strange seki

[John Fairbairn]

Now the pluperfect has virtually disappeared.

And the past historic in French is on the critically endangered list, too!

More and more, young people are rejecting the past. O tempora! O mores! Or in modern yob-speak: OMG.

[Pio2001]

The superko rule itself is also only an imperfect approximation of the real game.

What's the "real game" ? Since I'm playing in France, for me, the real game uses natural situational superko. That's the official rule.

But nowadays, superko is also necessary for artificial intelligence.

No it isn't. You just make void game a valid result.

Ok, but you need to decide a rule of some kind, that will define exactly when a game is void. Because the japanese style rules only states that it occurs when none of the players "is willing to stop the repetition", which is a subjective condition.
For example we can set a given number of repetitions a the limit. That's a "superko rule" in a broad sense.

I don't think anyone can reasonably argue even that superko is desirable, let alone necessary.

Why not ? With positional superko, a triple ko is exactly the same thing as a simple ko. You need to have enough ko threats in order not to be the one loosing the battle. Why is that bad ?

A void game has stranger consequences on the strategy : facing a choice between creating a double ko seki (A) or creating a triple ko (B), a player who is in a bad situation on the board would prefer the triple ko, allowing him to restart the game.
Why not letting him face the situation he's into and actually fight the triple ko ?

Triple ko.png (9.25 KiB) Viewed 13898 times

[John Fairbairn]

With positional superko, a triple ko is exactly the same thing as a simple ko. You need to have enough ko threats in order not to be the one loosing the battle. Why is that bad ?

Who said that was bad? It is the practical complexity of superko that is found to be unacceptable to nearly everyone.

You have a road. You have traffic that wants to travel along that road. So you makes rules. There are alternative rulesets. One ruleset is that everyone drives on the side nearest to the left, as in the UK. Other ruleset stipulates you all drive on the side nearest to the right, as in Europe. In both cases there are also restrictions on overtaking. Both rulesets work. Regrettably a few people do get killed. There are occasional other blips, such as roadworks and lane closures. But people put up with all that and don't change their rules. People even cross the English Channel with their cars and drive under different rulesets, and come home safely!!!!!

Then a mathematician comes along and says we don't need to do that. We can have a supertraffic rule. Everybody can drive and overtake where they like and so can make their journeys faster. It will be supersafe because all they have to do then is be supercareful, which people will be anyway because they don't want to be killed. That's been mathematically proven! This situation does actually exist in some countries, which are usually considered benighted. Next to nobody in the UK or Europe would therefore accept such a new rule, no matter how many supertraffic mathematicians jumped up and down in frustration, and no matter how many seconds of savings on journey times are claimed.

That's hyperbole, of course. But this sort of approach does apply in trivial situations, You see, even in the rather safe confines of a game of go, people - even pros - find it hard to be supercareful. They can't always be sure they play the kos in the right order, especially in fast games. Such mistakes have occurred in pro play. might they want to accuse a professional colleague of making a mistake, or even cheating. No doubt they would also be reluctant to divert a large portion of their earnings to pay for a cadre of referees who will just be called on once in a blue moon when a superko arises. Then you have the problem of whether the referees can stay awake waiting for such a rarity. Unlike major sports that haven't got the finances to install video replays, and no doubt they'd rather not have a digital record of picking their toenails (a la Sakata). In short, they have already decided to make life safe, sane and simple. They ignore superko. Japanese '89 rules are an abomination. OK, they ignore them.

There comes a point in most things in life when it's best if common sense just takes over.

And where common sense does not quite work, no problem: we just apply supercommon sense.

[jann]

What's the "real game" ?

Maybe what is played by most players, has closest ties to history and almost all pros agree with (wrt repetition rules)?

With positional superko, a triple ko is exactly the same thing as a simple ko. You need to have enough ko threats in order not to be the one loosing the battle. Why is that bad ?

I don't think there is anything wrong with inventing new rules and new variants. But given the wide consensus in Asia wrt triple ko handling, caution and modesty could not hurt either.

Besides the issue of applicability, one problem is taking away perfectly legal and desirable moves, forbidding players to defend themselves in some ko shapes. An unnatural and unnecessary change to the game, where the least of such artificial restrictions may be preferred.

[RobertJasiek]

It is the practical complexity of superko that is found to be unacceptable to nearly everyone.

Several aspects are involved WRT practical complexity.

1) Practical complexity of identification of repetition: it is similar for superko and void result ko rules because both require recognition of repetition of the whole board position. The latter is a bit relaxed because one need not recognise at / just before the first moment of repetition but players can be dull to recycle several times until both recognise repetition. This raises the additional question why players are supposed to take pride in being dull in a game in which they supposed to be bright.

