesets both players can agree on seki. If one disagrees ,
the sequence has to be played out including removal of ko threats and subsequent
ko fight. However , there are cases with irremovable ko threats such as smaller
sekis , double kos etc which might enable the second player to win the ko.

Under Japanese rules these players are screwed although the game is favorable to
them just because they happened to choose a ruleset which didn't consider this
particular case.

amnal wrote:

As wms points out, whilst you are factually correct, your point is not a good argument when noting that this is more often a problem in all other rulesets.

Concerning the false eyes scoring problem ... I don't know about you , but I much
rather prefer a system where all liberties must be filled at all times to one
where only certain kinds of liberties must be filled.

Andd wrote:

This question seems a little unfair, like asking "Do you still beat your wife?".

Unlike your example , my question implies nothing. By answering you don't admit
to anything beyond expressing your opinion on the matter.

Andd wrote:

I also agree that it's somewhat standard on go servers and there's some merit to coherency.

Well sure , but just because everybody is using them doesn't mean we should prefer
Japanese rules. A friend of mine is programming a Go server of his own and I'll
be sure to tell him to choose a different standard ruleset ^^

amnal · **#28**

Woodstock wrote:

amnal wrote:

Whilst this is techincally true, I've never ever seen it happen and be a problem. And I can't envisage a situation (well...I could maybe make a really contrived one) where it ever would be. You aren't losing time if the game is in the scoring phase, so that isn't a problem.

Disputes over the status of groups are a frequent occurrence with beginners ,
however. Also , during scoring phase a player can press undo and restart the game
along with the timer without consent of their opponent while he / she is talking
to an admin. This can be a problem.

Of course this particular issue doesn't depend on rulesets. But take a rather
complex , yet common dispute : Bent 4 in the corner.
Lots of people have never heard of it , thus insist on living in seki. Under Japa-
nese rules the group is dead automatically without having to play it out. All this
takes explaining and convincing and is pretty much unfeasible in a blitz game.

Afaik , with other rulesets both players can agree on seki. If one disagrees ,
the sequence has to be played out including removal of ko threats and subsequent
ko fight. However , there are cases with irremovable ko threats such as smaller
sekis , double kos etc which might enable the second player to win the ko.

Under Japanese rules these players are screwed although the game is favorable to
them just because they happened to choose a ruleset which didn't consider this
particular case.

Everything you say is technically true, but the extreme unlikelihood of it ever being important makes me doubt its relevance. Maybe a bent four in the corner situation will lose a few beginners a few games, but so will stuff like missing atari, failing to punish stupid invasions, filling in their own eyes in seki etc. I won't deny that the Japanese treatment of bent four is not ideal, but I don't think it is a terrible burden for the beginner to bear.

Quote:

amnal wrote:

As wms points out, whilst you are factually correct, your point is not a good argument when noting that this is more often a problem in all other rulesets.

Concerning the false eyes scoring problem ... I don't know about you , but I much
rather prefer a system where all liberties must be filled at all times to one
where only certain kinds of liberties must be filled.

In the KGS system, all liberties must be filled at all times or the score may be wrong. I have no problem with that. But sometimes I don't play the game properly and leave liberties unfilled.

It is not the case that only certain kinds of liberties must be filled. It is the case that certain kinds of liberties must be filled to ensure that the score is correct. this perhaps promotes laziness, but it's hard to imagine a scoring system where this would not be true. Plus, again, the effect on anything is tiny.

Cassandra · **#29**

Woodstock wrote:

...
Bent 4 in the corner.
Lots of people have never heard of it , thus insist on living in seki. Under Japa-
nese rules the group is dead automatically without having to play it out.

Japanese rules have a local view on status evaluation, there is no room to "play something out using the whole board" after the end of "play". The sole and only question to answer during status evaluation is "Can this one and only one chain of stones be taken of the board ?" for each chain of stones on the board seperately.

In most cases, not in all, concluding status of Bent-Four will be found to be "dead".

Harleqin · **#30**

Cassandra wrote:

Japanese rules have a local view on status evaluation, there is no room to "play something out using the whole board" after the end of "play". The sole and only question to answer during status evaluation is "Can this one and only one chain of stones be taken of the board ?" for each chain of stones on the board separately.

