Life In 19x19
http://www.lifein19x19.com/

No result game without loop (in japonese rule) ?
http://www.lifein19x19.com/viewtopic.php?f=45&t=18275
Page 1 of 6

Author:  Gérard TAILLE [ Fri Jul 16, 2021 1:10 pm ]
Post subject:  No result game without loop (in japonese rule) ?



Assume komi = 0.5
If the game ends in this position and if the 110 black stones in the upper left corner are considered dead then white wins by 0.5 point.
That means that if, in normal play, white takes effectively these 110 black stones then black is sure to win (because by taking the black stones white plays in her territory and looses one point).
As a consequence, in the position above white has no choice : white passes and the referee has to decide who is winning.
What is the problem. Nobody knows if black can live in the upper left corner (after white makes the capture) and more over nobody knows if black can live under hypothetical play!
Maybe we can conclude to a no result by ignorance (and not due to a explicit loop) ?

Author:  Bill Spight [ Fri Jul 16, 2021 1:47 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

See https://senseis.xmp.net/?10x10CornerGame1 . :)

Author:  Gérard TAILLE [ Sat Jul 17, 2021 1:47 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Thank you for the link Bill. It allows me to give you a more convincing position:

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


If white stones are dead then the black score is (17 * 17 * 2) + 5 = 583 points
Assuming now that 582 black stones have been captured, and assuming a komi 0.5, then the result is a black win by 0.5 points.
That means that black must avoid to capture at "a" because she will loose one point and the game (white will simply pass without trying to live)
Both players agree to stop the game and now the confirmation phase begin. The problem is now the following : nobody knows the god sequence saying if white stones are really dead under hypothetical play. The result of the game is simply unkonwn because the humans are not strong enough! BTW I doubt the current computers are able to give us the correct result.

Author:  luigi [ Sat Jul 17, 2021 3:35 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Your last diagram is the perfect summation of everything that's wrong with Japanese-style rules. :bow:

Author:  RobertJasiek [ Sat Jul 17, 2021 3:54 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

For further mistakes in Japanese rulesets and their corrections, see

http://home.snafu.de/jasiek/j1989c.html
http://home.snafu.de/jasiek/wagcflaw.html
http://home.snafu.de/jasiek/j_verbal_status.pdf
http://home.snafu.de/jasiek/j2003.html
http://home.snafu.de/jasiek/j2003com.html
http://home.snafu.de/jasiek/j2003inf.html
http://home.snafu.de/jasiek/wagcmod.html
http://home.snafu.de/jasiek/wagcinf.html
http://home.snafu.de/jasiek/wagci.html
http://home.snafu.de/jasiek/sj.html
http://home.snafu.de/jasiek/rules.html

Author:  Gérard TAILLE [ Sat Jul 17, 2021 7:29 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

RobertJasiek wrote:


What an impressive list Robert. Congratulations.
Did you already mentionnend my point (humans may be not strong enough to give the correct result of an hypothetical play)?

Author:  RobertJasiek [ Sat Jul 17, 2021 7:52 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Probably in 1997 on rec.games.go :) When I asked one of the two J1989 authors about it in 2006, he admitted the problem then suggesting a time limit.

Author:  jann [ Sat Jul 17, 2021 8:03 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Gérard TAILLE wrote:
The result of the game is simply unkonwn because the humans are not strong enough!

But this is only a practical limitation, not a theoretical flaw in the rules. Such cases and anomalies - which exist for almost any ruleset, including Japanese OC - seem much worse.

Author:  Gérard TAILLE [ Sat Jul 17, 2021 8:22 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

jann wrote:
Gérard TAILLE wrote:
The result of the game is simply unkonwn because the humans are not strong enough!

But this is only a practical limitation, not a theoretical flaw in the rules. Such cases and anomalies - which exist for almost any ruleset, including Japanese OC - seem much worse.


Yes Jann it is a practical limitation but practice is important isn't it?
An ideal rule in practice is a rule which allows the referee to get the correct result of a game even if the referee is a beginner (providing she understands all the articles of the rule).

