Practical superko rule - force adjudication

[hzamir]

I'm wondering about the following:

• Superkos are tidy--as rules--, but unworkable in practice, since less trivial cycles are not necessarily detected by humans before they are committed, and sometimes not until some moves after that, prompting the argument that "superkos are bad game design"
• The alternative that a game can be no-result (or draw), has the opposite problem: it's agreeably lax detection, but produces no positive result.

So what I wonder is, why have I not seen a suggestion along these lines--when there is an available referee?

A player who believes he has witnessed a cycle, regardless of his participation, may for his next move--provided it follows an actual play by his opponent--call for the game to be adjudicated, immediately ending play, and the referee makes one of three determinations:

• He erred about the fact of a cycle, and he loses.
• It is an adjudicable game with a cycle, thus determining the final score.
• It is an inadjudicable game at that point, possibly awarding the game as an incentive for breaking the cycle.

The proviso that calling for adjudication cannot follow opponent's pass is that a pass should be treated as an attempt to end the game without resorting to authority.

Pass lifting ko bans is not considered, as it's another flavor of superko rule.

I would say there are two kinds of adjudicable games

• one reducible to a true score
• a lopsided win reducable only to black/white wins by at least n points

One last issue: Could it be that forcing adjudication of a game is just culturally unacceptable idea in the heartland in those places where the game has the most history?

[RobertJasiek]

Superkos that are unworkable in practice in less trivial cycles not necessarily detected by humans do not occur in practice. The most complicated practically occurring cycles have been a) quadruple ko (with easier cycles than sequences in complicated life and death problems) and b) molasses ko (which has occurred only exactly once in the history of go). Practically occurring rare cycles of lengths 4 to 6 (of which some, namely triple ko, can be played like a basic ko with ko threats after exactly 1 ko stone) are not difficult in practice because they would be unworkable but because quite a few go players refuse to study superko application before occurrence, as if they refused to study ordinary life and death tactics with sequences of lengths 4 to 6 before their occurrence. The practically comparatively frequent cycles of lengths 1 to 3 are not unworkable, either. In fact, a cycle of length 3 is easier to understand under superko than under no-result rules.

There is not only "the one alternative that a game can be no-result (or draw)" but this just happens to be the traditional alternative. There are quite a few alternatives handling or preventing long cycles. Three decades ago, I have proposed one such alternative: the Basic-Fixed-Ko-Rules, which is the combination of the basic ko rule and the fixed ko rule with its prohibition to continue a cycle as soon as it is created and the strategic consequence that no long cycle is ever worth starting to play at all. Apart from the unwanted prohibition of all kos (including the basic kos), this is the simplest possible handling of long cycles. (Additionally, a forgiving tournament rule might provide that accidental recycling is undone or ignored.)

If you suggest some alternative that requires presence of a referee, it may be interesting from a point of view of expert rules study but is more complicated than, e.g., my suggested alternative of the Basic-Fixed-Ko-Rules, which should work without referee because long cycles should not even be started (or, together with the tournament rule, would still be harmless).

Your suggestion has the great disadvantage of unpredictable arbitration.

[hzamir]

I've already read (and briefly reread) through your fixed ko rule page on seseis.xmp.net, and the "types of basic ko", but unfortunately didn't find it so easy to understand. The pure fixed ko rule jumps from a trivial example where retaking kos is allowed straight to an 8 move example, without bothering with illustrating how it applies to simpler examples. Nor do I care for it now declaring it illegal to reproduce a pair of successive board positions rather than a boar position.

I also don't care specifically for making any additonal moves illegal other than suicide, and immediately retaking a ko, as those are blindingly obvious.

It is true that the cyle adjudication rule should set expectations about stones might be in a removable state.

But I don't think adjudication is a big deal in this context because we are not speaking about the merits of adjudication in a vacuum, we are talking about a variant of a superko rule with different mechanics. With superko rules you are outright declaring the person who actually repeated a position the loser, something which could in fact be missed in the moment, otherwise no one would make the illegal move unless out of ignorance. Whereas under this adjudication rule, you could award any ambiguous result to the player who calls out the cycle first. IMO, not worse than superko rule. And my admittedly poor understanding is that one motivation for forcing cycles is for a losing player to forestall a loss.

I also kind of like the double edged risk/reward for the player who tries to end the game by means other than passing.

[John Fairbairn]

The alternative that a game can be no-result (or draw), has the opposite problem: it's agreeably lax detection, but produces no positive result.

It's not really a problem at all. Why must a game end in a binary way? Ternary works well in very many cases. Football has 0-0 draws, chess accepts draws, and you can even have two gold medals for a dead heat in the Olympics. Even go has always had three results. Jigo (or the Chinese equivalent) has a VERY long history.

The only problem for go has been in the TOURNAMENT rules. In a knockout tournament, especially where Round N+1 is played immediately after Round N, a drawn game can cause a problem with timing, but that's a problem with the tournament's rules, not go's rules.