It seems, then, that a server that implements "japanese rules", but does not implement such an analysis, actually has an incomplete rule set. In other words, one should not rely on the players to reach agreement through informal communication (as this only works when the players are sufficiently experienced), but, in the case of a dispute, start an analysis phase for each disputed string, probably with infinite takebacks. For this, you need to define "local", of course.

Cassandra · **#31**

Harleqin wrote:

For this, you need to define "local", of course.

I suppose that it is this need, together with some recursive elements needed to combine the single chains status results, that makes implementation so difficult.

This problem with "local" may be one reason, why Monte-Carlo-based programs are weeker in Semeai, compared to their overall performance.

wms · **#32**

Harleqin wrote:

Cassandra wrote:

Japanese rules have a local view on status evaluation, there is no room to "play something out using the whole board" after the end of "play". The sole and only question to answer during status evaluation is "Can this one and only one chain of stones be taken of the board ?" for each chain of stones on the board separately.

It seems, then, that a server that implements "japanese rules", but does not implement such an analysis, actually has an incomplete rule set. In other words, one should not rely on the players to reach agreement through informal communication (as this only works when the players are sufficiently experienced), but, in the case of a dispute, start an analysis phase for each disputed string, probably with infinite takebacks. For this, you need to define "local", of course.

The problem is that for the server to all on its own determine life/death is about as hard as creating a perfect go player. And having the server step the users through the resolution at the end of the game is incredibly complicated; it is usually newer players who need it, and explaining how the system works is way, way too complex for anybody but a rules expert to use. The old cgoban 1 app had such a system, but I threw it out when I went to cgoban 2 because it was worthless - anybody who knew enough to step through the system would know at a glance that bent 4 was dead so it was just an ugly confusing trap for people to accidentally turn on.

jts · **#33**

When people say that bent four in the corner is automatically dead under Japanese rules, what exactly do they mean? Japanese rules don't prevent you from playing out the ko, right? And if they don't prevent you from playing out the ko, what's the point of saying they're automatically dead? What could you do at the end of the game, other than play out the ko?

oren · **#34**

jts wrote:

When people say that bent four in the corner is automatically dead under Japanese rules, what exactly do they mean? Japanese rules don't prevent you from playing out the ko, right? And if they don't prevent you from playing out the ko, what's the point of saying they're automatically dead? What could you do at the end of the game, other than play out the ko?

This may help
http://www.cs.cmu.edu/~wjh/go/rules/Japanese.html

Life-and-Death Example 7-1: Bent Four in the Corner

jts · **#35**

oren wrote:

jts wrote:

When people say that bent four in the corner is automatically dead under Japanese rules, what exactly do they mean? Japanese rules don't prevent you from playing out the ko, right? And if they don't prevent you from playing out the ko, what's the point of saying they're automatically dead? What could you do at the end of the game, other than play out the ko?

This may help
http://www.cs.cmu.edu/~wjh/go/rules/Japanese.html

Life-and-Death Example 7-1: Bent Four in the Corner

So in other words "automatically dead" means "after two passes, it's dead"?

Dusk Eagle · **#36**

This is a bent-four position in this context:

Click Here To Show Diagram Code: [go]$$Wc $$ --------------------------------------- $$ | O O O X O . . . . . . . . . . . . . . | $$ | a X . X O . . . . . . . . . . . . . . | $$ | X X X X O . . . . . . . . . . . . . . | $$ | X O . O . O . . . , . . . . . , . . . | $$ | O O . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . |[/go]

If we were to determine its status based on whole-board play, white could first remove all his ko threats one at a time and then remove all of black's external liberties, then play at 'a'. Note that black cannot prevent white from playing at 'a' without killing himself. So after white plays 'a', a ko starts after black captures and white plays B19, with black having to find the first ko threat. However, black will have no ko threats, as white has removed them all. So in Japanese rules, this shape is dead as it stands.

flOvermind · **#37**

Harleqin wrote:

For this, you need to define "local", of course.

That's easy. For each disputed group, play out the situation on the whole board. When the group gets captured, it's dead. When it can't be captured, it's alive. Then, restore the board to the original position and remove the dead group.

