Yesterday's rule dispute in Korea

[jaeup]

Another fun point: the players have so many prisoners that the final score is black 2, white -4. After 6.5 komi, white wins by 0.5.

Sorry,
Is there a way to get the video for the full game or the SGF?
Thanks. Take care.

I don't think the full game video is on the internet. Only the short clips yoyoma uploaded are available now.

I uploaded the sgf file with the correct order in a previous message, and that should be enough to follow the game.

Unfortunately, non-Korean speakers will have trouble understanding what mistakes both players made at the end of the game. Simply speaking, Black threw away the game and White almost threw it back but managed to save it with the help of the rule allowing her not to make a reinforcement. (This video https://www.youtube.com/watch?v=FSo-zFNYyuk&t=1s explains many other possible moves, but only in Korean.)

[jaeup]

Hi,
In my opinion, the rules of go should fulfill three conditions :

1-Novice people who read them should be able to play a game and score it without help of a more experienced player.

2-The rules of play should be translated and published by national federations worldwide, and referees should be able to use these translations and score games in any tournament without dispute.

3-The rules of go should be programmable, so that we can have software playing go.

I know most people wish similar things, but they usually miss one important condition.

4. The rule should be free from anomalies.

Well.. what is anomaly? It is a subjective term which means one cannot define it rigorously.

However, it is a serious issue to make a rule acceptable for the professional players. I bet they will never accept a rule that a separated moonshine life ends up with a life or a rule allowing one to have a benefit from send-two-receive-one trick, even if such a thing happens rarely. They rather prefer a referee to come to the player and say "Hey, your claim is disgraceful, and there is no way for me to accept it. Give up the game now, or you will be in a serious trouble in your future career."

I guess it is one reason why the Eastern Asian rules are written in such an ambiguous way. They know the rule text is often contradictory, but they want to leave it as it is, and interpret the rule text the way they prefer. i.e. they are ready to abandon the literal meaning of the rule text if it is inevitable to prevent an abuser to try an anomalous game end.

I am not defending them here, but I am saying that it is really important to understand why they act so weirdly in their rule writing, if you want to make a ruleset acceptable for them.

[Ferran]

I uploaded the sgf file with the correct order in a previous message, and that should be enough to follow the game.

That's the one named "temp.sgf"? Sorry, I hadn't connected both.

Thanks anyhow. Take care.

[Edited to delete questions that were already answered; my apologies, I'd assumed the videos where from a different game witha similar issue; brain gasses]

[Bill Spight]

Hi,
In my opinion, the rules of go should fulfill three conditions :

1-Novice people who read them should be able to play a game and score it without help of a more experienced player.

2-The rules of play should be translated and published by national federations worldwide, and referees should be able to use these translations and score games in any tournament without dispute.

3-The rules of go should be programmable, so that we can have software playing go.

I know most people wish similar things, but they usually miss one important condition.

4. The rule should be free from anomalies.

Well.. what is anomaly? It is a subjective term which means one cannot define it rigorously.

Indeed. Is ending the game with an unfinished ko because of a ko ban anomalous or not? Shusai and Go Seigen thought not, but the Nihon Kiin 1949 rules forbade it. Is Three-Points-Without-Capturing anomalous? Berlekamp and Wolfe showed that theory supports that evaluation, but the Nihon Kiin 1989 rules do not allow it. Ing regarded as anomalous the fact that two double ko death positions can combine to produce a single superko, so that one position can live. (It is similar to Moonshine Life, where a double ko can be used as a ko threat to prevent capturing the group.) However, even before I developed the theory of multiple ko evaluation, I could show that there are positions that are not themselves double ko deaths that can break the double ko. From the standpoint of theory, it is the single double ko death that is exceptional.

However, it is a serious issue to make a rule acceptable for the professional players. I bet they will never accept a rule that a separated moonshine life ends up with a life or a rule allowing one to have a benefit from send-two-receive-one trick, even if such a thing happens rarely. They rather prefer a referee to come to the player and say "Hey, your claim is disgraceful, and there is no way for me to accept it. Give up the game now, or you will be in a serious trouble in your future career."

I guess it is one reason why the Eastern Asian rules are written in such an ambiguous way. They know the rule text is often contradictory, but they want to leave it as it is, and interpret the rule text the way they prefer. i.e. they are ready to abandon the literal meaning of the rule text if it is inevitable to prevent an abuser to try an anomalous game end.

This is the way most court systems work in the real world. Real legislation is often ambiguous and sometimes contradictory. Courts make rulings only as necessary, and the rulings themselves may not be logical.