2) Practical complexity in typical games: in almost all games (4999:5000, 19999:20000 or so), long cycles requiring superko consideration do not even occur so the practical complexity in typical games is extremely low.

3) Understanding why the practical complexity in typical games is extremely low: this requires understanding of the missing practical complexity of superko occurring in these occasional ko situations: sending-2-returning-1; double disturbing death (such as in a big nadare joseki), which, as under other typical ko rulesets such as void ko rules, provides an arbitary supply of ko threats for the dead player.

4) Practical complexity in the relatively most frequent (occurring in, say, 1:5000 or 1:20000 games) superko positions (triple ko or three kos behaving like a triple ko): there is no practical complexity because the ko fight should be exactly like a basic ko fight. There only is slight theoretical complexity, which is much less complex than your typical life and death problem of tactical reading: one must once understand why the ko fight should be exactly like a basic ko fight: due to otherwise occurring repetition, a ko threat sequence in a triple needs to end after, by choice of the ko threat player, 1, 3 or 5 successive plays in the triple ko.

5) Practical complexity in the second-most frequent (occurring in, say, 1:10000 or 1:40000 games) superko positions (quadruple, quintuple ko, moonshine life): now we are entering actual complexity in every one such shape because mindless play can be wrong; the actual move order and moment of next-move repetition matter. However, over all games, the practical complexity is extremely low because of the rarity of such shapes. In go, we take pride in the ability of thinking and in particular tactical reading. Such a rare shape with practical complexity in a game in which it occurs gives the players the chance to demonstrate their skills, which is an advantage compensating the disadvantage of the complexity of procedural handling in such a rare game.

6) Practical complexity in arcane shapes: in theory, superko strategy and tactics can become arbitrarily complex with extremely long cycles or ko fights. In practice, that is, for the sake of practical complexity, such is absolutely immaterial because such shapes (such as molasses ko) never occur (or at worst once in the entire world-wide history of go). Besides, there is the advantage that we have something to celebrate if indeed such a shape occurs.

They can't always be sure they play the kos in the right order, especially in fast games.

They also can't always be sure to play non-ko tactics in the right order, so what? Difficult move order in non-long-ko-cycle tactics matters many times per game while difficult move order in long-ko-cycles matters once every 1:10000th or 1:40000th game, so what?

Such mistakes have occurred in pro play.

Such mistakes in non-long-ko-cycle tactics in pro plays occur all the time, so what? Prohibit fast games?

Then you have the problem of whether the referees can stay awake waiting for such a rarity.

They even have to stay awake for relatively much more frequent incidents, such as self-atari, retracting a move or recapture of a basic ko.

Unlike major sports that haven't got the finances to install video replays, and no doubt they'd rather not have a digital record of picking their toenails (a la Sakata). In short, they have already decided to make life safe, sane and simple.

Wrong. That they make referee-handling simpler in 1:5000 or rarer long cycle cases is irrelevant as long as they make referee-handling more difficult in more frequent rules matters, such as not filling the last basic endgame ko.

Japanese '89 rules are an abomination. OK, they ignore them.

They ignore parts of them but apply other parts.

There comes a point in most things in life when it's best if common sense just takes over.

Common sense like "the simplest, shortest rule text ('a play may not repeat a position') is good enough in practice".

And where common sense does not quite work, no problem: we just apply supercommon sense.

Like going back to the position before the start of the long cycle and continue the game from there with carefulness and, if necessary, a few minutes of extra thinking time.

[John Fairbairn]

Robert: Everything you say is based on the premise you alone are logically right and everybody else is wrong. Common sense is based on neither right nor wrong but on what most people prefer. We use it because it a fundamental part of our DNA.

Common sense can be glaringly wrong, e.g. racial stereotyping. Then it is well worth trying to change the basic thinking. But it is incredibly difficult to change behaviour that is part of our DNA. If it is so difficult in such important cases, should we be getting worked up about trivial cases?

Common sense in go starts with the oft forgotten fact that it is just a game. I repeat: it is just a game. It is highly unlikely, therefore, that using our human predilection to apply common sense to all aspects of go is going to be in any way heinous. The game does not need self-appointed Einsteins to untangle its theory or the workings of its rules.

It is absolutely fine if, as an intellectual pursuit, budding Einsteins study go theory and rules. What is not fine is for those people to impose their mindsets on the bulk of players who prefer common sense. Report the results of the intellectual pursuit if you like. If those results are ignored, move on. Don't try to start a war. Learn to trust common sense. It is sometimes wrong. It is often right. It is always what most people prefer.

It is my view (having translated French rules into English as part of the BGA's decision to adopt AGA rules) that AGA rules failed on one simple issue: superko.

If the AGA had taken the decision to promulgate a ruleset that simply allowed people, as it were, to drive on different sides of the road according to the country they were in - Japanese-style counting or Chinese-style counting - and left it at that, people would have seen only the common-sense behind that view. After all, counting is implicit in every single game, and different countries are, well, different.

