And in this case, considering end game, if it’s white’s turn she would pass (black cannot force her to start playing inside to prove anything). White has 16 points inside. Black initiates the invasion and white keeps passing (eventually intervening to prevent black from making 2 eyes) and ends up with 19 points inside or even 20. However that’s not a pass-pass end game and white would happily accept that black believes he’s alive and the game goes on. So, no deep problems with any rules here, just a beginner’s blunder, right?
So in those cases, white should probably not pass after black’s first move, but play the other miai (although this may depend on the balance of ko threats):
Only now white can pass indefinitely and gain points under Japanese rules if black continues (as long as white ensures her outside stones stay safe).
A trickier situation arises when black starts at 1 and then both players pass (so the game stops) and black claims his stones are alive while white claims they are dead (hoping to gain a point under Japanese rules from black playing at 1):
In my understanding of the official Japanese rules this black group is dead when the game has stopped, because players can’t use ko threats elsewhere during hypothetical play.
So if black wants to fight it out, he should do that during regular play. This means that black should either resume the game to start a ko fight with 3 (to save his group, or at least reclaim the one point loss of black 1), or accept white’s claim that this black group is dead (and thus accept that black 1 lost a point).
Under Chinese rules this disagreement can be settled by just resuming the game and fighting the ko. Then again, under Chinese rules white does not gain points from refraining to play at J1 in response to black 1, so white should just defend at J1 before passing to avoid any shenanigans.
An even trickier situation arises when black plays 1, white passes, black plays at J1, both players pass and then disagree about the status of black’s group while refusing to resume the game:
In this case the situation is actually unsettled because the status of black’s group depends on who plays first under hypothetical play. If both players keep disagreeing about the status of this group and also keep refusing to resume the game to settle the situation, I think the official Japanese rules state that both players lose the game.
I don’t know for sure how this situation would be handled under Chinese rules, but I assume black’s group would be considered alive because white refuses to prove it’s dead by resuming the game and actually capturing it.
A similarly unsettled situation arises when black plays 1, white passes, black plays at 3, white captures, both players pass and then disagree about the status of black’s group while refusing to resume the game:
So it would be white to argue black is dead, hence white would be able to play first instead of black, hence black cannot start a ko. And if white refuse to prove it is dead, they would be considered alive (since you cannot prove it, you are declaring it is alive).
Aha, so when the game stopped like this under Chinese rules…
…, black would be considered dead (the same as under Japanese rules). But if black disagrees and white refuses to resume to play J1 and eventually capture black, black’s group would be considered alive.
Mind you that the trickiest situations (stopping the game under Japanese rules with an actually unsettled situation on the board and the players disagreeing about its status while refusing to resume the game) would only arise in real games when the players are novices or trolls.
In the Chinese rule there is no regulation or “examples” for which is dead, only depends on players’ claims. Hence not automatically dead, the one who claims to be dead has the burden of proof, otherwise, alive.
And the interesting thing is that they can both claim the other groups are dead and refuse to play them out, in that case, it would often invoke the second regulation, that there is still fighting and cannot be resolved, hence they are both alive (like seki).
I think it would only be considered “both players lose”, if the value is large enough to change the outcome. If not, then this specific shape is also given as a particular example of “anti-seki”, where both players are considered dead, but stones are not removed and no territory is counted.
With fractional komi (6.5) it seems hard to imagine such unsettled shapes to actually remain on the board under J89. If the position is scored as is (neither is alive) then one of the players will lose on points. He therefore has no reason to accept the scoring, and will rightfully claim that there is an “effective move” and the game cannot end. The opponent will then have to decide: If he can still win like that, then he will request resumption, and the losing player (going first) will settle the shape. If not, then both lose. In neither case will the group actually be scored as seki (assuming optimal behavior from both).
Apologies for the delay in response, things came up in life.
Regarding the questions about how territory works, I think the clearest territory rules are Jasiek’s New Amateur-Japanese Rules. These are relatively easy to motivate as a progression from stone scoring and don’t (directly) have the complicated “ko-pass” and other stuff that makes the pro rules complex for beginners.
Going back to the discussion about Taiwanese rules options:
As much as I like the idea behind Ikeda I hybrid rules, people regularly fail to implement them correctly. E.g. they use Lassker-Mass which has pass fight issues. Even with Ikeda I, if you want to avoid the occasional strange end-game outcome. You either have to do a 3-phase thing like Kata Go does, or you have to require 3 passes to end each phase and have a non-superko repetition ban (like Jasiek’s basic + fixed ko rule) that lets you dissolve end game kos during play.