The short history of written go rules shows that attempts to eliminate anomalies can introduce complications and ambiguities, and can even produce new and different anomalies down the road. Under the J89 rules it can be correct to leave a group with two eyes on the board with liberties as a new kind of seki. Anomaly or not? To try to avoid very rare but perceived anomalies Ing introduced Fighting Kos and Disturbing Kos, a distinction which he never managed to demonstrate logically. There are kos which could be either. Disturbing kos can be very disturbing.

OTOH, Ing's 4 pass rule, designed to prevent anomalies, is actually a very beautiful way to deal with end of game issues. However, most players do not understand it, regarding the 3d and 4th passes as mere formalities. Since the difficulties which the rule addresses occur infrequently, that lack of understanding is rarely a problem. Still, the rule itself has produced a rules dispute.

Now, I am an advocate of Button Go, and it has been used at least once in international competition, but not under that name. I doubt if the organizers had even heard of button go. With one simple alteration AGA rules can implement button go. But human beings are quite good at having rules or principles with exceptions. Despite being criticized as illogical, the Japanese 1949 rules were onto something. Instead of having difficult to understand and apply rules to prevent infrequent anomalies, have easy to understand and apply rules, with any known anomalies listed as exceptions. And if theorists come along later with rationales for the rulings on anomalies, so much the better.

[Pio2001]

However, it is a serious issue to make a rule acceptable for the professional players. I bet they will never accept a rule that a separated moonshine life ends up with a life or a rule allowing one to have a benefit from send-two-receive-one trick, even if such a thing happens rarely.

Professional players in China seem to be ok with chinese rules.

Are there some examples of moonshine life, or send-2-receive-1 benefit, with positional superko ?

I guess it is one reason why the Eastern Asian rules are written in such an ambiguous way. They know the rule text is often contradictory, but they want to leave it as it is, and interpret the rule text the way they prefer. i.e. they are ready to abandon the literal meaning of the rule text if it is inevitable to prevent an abuser to try an anomalous game end.

But this would prevent computers to play go, in contradiction with the third requirement.
I would prefer having strict rules, with some extremely rare oddities such as triple ko or molasses ko.

Triple ko doesn't bother me. With the positional or situational superko rule, it just behaves as a regular ko.
Molasses ko, on the other hand, is indeed a monster. Let computers play molasses ko until the end, and keep the chinese option : in real life, if needed, the referee may decide what happens is case of molasses ko.

According to the statistics in the Go Player's Almanach, triple and quadruple ko occurs in one game out of 10,000 (quadruple kos are as frequent as triple kos).
Molasses ko and Round-robin ko have never been reported in 100,000 games.

The simplest form of Button Go has a token called a button, which is worth ½ pt. by area scoring. At her or his turn a player may take the button instead of making a play on the board. Taking the button lifts a ko or superko ban, just as a board play does. Normally the button is taken after the last dame is filled and has the effect that it does not matter who gets the last dame.

If I understand correctly, Button go uses area scoring. Which means that, in it's simplest form, it forces the players to count all the area. That will be difficult to swallow.

AGA style Button Go:

You can implement Button Go with AGA pass stones by making the last pass special. If the player to make the last pass also made the first pass, she or he does not have to hand over a pass stone. White does not have to make the last pass. Because of that, we may consider that Button Go rule as a simplification of AGA rules.

I don't understand : you mean that at the end, there won't be an equal number of black and white stones on the board ?
In this case, the AGA rule doesn't work. You can't have the score defined by area and fill territory with prisoners if the total number of black and white stones are not equal (except in handicap games).
Or am I missing something ?

[Bill Spight]

The simplest form of Button Go has a token called a button, which is worth ½ pt. by area scoring. At her or his turn a player may take the button instead of making a play on the board. Taking the button lifts a ko or superko ban, just as a board play does. Normally the button is taken after the last dame is filled and has the effect that it does not matter who gets the last dame.

If I understand correctly, Button go uses area scoring. Which means that, in it's simplest form, it forces the players to count all the area. That will be difficult to swallow.

This form of button go, with a token, uses area scoring. Which is one of the options of AGA rules.

AGA style Button Go:

You can implement Button Go with AGA pass stones by making the last pass special. If the player to make the last pass also made the first pass, she or he does not have to hand over a pass stone. White does not have to make the last pass. Because of that, we may consider that Button Go rule as a simplification of AGA rules.

I don't understand : you mean that at the end, there won't be an equal number of black and white stones on the board ?

Not necessarily.

In this case, the AGA rule doesn't work. You can't have the score defined by area and fill territory with prisoners if the total number of black and white stones are not equal (except in handicap games).
Or am I missing something ?