But as soon as superko was mentioned, they lost most of the "common-sense people". It was an instinctive reaction. Whether it's justified or not is irrelevant. People just don't like being told what to do by interfering outsiders. We are seeing this every day all around the world as people defy CV lockdown rules. In this case we need rules because several areas where we are used to applying common sense (save lives, save jobs, save mental health) overlap and are in conflict with each other. But even where we apply such rules, those imposing them strive to keep things simple ("hands, face, space"). That control too is an area where we apply common sense. And as soon as the vaccines arrive and work, what will be the first response: ditch the unnecessary rules.

Go is not even remotely near being something where areas of common sense conflict. It does not need extra rules. IT IS JUST A GAME.

[Bill Spight]

Common sense like "the simplest, shortest rule text ('a play may not repeat a position') is good enough in practice".

Preferring "the simplest, shortest rule text" is an esthetic choice. But most human communication is 50% redundant. Keeping it simple is a good idea, but not short. And then there is the question of the cognitive difficulty of what is essentially a bookkeeping task. A go board has 3³⁶¹ possible states, depending on whether a point is empty, occupied by a Black stone, or occupied by a White stone. (OC, not all of those states are legal, but there are a humungous number of potential positions to remember.) Who keeps track of that? We can, however, keep track of a local double ko with only 4 relevant states. Whether the rest of the board is the same or not, we can keep track of by the fact that plays there do not lead to repetitions, else they would be part of the ko or superko. And a local triple ko, where losing one ko loses them all, is easier to keep track of than three separate kos, where losing one does not affect the other two. Besides, a local triple ko has 6 relevant states, while three separate kos have 8. It does not take much complexity to reach human limits. OC, go itself is very complex, but keeping track of the state of the board is a separate task. I submit that it is up to the players as a group to say what they need to keep track of. A rule, no matter how short and simple to state, that does not take into account what it requires of the players is a bad rule. True, it is undesirable to have a game that repeats forever. But a superko rule, of whatever ilk, is not the only solution to that problem.

[RobertJasiek]

Meta-discussion (just a game, Einstein etc.) ignored. (It is a common trick to resort to meta-discussion when lacking arguments for discussion.)

You claim to know what common sense and majority of perceiving it were but you don't prove it. You ignore that common sense prefers fewer rules.

"It does not need extra rules. [...] ditch the unnecessary rules." Then why do argue in favour of extra ko rules? (One or several separate rules for long cycles and hypothetical ko rules are extra rules in Japanese, Korean, Chinese, Ing rules.)

You claim that there would be a common understanding of ko rules in Asia but the opposite is true: different Asian countries have different exceptional ko rules and each Asian country has had different exceptional ko rules in different rulesets. You may be following some idealisation of how you would like to be Asian style ko rules for your understanding of what should be Asian common sense. What are the imagined ko rules of your common sense idealisation?

[John Fairbairn]

You claim that there would be a common understanding of ko rules in Asia but the opposite is true

When lacking arguments for discussion, it is a common trick to resort to saying somebody said something they didn't say.

Go has two kinds of rules. There are the rules of play on the board. Although different versions exist, each version can be kept very simple - imperfect but practical and simple.

There are also administrative rules. For example: komi; what to do in void games; time limits; etc, etc.

There is no common understanding of administrative rules even within a single country. Sponsor A (sponsoring say a league) might decide a triple ko is a draw, half a point each. Sponsor B, sponsoring a knockout, might demand a replay with colours reversed. Sponsor C, also sponsoring a knockout, might demand a replay with colours not reversed. Sponsor C, promoting rule change or standardisation, might demand electronic surveillance and a PhD in reading rulesets.

At one time, pro go in Japan had three komis in operation simultaneously: 0. 4.5, 5.0 and 5.5 (and some amateurs used 6.5).

But all these are administrative rules that fall within the bailiwick of sponsors, administrators and professional guilds. They are not the concern of people who buy a beginner's book to see what the game is like, and then spend their lives playing socially if they do like it.

You claim to know what common sense and majority of perceiving it were but you don't prove it. You ignore that common sense prefers fewer rules.

Well, it's common sense that I know what common sense is. It's not something you have to prove. But if you are not used to applying common sense, you say daft things like "You ignore that common sense prefers fewer rules," ignoring that is actually what I said (remember the ditching rules comment?) and ignoring that "few" is not the only necessary attribute: we need to add attributes such as simple, understandable, workable....

Here's a simple test to see whether you are using gumption and common sense or you are being driven by intellectual hubris.

You can play go with an opponent entirely in your heads, just relaying the moves (A4, K11 etc). Or you can use a board and stones. Which method would you use? The much cheaper one with fewer components or t'other? Then ask yourself why you made the choice you did.