As an example:

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

For an experimented player it is quite easy to see which stones are dead. But I think a beginner knowing perfectly the rule, will not be able to give the correct result without an help from an experimented player.
Japonese rule is probably perfect for good players but maybe a nightmare for beginners. That does not mean japonese rule is not correct but certainly it is not an ideal rule.

Author:  jann [ Sat Jul 17, 2021 8:39 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Gérard TAILLE wrote:
That does not mean japonese rule is not correct but certainly it is not an ideal rule.

Sure, but if the ideal rule would exist it would be widely known and dominant by now. Generally, you get to choose between problems and complications before or after game stop. Japanese is the simplest before stop (and one of the worst for after stop issues).

Author:  Pio2001 [ Sat Jul 17, 2021 11:12 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

jann wrote:
Gérard TAILLE wrote:
The result of the game is simply unkonwn because the humans are not strong enough!

But this is only a practical limitation, not a theoretical flaw in the rules.


Hi,
The main drawback of such a limitation is that the rule is not programmable. A software can't always tell who's the winner of a game. This is a major problem for deep learning software, which need to assess who's the winner of thousands of self-play games (some of them ridiculous, when they start self learning from zero) without any human intervention.

jann wrote:
Gérard TAILLE wrote:
That does not mean japonese rule is not correct but certainly it is not an ideal rule.

Sure, but if the ideal rule would exist it would be widely known and dominant by now.


If we look at official rules, Korea is the only country besides Japan that still have an official rule in the Japanese style. All others that have taken the time to write a complete rule are in the Chinese style : China, New Zealand, SST, AGA, France, UK.
Among competitions with a separate official rule set, one use a Japanese style rule (World Amateur Championship), the other use a Chinese style rule (World Mind Sports Games).

What's interesting is that I don't know any example of a federation or competition that would have switched from a Chinese style rule to a Japanese style rule. They all do the opposite.
And the AGA style (Chinese scoring counted with pass stones) seems to be popular in the recent years (adopted by UK in 2006 and by the European Congress in the late 2010').
The UK rules text is one of the shortest and clearest.

Author:  gennan [ Sat Jul 17, 2021 11:29 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

If area scoring is really superior to territory scoring, then why can't China, New Zealand, SST, AGA, France, UK agree on one "perfect" rule set?

Author:  jann [ Sat Jul 17, 2021 11:53 am ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Pio2001 wrote:
This is a major problem for deep learning software, which need to assess who's the winner of thousands of self-play games

The problem in that case handling the simple, ordinary positions, not rare beasts. ML is robust enough even if you'd randomize the result of 0.01% of games.

And again: problems/complications before stop vs after stop. The optimal choice is different for avg players (always in agreement) and beginners or bots.

Quote:
complete rule are in the Chinese style : China, New Zealand, SST, AGA, France, UK.

Western rules tend to be beginner-oriented, I wouldn't explain deeper meaning into this.

Author:  RobertJasiek [ Sat Jul 17, 2021 12:10 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

gennan wrote:
If area scoring is really superior to territory scoring, then why can't China, New Zealand, SST, AGA, France, UK agree on one "perfect" rule set?


The issue of unification for a scoring system is not one of superiority. Area scoring rulesets are split, territory scoring rulesets are split etc.

Instead it is a matter of political willingness to unify. Thus far, most rule-making has been made by national associations with only or almost only one association's internal view.

The greatest attempt of unification was the International Go Rules Forum. The Japanese and Koreans did not want to unify so the main purpose of the forum was lost. Afterwards, we tried to unify at least area scoring rulesets. However, the remaining open issue were ko rules, the forum was initiated by Ing people and they do not want to completely abandon Ing Ko Rules. Hence, there was not enough good will from enough sides to unify.

The IGF is another body which might unify rules. However, the IGF is still dominated by Japanese-style rules traditionalists who rather want to keep WAGC rules than get anything better. Those who want change need to apply as IGF directors and replace people like Yuki Shigeno (a hard core Japanese-style rules traditionalist).

Another relevant body is the EGF. Its choice of rulesets has been driven by tolerance, sponsorship and clarification as far as possible within this scope of compromise. The EGF rather uses several rulesets instead of unifying them into one.