That way, you can theoretically lose every other group on the board, as long as the disputed group stays alive. Also, when you look at it in that way, it becomes clear why in the hypothetical play phase of Japanese rules, only a pass is considered a valid ko threat. That rule, by the way, is also the real reason why bent four is dead in Japanese rules

Cassandra · **#38**

flOvermind wrote:

... in the hypothetical play phase of Japanese rules, only a pass is considered a valid ko threat. That rule, by the way, is also the real reason why bent four is dead in Japanese rules

Unfortunately, it is not so simple.

Click Here To Show Diagram Code: [go]$$W $$ ---------------- $$ | O O O . X W C $$ | . X X X X W C $$ | X X W W W W C $$ | W W W C C C C $$ | C C C C C C C[/go]

The 7.1-example of Nihon Kiin 1989 (with changed colors) assumes that White's outside group is alive, not showing the content of the circled area.

There are later examples, in which Bent-Four is alive in Seki or even alive with territory. This is due to the existence of some "local" Ko-treats.

Any unconditionally alive group (i. e. with two eyes) borders "local". No move beyond these "borders" does have any effect on the evaluation of life and death.

Click Here To Show Diagram Code: [go]$$W $$ -------------------- $$ | O O O . X O . O C $$ | . X X X X O O O C $$ | X X O O O O . O C $$ | O O O C C O O O C $$ | C C C C C C C C C[/go]

Giving the outside group 2 eyes would have been better. Now it is really uninteresting, how the circled area looks like.

Click Here To Show Diagram Code: [go]$$W $$ -------------------- $$ | O O O 2 X O . O . $$ | 1 X X X X O O O . $$ | X X O O O O . O . $$ | O O O . . O O O . $$ | . . . . . . . . .[/go]

White can take Black's stones off the board, ...

Click Here To Show Diagram Code: [go]$$W $$ -------------------- $$ | 4 3 7 X X O . O . $$ | 5 X X X X O O O . $$ | X X O O O O . O . $$ | O O O . . O O O . $$ | . . . . . . . . .[/go]

6 = Pass or Tenuki
... independent from what happens in the rest of the board

Click Here To Show Diagram Code: [go]$$W $$ -------------------- $$ | W W W . Z W . W . $$ | . Z Z Z Z W W W . $$ | Z Z W W W W . W . $$ | W W W . . W W W . $$ | . . . . . . . . .[/go]

Returning to the original position, we have Black stones, which can be captured, between White stones, which cannot be captured (= are said to be "alive"). Concluding, Black's group is dead as it stands.

As you have written already, status assessment does not change the score (so it does not matter, if something may have been captured in the process). The results "cannot be captured" / "can be captured" are combined using the original position after the end of "play".

RobertJasiek · **#39**

wms, after the filling of all teire, providing "having the server step the users through the resolution at the end of the game" is incredibly easy (not complicated, as you think) if you base it on the players' agreement:

1) Pressing Done invokes scoring iff both players have pressed Done successively without any intervening marking of a string or Undo.

2) Undo during the agreement phase requires the opponent's confirmation.

3) Optional: When a string is being marked, automatically mark also all same-coloured strings within the same empty-or-same-colour-string-region.

Kirk, with fake false sekis eyes I refer to a circular shape like in http://home.snafu.de/jasiek/j_verbal_status.pdf

Cassandra, repeating "Japanese style rules do not have global play in analysis" many times does not make it correct. - What is being analysed is local with respect to strings. How it is being analysed allows for global play.

jts, bent-4 belongs to the rules forum. In short, different Japanese rulesets treat it differently.

Cassandra · **#40**

RobertJasiek wrote:

Cassandra, repeating "Japanese style rules do not have global play in analysis" many times does not make it correct. - What is being analysed is local with respect to strings. How it is being analysed allows for global play.

It's just a matter of taste. And doesn't really matter.

You cannot win a Semeai with moves, which are too far away. You have to play near by (= "local"), so that your moves can have an effect on the Semeai.

It may be possible that you receive compensation elsewhere on the board, when playing non-local, may be that you even win in this kind of exchange, because your moves gain more than those of your opponent. But you will not win the Semeai.

So the view on a Semeai is "local", not "global". It is possible to exclude part of the board from it's investigation.

The same is true for the status evaluation (typical Japanese style) of a single string of stones.

Do Japanese Rules ( or their KGS implementation ) suck ?

Who is online