Without such adjustments, you still get novel end game behavior in a few cases, not unlike your example of the problems with button go.
I would be fine with a 3-pass variant, and it has the merit of being very close to the rules used by AI. But I don’t think the mechanics are all that transparent and strategically clear, unlike those proposed territory rules. Where we can explain them like this: “In the event of a dispute, we score the position the players actually passed in based on who would control that territory if play continued. And we look at it from the perspective of both colors because we want the score to be the same for the board we have as it would be if we flipped the colors and reached the same position, i.e. the sequence of moves that got us to the final position and prisoner count, should not change the score.”
I think your example of a problem with button go is a good illustration of why the World Mind Sports rules were what they were: they have a restriction that a player (generally) cannot pass while a stone is in ko. Which, unless I am misremembering the sequence in your example, would force white to end by filling the ko in either sequence, and if the “white pass first” compensation is only applied to the final pair of passes, then those rules “work” for your example since black will always pass first after the ko is filled.
However, according to Jasiek, correctly explaining the nuances of this “can’t pass while in ko” thing quickly turns into something similar to the Ing Ko rules. Or least the kind of ko rules that track the parity of the repetition. This seems like a cure worse than the disease.
So, going back to the Taiwanese rule of “last competitive play”, if you have a suggested rules text for implementing it, I’d love to see it. I like the idea, I just don’t know how to implement it in practice.
Conceptually, it seems simple enough: we rewind the game and if the last stone played that caused the final score to change was played by black, then white gets an extra point of komi. This seems like it would always correspond to the last play by a player after which that player only passed or played inside of their own territory.
It feels like we want to say that while the board has 361 playable positions, we will only consider 360 of them to be scorable. So, in an otherwise, tied game, if both players have 180, we do not allow black to get to 181 by virtue of having played more stones than white. (i.e. there should be the same number of stones played).
My concern is that it may have unforseen consequences similar to button go. The issue with button go rules in your example is that white can manipulate the pass parity and create a situation where the players have played an unequal number of stones, which messes up the math behind the parity correction (since that math assumes no odd seki and no uneven passing.)
In particular, based on the more general math here, the issue is that you can “fake out” a parity correction that only looks at passes or last plays by having an uneven number of moves of an offseting parity. (The same problem would likely arise if you tried to rely on territory parity directly instead of “last play” or “first pass”.)
Your “rewind to find the last play that scored” idea avoids the problem in your example, but there are probably similar issues to button go because, like button go, it only tracks the parity bit of the last play and doesn’t factor in the ability to shift the apparent parity via uneven passing.
So, we would be assuming that there is no sequence of plays whereby white can do something that does not change his score, but causes black to play the last scoring move.
Hence my proposal above that you use area rules, but, up until the first pair of passes, you correct for the full amount of any differences in stones played. (And after the first pair of passes, the dynamic komi is locked in. If play resumes, then there is no more adjustment. Since the score should not be changing at all during dispute resolution and we don’t care about how many moves it takes to demonstrate something.)
I mentioned pass stones because I was essentially using them “in reverse” since it’s easier to track imbalanced passes than actual stones played. But it amounts to the same thing.
Since the ultimate logic seems to be that white and black should play an equal number of stones. We get what we want in a transparent way if we can simply enforce that (and can avoid pass fights or any other “odd” behavior that makes it strategically optimal to pass and then make some play that scores in the second phase.)
That said, I don’t know if this does actually avoid odd end game behavior. And it might be equivalent to one of the flawed hybrid systems above instead of what we actually want.
So, it might make more sense to only apply a 1-point parity correction instead of the full difference in moves played: if the difference between white and black passes (with white always going last) is odd, then white gets 1 extra point of komi. Since white will always pass last, then the only parity difference between area and territory can be in the number of moves / passes. So correcting for odd parity here should correct for the parity of the entire board. (If I did the math right, it also happens to correct for parity differences due to odd seki since we are basing our parity correction on the precisely the set of situations where the parity in area and territory scoring differ.)
What do you think? Does this give an acceptable answer in your button go example and other cases like it?
P.S. I think another way to state what I am proposing that doesn’t rely on equivalence scoring is as follows:
Area scoring is used. The number of passes is tracked. After two passes in a row, the game ends and komi is adjusted in two ways: 1) if white made the first of the final two passes (or black made the last pass), white is paid 1 additional point in komi, and 2) if the difference between the number of passes made by both sides is odd, then white is paid 1 point less komi.