Nevertheless, change has occurred (and I have played a role there especially towards clarification as far as politically possible) but it is all subject to politics so change progresses slowly like a molasses ko.

Author:  Pio2001 [ Sat Jul 17, 2021 12:13 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

gennan wrote:
If area scoring is really superior to territory scoring, then why can't China, New Zealand, SST, AGA, France, UK agree on one "perfect" rule set?


EDIT : cross-post with Robert Jasiek.

The fact that they use different rules doesn't imply that they belong to a category (Chinese style) that is worse than the other category (Japanese style).

I would indeed support the idea to adopt a universal rule of go.
Historically, it seems to me that each new rule brings an improvement over the previous ones (I mean rules officially used in tournaments).
I think that Chinese style rules without pass stones won't become universal because it is natural to count territory minus prisoners minus dead stones instead of area. The problem is the "dead stones" that are not yet prisoners, and that cause nearly insolvable problems in territory style rules.
AGA style rules seem popular, as they were adopted three more times after they were created (France, UK, European Congress).

In this category, the three rules officially used in tournaments are AGA, French, and British rules.
French and British rules are shorter than the original AGA rule. This is a step in the right direction to become universal. They still have minor flaws each, but I think that they are close to perfect.

Author:  RobertJasiek [ Sat Jul 17, 2021 12:14 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

jann wrote:
Western rules tend to be beginner-oriented, I wouldn't explain deeper meaning into this.


There is more meaning involved, such as
- different views on which are the best details (suicide or no suicide, which superko variant, how tolerant should the agreement phase be),
- specific language needs after translations,
- historical context of creation.

Author:  RobertJasiek [ Sat Jul 17, 2021 12:22 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Pio2001 wrote:
it is natural to count territory minus prisoners minus dead stones


Haha, good one;) The natural thing is either what the New Zealanders do or fill until a player can't fill any more:)

Author:  Pio2001 [ Sat Jul 17, 2021 12:36 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

RobertJasiek wrote:
Nevertheless, change has occurred (and I have played a role there especially towards clarification as far as politically possible) but it is all subject to politics so change progresses slowly like a molasses ko.


Thank you for these interesting infos !

Author:  Pio2001 [ Sat Jul 17, 2021 12:47 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

RobertJasiek wrote:
Pio2001 wrote:
it is natural to count territory minus prisoners minus dead stones


Haha, good one;) The natural thing is either what the New Zealanders do or fill until a player can't fill any more:)


Oops, I didn't see the irony of my sentence.
What I was meaning is that on a real board, it is more convenient to count the final score in the Japanese style, and endgame values are used to be counted in territory.
I don't think that go players would accept to stop counting using prisoners and adopt a pure area counting method, like we can see here in the Ke Jie vs Alphago match (counting performed from 4:52:11 to 4:54:35 in the video) : https://youtu.be/Z-HL5nppBnM?t=17529

Author:  Cassandra [ Sat Jul 17, 2021 1:00 pm ]
Post subject:  Re: No result game without loop (in japonese rule) ?

Gérard TAILLE wrote:
Thank you for the link Bill. It allows me to give you a more convincing position:

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


If white stones are dead then the black score is (17 * 17 * 2) + 5 = 583 points
Assuming now that 582 black stones have been captured, and assuming a komi 0.5, then the result is a black win by 0.5 points.
That means that black must avoid to capture at "a" because she will loose one point and the game (white will simply pass without trying to live)
Both players agree to stop the game and now the confirmation phase begin. The problem is now the following : nobody knows the god sequence saying if white stones are really dead under hypothetical play. The result of the game is simply unkonwn because the humans are not strong enough! BTW I doubt the current computers are able to give us the correct result.


Even assuming that White did NEVER intend playing the double atari of a (to win), it is ABSOLUTELY IMPOSSIBLE to reach that position by alternating play (NO passes in between).

It is EVEN THE MORE IMPOSSIBLE that White could have generated a SURPLUS of 582 prisoners. Simply because White's stones will have to remain on the board (assuming that we will generate Black prisoners only) after having captured Black ones.

You cannot "prove" a supposed weakness of a rule set by using an example that can be reached only by very young kids putting Black and White Smarties onto the go board.

Page 1 of 6 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/