Well, as the guy who introduced the idea of pass stones in the AGA journal in the 1970s, I know what I'm doing. (BTW, I was probably not the first person to come up with the idea of pass stones; I was not the only one, OC.) This method requires AGA territory scoring, not area scoring, which requires a token or other adjustment. But, IIUC, most players use territory scoring anyway, and the idea of requiring White to pass last has met with confusion and opposition.

Button go is a hybrid between area and territory rules. This is accomplished by requiring an equal number of pass stones or other stones played after the button instead of an equal number of stones on the board for scoring. Basically, before the button the game is played by territory rules, and afterwards, which is where nearly all anomalies and other difficulties arise, it is played by area rules, by which it is normally easy to resolve life and death and kos. This has little effect on strategy, because, with few exceptions, correct play by this form of territory scoring is also correct by area scoring. Button go is not area go by another name, nor is it a form of territory go, but a combination of both.

Edit: Using area counting with the ½ point token has the effect of using territory rules before the button because it means that who plays the last dame does not matter, as a rule.

[jaeup]

The short history of written go rules shows that attempts to eliminate anomalies can introduce complications and ambiguities, and can even produce new and different anomalies down the road.

Yes, I agree. The first thumb rule I learned from my rule research was this: "Every rule has anomalies".
The second thumb rule is "If you add one sentence in the rule to eliminate one anomaly, it almost certainly creates a new anomaly.

Now, I am an advocate of Button Go, and it has been used at least once in international competition, but not under that name. I doubt if the organizers had even heard of button go.

I am a supporter of the pass stone and button go, or any other variants sharing the same philosophy. I often see pros and amateurs saying that "The Chinese rule itself is OK, but I just can't bear 7.5 komi. Unfortunately, 6.5 komi is unacceptable." They are not true rule beginners (The true beginner will simply insist applying 6.5, not knowing its problem.), but I am so surprised that none of them really know that a simple modification enables the introduction of 6.5 komi. (and the players in the Chinese tournament do not need to wait for the referee to do the counting.)

Well, as the guy who introduced the idea of pass stones in the AGA journal in the 1970s, I know what I'm doing. (BTW, I was probably not the first person to come up with the idea of pass stones; I was not the only one, OC.)

When writing my book, I always had trouble identifying who did something first. I will at least add one sentence mentioning that you suggested the introduction of the pass stone in the AGA rule.

Regarding this issue, old Chinese records with group tax are interesting. It seems that they force White to make the last move, which suggests that their scoring method might have been close to the current AGA territory scoring. (Of course, minus the group tax.) A clever person may have developed the idea of pass stone from that experience, because filling in your territory is practically the same thing as giving a prisoner to your opponent, especially for the last move of the game. Of course, I don't have any more clues on it.

[jaeup]

Professional players in China seem to be ok with chinese rules.
Are there some examples of moonshine life, or send-2-receive-1 benefit, with positional superko ?

whites_trap.png (150.72 KiB) Viewed 9709 times

This is one well-known situation. For all three types of superko rules, Black fails to protect the upper group. When I demonstrated it in front of 40 pros, they simply laughed. In their mind, "PSK prohibits Black to play at X" is an unacceptable flaw of the rule.

Why are Chinese players fine with the current rule? As I mentioned, they are ready to ignore the written text of the rule. Here is a sarcastic summary of the Chinese rule. If a whole board repetition(WBR) occurs
1) sometimes the game ends with a draw
2) sometimes the player who made WBR gets the penalty
3) sometimes the opponent gets the penalty (or at least the game continues)

Black 10 at X makes a clear WBR, but they know Black should not be penalized. Thus, a human referee will come up to choose 3. Is there a clear algorithm distinguishing cases 1, 2 and 3? The answer is No.

According to the statistics in the Go Player's Almanach, triple and quadruple ko occurs in one game out of 10,000 (quadruple kos are as frequent as triple kos).
Molasses ko and Round-robin ko have never been reported in 100,000 games.

Considering how many draws Lee Sedol has made in his career, it definitely depends on one's strength. (and of course on one's play style) Players who know how to make a draw (and not to lose a game) tend to make it more often.

[Bill Spight]

Well, as the guy who introduced the idea of pass stones in the AGA journal in the 1970s, I know what I'm doing. (BTW, I was probably not the first person to come up with the idea of pass stones; I was not the only one, OC.)

When writing my book, I always had trouble identifying who did something first. I will at least add one sentence mentioning that you suggested the introduction of the pass stone in the AGA rule.