I think this approach has the general flaw of allowing certain lines of play that are strategically bad in normal go, but have better consequences if only control is reflected back to the original position and its original borders (like in J89 example 2). Hypothetical play only works correctly when its relation to normal go - esp its objectives - is maintained carefully.
If the board would be even (like 360), area scoring would still suffer from the same parity problems, except in that case it would be odd territory scores getting rounded up to the next even area scores with B’s surplus stone. Since territories can be either odd or even, in one of the cases stones will always be odd.
This is always possible due to oversights, but is also why design methods matter. Focusing on the real problems and their theoretical roots is usually safer than ideas with only indirect connections. For area parities, the real problem is B’s extra stone and its area point, so:
direct/minimal rule: the extra point of komi depends on whether B played more stones in main game
historical Taiwan rule: “who played last” - which will usually correlate to the extra B stone, but not always, eg. could create pass fights without the “competitive” hack
button go: “who passed first” - which will usually correlate to who played last, but not always
Wrt stops with open ko and switch to pass stone encores, imo the point is that players will NOT learn complicated rule inventions. They only know the capture rule and the simple ko rule, and will stop games on two passes (with some vague idea of resumptions being possible).
They don’t expect a dispute in advance, and will not fill dame in the main game before passing. Thus slower and more careful transition, like more phases with two pass stops and more resumptions seem better than fewer phases with three pass stops (since the extra phases won’t happen in practice anyway). And when done correctly this can also solve the open ko problem that LM has.
I’m unclear what you mean in context. Are you saying you disagree with Jasiek’s NAJ rules or that you disagree with my simplified / beginner explanation of why we do hypothetical play to determine life and death, but score the final position instead of the result of continued play? It seems like you mean the latter, but I am unsure.
Edit: For clarity, do you mean “Life-and-Death Example 2” here? Because I am under the impression that this scores correctly with Jasiek’s NAJ rules. See here, specifically “traditional name: seki”.
Here is my bottom line position:
If I understand the argument for why this parity matters correctly, what we care about is move parity (stones + prisoners + passes). The difference between white and black can only ever be 0 or 1. And in the case where it is 1, we want to give white an extra point of komi.
If we do the math, I think this works out to mean that we only need to track the parity of the passes. (Sum or difference doesn’t matter). If pass parity is even, and black passed last, then white gets the extra point of komi. If pass parity is odd, and white passed last, then white gets the extra point of komi. Otherwise, white does not get the extra point.
All of the rule sets that attempt to track this more simply end up being wrong because if you only track one source of parity changes, then the players can always “hide” the shift in parity in something you don’t track.
Do you agree with the general idea and do you think my math is correct? Is there any fundamental problem with calculating this directly instead of using one of the various shortcuts?
I’m going by the discussion on Jasiek’s website and his criticism of how that rule basically brings in Ing-style ko rules. See pages 6 and 13 of this PDF according to that document, the official wording was: “For controversies over half-point ko before the endgame, the owner of the ko must end the ko and make the last play, despite the outcome of the ko.”
So, essentially, the rules that Kata Go uses, where phase 2 changes the ko rule to have “pass for ko”, and then phase 3 adds pass stones?
In J89 example 2 (in the examples section, not in commentary), w can take the corner in exchange of giving up two outer stones. In normal play this is a neutral exchange (tho either the corner or the outside can be bigger to change this) so W usually leaves things as seki. This correct outcome depends on hypothetical play behaving similarly to normal play strategies and not creating new loopholes.
For example, if there are two analyses as in naj, W may score the corner points without having to fully give up the outside (since he can play differently in the two analyses). For other rules that try to project back per-intersection control to the original position and its possibly incompatible borders, a similar problem is that this projection can result in pathological strategical values, creating lines of play that differ from normal play.
I think it is simply about played stone parity. Prisoners are irrelevant since they are equally accounted in area and territory, and passes also have no direct role, at most an indirect correlation again.
Not necessarily that particular one, but similar. The second phase cannot have pass stones yet (there can still be unfilled dame). “Pass for ko” is a heavy burden and only necessary if one aims at max Japanese compatibility - it is generally not for pass stone playout rules.
Multi-phase disputes don’t completely solve the pass stone switch with open ko problem though, only mitigate it, so some additional rule may also be necessary (the issue is how to distinguish such open ko from double ko sekis or 0-sided kos). But any extra rule should affect dispute phases only, and not fiddle with the normal and safe stop of the main phase imo.