[RobertJasiek]

Again, what are the imagined ko rules of your common sense idealisation?

[John Fairbairn]

Again, what are the imagined ko rules of your common sense idealisation?

You've got to learn to get out of your straightjacket mindset.

I neither set nor imagine any rules. Nor is it the goal of common sense to be ideal. It is to be practical, useful, applicable etc.

The common sense applications for ko anomalies are, for example, accepting whatever rules pros and/or sponsors choose to set in pro tournaments, accepting or not accepting (by choosing whether or not to attend) whatever organisers set in amateur tournaments depending on how much any extra rules trivia one wishes to accept, or, in social games, having a laugh and maybe a beer and a friendly discussion if an anomaly arises. None of them are about controlling other people. What's so hard about any of that? IT IS JUST A GAME.

[HermanHiddema]

I think Bill makes a good point. In theory, to play superko you need to memorize every board position that has occurred in the current game. An easy task for a computer, an impossibility for humans.

So that's not what we do. In practice, we recognize that there a certain patterns that create repetition. Basic ko, Sending two returning one, eternal life, triple ko, quadruple ko, quintuple ko, etc.

And of course, even among those there are variations. E.g. triple ko might be three basic ko shapes between the same two groups, or it might be a double ko seki combined with another ko.

As humans, it gets ever harder to play the repetition correctly as the cycle gets longer, because we need to memorize more positions that have or have not occurred.

Traditional rules in Asia take the approach to only deal with short cycles of length 2 or 3, basic ko and sending two returning one, and to ignore the ever rarer longer cycles.

New Zealand rules or AGA rules would leave it up to the players to e.g. play a septuple ko, and good luck to them.

I'm sure even the strongest players would, at some point, consider a repetition too complex to play in a reasonable time frame. It's just that the really really long cycles are so rare, that we never have to think about them.

But if you're serious about preferring the "simple" rule, you should be willing to consider whether it is a good idea to have players play the septuple ko. Is a rule set truly "simple" if it asks that of the players?

[RobertJasiek]

John, thanks for clarifying what you want. When you spoke about common sense, I expected rules (or rules application) that are common among all or most players. Now you clarify that you want to allow a wide range of rules applications, basically whatever is set as rules for a tournament or (implicitly) agreed between players in other games. I see. However, this includes such tournaments with superko. You have expressed your disliking of superko but consider it common sense to accept superko whenever set in the announced rules. I hope I have understood you correctly about this.

Herman, what one wants WRT ability of remembering positions depends on preferences. One can give a high priority to the observation of rarity of hard-to-remember-cycles and conclude that, in practice, superko is perfectly applicable. Or one can give a high priority to the requirement that ko rules must be reasonably applicable even in the rarest case and conclude that superko is not suitable. Etc.

"Traditional rules in Asia take the approach to only deal with short cycles of length 2 or 3, basic ko and sending two returning one, and to ignore the ever rarer longer cycles." No, that's not what Asian ko rules do. It is what my 3-plays-ko-rules do: http://home.snafu.de/jasiek/3_plays.pdf Asian ko rules differ. Chinese ko rules can be interpreted to have a rule for 2 plays, a rule for 3 plays and exceptional rules for more plays. Japanese and Korean rules have a rule for 2 plays, rules for more plays (of which some classify so that 3 plays belongs to a certain class) and exceptional rules.

A rule can be simple to learn but some such simple rules (e.g., superko) can sometimes be hard to apply. You might demand relative simplicity of both the rule and its application.

[John Fairbairn]

common sense, I expected rules (or rules application) that are common among all or most players

It did occur to me that you were misinterpreting "common". I dismissed that idea because I know the English phrase "common sense" is used in German. The very fact that Germans have borrowed the phrase suggest they may have felt the lack of the phrase in their own language. However, I don't know how much it I used, or even whether it is used the same way as we use it. After all, 'fair play' is a similar example of such a borrowing, into various languages, and that has certainly ended up with different meanings from ours.

Was soll es bedeuten? Of the various German close parallels I know, I suspect Vernunft might have been uppermost in your mind (and way off, if so!). Menschenverstand seems a bit closer to empathy to me, unless you qualify it with gesunder - but then that sounds a bit prim. My own preference would be for Hausverstand (or maybe Alltagsverstand). I have not checked, but my guess would be that Hausverstand has even been borrowed into (?American) English as horse-sense, which is passably close to common sense.

What's your personal usage? (And no, I will not accept Hausverstand-1 and Hausverstand-2 ) Are there any good dialectal versions you (or anyone else) know?

[RobertJasiek]

"gesunder Menschenverstand" or, with a restricted scope of usage and not applicable here, "Allgemeingut".