There are easy ways to get round even that. Baseball has extra innings, football has penalty shoot-outs, golf has extra holes, go and chess have blitz replays. Far from that being a problem, this usually adds to the journalistic drama, which is a good thing in games that rely on sponsors and thus media coverage.

The arguments about superko and so on are all really about people playing mathematics with a go board. Not a game many of us want to play.

[RobertJasiek]

hzamir, the pure fixed ko rule is very different from its combination with the basic ko rule. So do not dismiss the latter by only studying the former!

John, it is not only about tournaments or maths. It is also about beginners trying to learn the game on their own. No gaps / exceptions in the rules ease their access.

Similarly, I was put off chess when, in a tournament, I learned that I was unaware of the en passant exception, which had never been taught to me before. It may make the game of chess more interesting (just guessing) but such is a hurdle for beginners. Whatever long cycle ko rule is used for go, it does not (except for experts and sensationalists) make the game more interesting because long cycles with different strategic behaviour due to different long cycle rules are rare. There simply is no need to complicate the rules, other than to preserve professional tradition in rare cases (ignoring that such tradition has changed).

[hzamir]

The alternative that a game can be no-result (or draw), has the opposite problem: it's agreeably lax detection, but produces no positive result.

It's not really a problem at all. Why must a game end in a binary way? Ternary works well in very many cases.

• I agree it is not a problem where obtaining a win/lose isn't strictly required.
• I personally prefer that a draws be possible in a truly even game (e..g komi=7).
• I do think it a problem that someone otherwise demonstrably behind can force a cycle to prevent counting.
• I do think it a problem that superko should resolve that problem by either:
  • one player preventing himself from making an illegal superko
  • another player immediately recognizing his opponent's illegal move
  • a later realization of a cycle that must be rewound and attributed to the offending player

This is why I asked if it is not more reasonable that creating a superko not be in itself illegal, but that the first player who claims it occurred can end the game on his move, with the attendent risks/rewards for doing so (with risks of losing for being mistaken, accepting his own loss if adjudication goes against him, and reward of winning if correct and he wins adjudication or ambiguous cases).

[RobertJasiek]

[*] I personally prefer that a draws be possible in a truly even game

One of my other inventions is the Long Cycle Rules:
- prohibited short cycle of lengths 2 or 3,
- tie in a case of a long cycle of length 4+.
Again, the practical differences to other ko rules are marginal.

(This happens to agree to Chinese Ko Rules if their ambiguity is removed by a) clarifying their intention of [what is stated as if it was] superko only applying to short cycles and b) the referee always declaring a draw in every case of a long cycle.)

[jann]

The alternative that a game can be no-result (or draw), has the opposite problem: it's agreeably lax detection, but produces no positive result.
...
One last issue: Could it be that forcing adjudication of a game is just culturally unacceptable idea in the heartland in those places where the game has the most history?

I think what you overlook here is the importance of simplicity. True, conceptual simplicity (not textual conciseness which superko aims at for example). People tend to overestimate the acceptable complexity of rules - one tricky rule invention is about one more than acceptable.

Imagine what would happen if the rules would not include ANY special rule about repetition (just the normal ko rule, and territory scoring so prisoners matter). This simple ruleset would theoretically behave like current Japanese rules: in some games neither side could force scoring in a won position, leaving those without a winner. Saying "no result" just acknowledges the fact that the game continues forever. Not an additional rule changing anything, just a consequence of normal go play.

Compared to this, any adjudication scheme (or other rule invention) will likely be seen as unnecessary complication.

It's not really a problem at all. Why must a game end in a binary way? Ternary works well in very many cases. Football has 0-0 draws, chess accepts draws, and you can even have two gold medals for a dead heat in the Olympics. Even go has always had three results. Jigo (or the Chinese equivalent) has a VERY long history.

The only problem for go has been in the TOURNAMENT rules.

How bad this is in reality? When the tournament cannot accept a draw, a rare triple ko with an immediate replay drags things out for sure. But if the repetition arise in middle game, it amounts to something like an 50% length increase - which may not be much more than what normal game length variance could produce.

[RobertJasiek]

How is what you call "normal ko rule" not a "special rule about repetition"?

What is what you call "normal ko rule"? (Do or do not passes lift bans?)

Only having either basic [your "normal"] ko rule, a player is not just faced with exceptional outcomes per se but, more importantly, is also faced with strategic undecidability whether to choose options involving or not involving such possible exceptional outcomes.

[jann]

The ko rule is not "special" in the sense it does not face adoption resistance, thanks to the last millennium.

But from new inventions, the only thing that recently gained noteworthy support may be the pass for ko rule of J89, and even that is unknown to most players. Territory scoring with simple ko and L/D on agreement works too well in practice, for any signifcant new invention to gain momentum.