If you only count played stones, (stones on board + prisoners) then it seems like it will always be possible to use a ko to force extra passes and change the parity of played stones without altering actual board parity. (Edit: this was wrong, see my demonstration below showing that the two are equivalent)
Implementation issues aside, can you give an example where my formulation does not work while your “stones only” version does?
As for passes being “indirect”, I’m taking advantage of the parity math to derive the actual parity of the game from the information we have most readily available. You can use whatever combination of the factors you want from that page to arrive at the same result.
Even under normal life-and-death territory rules, you consider both cases of white and black starting. So I don’t see how “two analysis” can possibly be the source of a problem. I’m also not convinced that it can result in lines of play that differ outside of the rare cases he documents.
In particular, unless I have misunderstood what playout sequence you are envisioning, then your application of his rules doesn’t match his own commentary’s treatment of that position. (And I find it hard to believe that he didn’t meticulously check every single example in the 89 rules given that he does list the examples where his rules differ. So essentially, you are saying that he applied his own rules incorrectly. Is that correct? Or is there something I am missing?)
Specifically, there is no sequence where black plays first and that ends in 3 passes wherein white can turn those stones into an immortal shape (one that black cannot capture if white exclusively passes). Black will play at a (the dame) and after that pass until the end of the analysis phase. If white attempts to capture the black stones in the corner, they will put all of their own stones in atari and black will capture them. If white also passes, the shape will still be seki at the end, because with white exclusively passing, black can easily put the stones into atari and capture them. Thus the white analysis shows that white controls no territory in that shape. With the black analysis, white is allowed to go first to determine whether black can make any of his stones immortal. White will play the dame and then pass until the end of the phase. If black attempts to capture white, it will put their stones in atari and thus white will capture them, showing that black controls no points in this shape. If black also passes, at the end of the analysis, the stones will still be in seki, and white can easily capture them.
So this position scores correctly. Can you provide a more optimal sequence for white and black such that the score is otherwise? Regardless, none of this is unnatural or a deviation from normal go strategy. It’s very basic territory scoring hypothetical play.
Edit:
Upon reflection, I think your actual plays version and my rule are equivalent:
Practically speaking, my phrasing seems sub-optimal because of what happens if one of the players misreads the end game and/or misunderstands the strategy and tries to shift the parity by playing in their own territory or their opponent’s. You can propagate the parity math and talk about it in terms of the first pass, or probably ideally, you would use equivalence scoring and use the territory parity itself. Since you are already counting that number anyway, it would make for cleaner bookkeeping and probably be less error-prone in practice.
Regardless of how you arrive at it, the goal with my version is to specifically derive the “P” value in the formula on that scoring page that accounts for the difference between territory and area rules
Looking at your “board plays only” version for comparison.
First off, it is possible for black or white to make multiple extra board plays. So the difference is no longer exactly 1. Setting that aside, and stipulating that we only care about parity, then there are still 2 cases to reconcile the methods:
Black passes last, this means that the difference in moves is odd. So: If pass parity is even, then play parity is odd, so you award white a point, but if pass parity is odd, then play parity is even, and so you do not award white a point.
White passes last, meaning that means move parities are equal. If pass parity is even, then so is play parity, and thus you do not award a point. But if pass parity is odd, then play parity must also be odd, and so you award white a point.
This seems to be the same as what I said, but by a different route. In both cases:
Black passes last & pass parity odd (play parity even): no compensation
Black passes last & pass parity even (play parity odd): white gets 1 point
White passes last & pass parity odd (play parity odd): white gets 1 point
White passes last & pass parity even (play parity even): no compensation
Does that make sense to you?
Also, what I mean by “sub-optimal phrasing” is this:
Suppose the game is over and white has made the first pass. Normally, black would also pass, and the game would be scored. At this point, the territory parity is locked in. But black might think they can change this by playing in their own territory (or in white’s). And thus when white passes and then black passes, the pass parity will be “wrong” and he will have “tricked” the rule. In reality, white can also play in their territory (or black’s). And put the parity back like it was. Either black will pass, followed by white, and white will still get the point. Or the situation will repeat until we have filled in everything. And since pairs of plays can’t alter the territory parity (or play parity in general), the score will not change.
But for online usage, I could see trolls trying to force the other player to give up the parity correction point by trying for force this sequence and being aggravating enough to cause the other player to give in.
So, the phrasing and implementation should probably be different as a practical matter to make it so that it does not matter whether white responds with a stone or a pass.