Then let me be clear. I wrote a very short article explaining the current Taiwan (Ing) rules, which had a superko rule. At the end I suggested that people who wanted to try out those rules, but were used to Japanese rules, could use pass stones (which I called bookkeeping stones) with White passing last, so that they could count territory as they usually did. It was only much later that the AGA rules were formulated and used pass stones. AGA president Terry Benson was the editor of the AGA Journal when they printed my article (summer 1977, IIRC). I have not asked him, but he may have suggested the use of pass stones when they were formulating the rules. But, as I said, other people have certainly come up with the idea, so the AGA could have gotten the idea from several sources.

Regarding this issue, old Chinese records with group tax are interesting. It seems that they force White to make the last move, which suggests that their scoring method might have been close to the current AGA territory scoring. (Of course, minus the group tax.) A clever person may have developed the idea of pass stone from that experience, because filling in your territory is practically the same thing as giving a prisoner to your opponent, especially for the last move of the game. Of course, I don't have any more clues on it.

I disagree with Chen Zuyuan on that point. John Fairbairn published a translation of some of Chen's writings on the history of weiqi, which was available in GoGoD. (I don't know if it still is.) The oldest known description of weiqi is unclear. It talks about putting stones on the board with "two overflowing" or "both overflowing". That strongly suggests a form of stone scoring, and to some suggests a group tax. (I disagree. If you don't already know about the group tax, how do you get the idea from that phrase?) However, the oldest known scored games (centuries later) have scores for each player that are not only smaller than scores by any form of stone scoring, they are smaller than modern territory scoring by the amount of the group tax. (Again, if you didn't already know about the group tax that fact would be puzzling.) The few scored game records we have also indicate that each player made the same number of plays. But you cannot derive pass stones from that fact, because all of the games end with dame unfilled. Edit: In addition, you have to come up with the idea of keeping captured stones and using them to fill in territory for counting. Edit2: Removed further speculation.

Chen believes that territory scoring with a group tax derived from stone scoring with a group tax, via the use of pass stones. I see no evidence of that. In addition, I knew, long before hearing about ancient Chinese territory scoring, that no pass baduk produces territory scoring with a group tax. That is true for simple no pass baduk and also for no pass baduk with prisoner return, a form proposed by the late Professor Berlekamp. In fact, in Mathematical Go, Berlekamp proposes a special encore to allow the eyes necessary for life to be filled, which eliminates the group tax. A simple example of territory scoring with a group tax is the Capture Game, with a no pass rule. Once all the dame are filled, the players have to start filling in their territory, and have to leave the points necessary for life unfilled. Instead of filling in territory, they could stop play by agreement and determine the winner by counting territory with a group tax.

[Bill Spight]

Professional players in China seem to be ok with chinese rules.
Are there some examples of moonshine life, or send-2-receive-1 benefit, with positional superko ?

whites_trap.png

This is one well-known situation. For all three types of superko rules, Black fails to protect the upper group. When I demonstrated it in front of 40 pros, they simply laughed. In their mind, "PSK prohibits Black to play at X" is an unacceptable flaw of the rule.

{snip}
Why are Chinese players fine with the current rule? As I mentioned, they are ready to ignore the written text of the rule. Here is a sarcastic summary of the Chinese rule. If a whole board repetition(WBR) occurs
1) sometimes the game ends with a draw
2) sometimes the player who made WBR gets the penalty
3) sometimes the opponent gets the penalty (or at least the game continues)

Black 10 at X makes a clear WBR, but they know Black should not be penalized. Thus, a human referee will come up to choose 3. Is there a clear algorithm distinguishing cases 1, 2 and 3? The answer is No.

Emphasis mine.

Well, Ing rules treat this Sending-Two-Returning-One sequence as a disturbing ko, and prohibit White from repeating it. Ing does not use a superko rule by that name, OC. Spight rules (version 1) do allow the use of a superko rule. However, passes lift ko and superko bans. So under those rules after Black passes, lifting the superko ban. We start from the current position, with no ban on previous positions. Then, with a positional superko rule, White can play ("sending two"), can capture the two stones, but then ("returning one") would reproduce the position after the Black pass, which a positional superko rule would prohibit. A situational superko rule would allow it, but then Black could pass again, ending play by passing twice in the same whole board position. Kee rules also take care of such sequences. ( https://senseis.xmp.net/?KeeRulesOfGo ).

[RobertJasiek]

Not all professionals have the same opinion on anomalies; there are also those saying that rarities should not be over-emphasised.

Those professionals rejecting shape anomalies or special case ruling every shape by another exceptional rule or arbitrary decision claim that there should not be anomalies but they hide that they insist on these systematic anomalies in the rules:

1) different rules for alternation phase and analysis phase,

2) global dependence of life status and global ko threats allowed during the alternation phase but local dependence of life status and (sort of) local ko threats allowed (or made effective) during the analysis phase.

There is no justification for special treatment of shape anomalies while there is no explanation for requiring systematic anomalies in the rules.

Instead of fulfilling traditionalists' wishes, a unified ruleset that shall also make them happy must also have their explanation and dissolution of systematic anomalies in the rules.

E.g., it is all fine and well to modify area scoring rules so as to remove anomalies related to pass or ko rules, or remove strings with 1 liberty by adding a rule that the game may not end until there are none. However, modifying territory scoring rules (of the traditional kind) to eliminate all anomalies seems impossible.

[jaeup]

E.g., it is all fine and well to modify area scoring rules so as to remove anomalies related to pass or ko rules, or remove strings with 1 liberty by adding a rule that the game may not end until there are none. However, modifying territory scoring rules (of the traditional kind) to eliminate all anomalies seems impossible.

As a Korean, my first try was to modify the existing territory scoring rule. My temporary conclusion is the same as yours. Making a anomaly-free ruleset with a programmable logic seems to be impossible, and what is worse is that proof of impossibility is also impossible because of the ambiguity of the concept "anomaly".

I think modifying area scoring rule to fulfill this purpose is also a tough job. I have seen ideas floating here and there, but they are rarely assembled to form a complete ruleset. (Again, because nobody defined what "anomaly" is, whether a ruleset is anomaly-free is a subjective judgment.)

[Matti]

A player says there is a rules anomaly, when he does not understand why he lost.

[Matti]

Professional players in China seem to be ok with chinese rules.
Are there some examples of moonshine life, or send-2-receive-1 benefit, with positional superko ?

whites_trap.png

This is one well-known situation. For all three types of superko rules, Black fails to protect the upper group. When I demonstrated it in front of 40 pros, they simply laughed. In their mind, "PSK prohibits Black to play at X" is an unacceptable flaw of the rule.

This situation is based on the fact that after white 9, black has ran out of moves. On 19 * 19 board it wuld be very unlikely to happen. If one explain this to the professionals, maybe the would have a different opinion.

[Bill Spight]

I submit that there are at least two forms of baduk without anomalies. Both are forms of territory baduk.

The first is the Misere Capture Game. The first player who captures a stone or stones loses. Suicide is not allowed. (We could allow suicide to win, but that's a different game.) A player whose turn it is but who cannot make a legal play also loses. Obviously, this is a form of No Pass Baduk. As with all known forms of no pass baduk, it has a form of territory. Actually, it has two forms of territory. One is a single point eye such that filling it is not suicide. The other is a point that connects a string in atari to another string, such that the connection is not suicide. This second form of territory is nothing like territory in regular go, but deserves the name. Why? Because it belongs to one player and not the other. One player can play there, the other cannot. Also, this game has a group tax. Consider a group with two one point eyes. It does not have two points of territory, because the player can fill only one eye, filling the second would be suicide. In this case the group tax is only one point. There are also sekis, positions with one point such that neither player can play there because doing so would capture the other player's string or be suicide. Sekis in this game have one dame instead of two.

OC, a regular baduk player may consider the forms of territory in this game, the form of seki, and the group tax to be anomalous. To which the answer is that they are characteristics of Misere No Pass Baduk without Suicide, not anomalies. This is the attitude of John Tromp about Tromp-Taylor rules and Shusai and Go Seigen about "unfinished" kos at the end of play. The rules lead to such positions, what is your problem?

The No Pass Capture Game is more familiar. The first person to capture a stone or stones wins. Territory in the capture game is familiar, but it has a two point group tax. Life is unfamiliar. Any eye with at least two points is alive, but if there is one or more opposing stones inside, it may be seki. Again, such features may be regarded as anomalous, but the answer is that they are part of the game.

AFAIK, anomalies in baduk involve at least one of the following: life and death, seki, ko (including superko), suicide, passes, territory, or scoring. That's a lot of ground to cover, and opinions differ. It seems impossible to please everybody. At the same time, I do not agree with John Tromp's attitude. Just saying, these are the rules, like it or not, is not enough, IMO. I think that Ing was on the right track. Derive the rules from principles, then if you agree with the principles, accept the rules and their results. The trouble with Ing rules is that when they produced results that seemed anomalous to Ing, he changed the rules and, IMO, was not able to square the new rules with his principles. There was a principle that was not available to Ing, that of evaluation. Evaluation is a general enough principle to include all of the above except suicide. It was only in 1998, with my paper on the evaluation of multiple kos and superkos, that the evaluation principle was strong enough and general enough to use as a basis for the rules of baduk.