You could have it so that once white has made his stones immortal, they can just click some “pass forever” button and the game will get scored correctly. (Same with black in the opposite parity scenario).
But I think that simply using equivalence scoring fixes this since “give a pass stone” and “play in your own area” are made equivalent.
In that case, you replace “black passed last / white passed last” with “white had to / did not have to give an extra pass stone to end the game” and you reverse the pass parity rules.
Thus, white gets the extra point, if they had to give an extra pass stone and there were an odd number of pass stones exchanged. Or if white did not have to give an extra pass stone, but there were an even number of pass stones exchanged.
I overlooked that he starts analyses with defender, so in the simplest case the throwin saves. But what if there are two such sekis on the board? Or what if the defender has no useful move, like in J89 example 4, or this funny one from L19?
Generally, projecting back control to the original position and original borders may allow pathologic sequences since the strategic values change by not caring what happened to those borders for example. In J89 the enable rule keeps hypothetical play reasonably compatible with real go.
Earlier you seemed to mean aga-style pass stones with W must pass last, so this doesn’t seem clear, nor is it clear exactly what you aim at. As you wrote, pass parity might be manipulated by playing into territory for the same area value as a pass. Also, what if the board is almost finished with one dame left, it is B’s turn but he misses the dame and passes instead, letting W take it (normally without Taiwan compensation)?
The complications mentioned in this thread is why one should chose area rules in OGS. Better if AGA or New Zealand because the standard text is in English. Area rules allow all positions to be played out with no difficulties. Territory rules can also allow playouts. Ikeda territory rules and Kata Go Japanese-like rules do but Japanese rules do not. Japanese rules are simply bad and should be avoided. The Go world would be better without Japanese rules.
That’s denying why these kind of rules exists. A bit like walk and don’t use cars because cars break sometimes.
I mean you can understand the attractiveness of not counting the stones but just what is between them. That’s already what you do during the game to evaluate the score, you are not going to count the stones, aren’t you?
Although it’s not so simple as it looks.
But for the most, players are happy with that, even if maybe in a few cases their game will lead to some convoluted result.
Now tell me what do we care about a one point difference sometimes in our amateur world? Of course we can appreciate and study and comment on OGF and argue… Well I get that some care, some prefer do something else.
I believe all of those are addressed in his commentary. And the projection does care about the original borders, you can’t score territory that isn’t surrounded by living stones that were in the final position, etc.
Also, Japanese rules do allow for adjustments of actual borders to impact life and death. In example 2, their reasoning is that it’s seki because black can play an unremovable stone on one of the contested points – i.e. the fact that he can move the border is why no one’s stones are dead.
I dropped the AGA requirement because you said you didn’t like it. There are probably dozens of ways to compute the same result. I like the idea of using pass stones because it makes the accounting cleaner by making passes equivalent to plays in your own territory. So it limits the avenues for things to go awry by taking away one avenue by which parity differences can occur.
My example of a player attempting to troll could be handled in other ways and with other calculations. But the pass stones have clear strategic ramifications and keep the game connected to a hypothetical stone scoring version that gives a bonus for the single point eyes left at the end. (And if you want to add in a correction for odd seki, staying connected to this version seems like it makes the explanation for why you care about odd seki cleaner.)
The way I wrote it that you quoted was an attempt to explain the reasoning (under optimal play) that connected what I’m saying to what you are saying. I.e. I dropped the pass stones, I focused on move and play parity, etc.
But, as I stated, the formulation is sub-optimal as a practical matter. The players can’t actually shift the pass parity under correct play. It’s just that the way I wrote it makes it possible for a sore loser to be obnoxious.
Alternative calculations wouldn’t have this issue. This is also why I originally used pass stones, I think they are the cleanest way to handle things because it limits the amount of parity information you have to track.
If black takes it, he gets a point, but also passes last. So the point is offset by what is awarded to white.
If black does not take it and white passes. No one gets the point.
If black does not take it and white takes it, then black has made an error, as under normal area scoring because white will score the point and still get to pass last with odd pass parity, getting the point they would have gotten had black played optimally.
Dame must still be filled because (by design) the strategy under this rule is the same as area scoring. We are only correcting for the parity of the full board so that the score is granular and recognizes the extra point of territory difference that we care about.
My goal was not to make area end game reasoning closer to territory end game reasoning, I wanted to calculate the parity difference while leaving the strategy completely unchanged. As a result of the math, correct play under area scoring will remain correct play, and incorrect play will remain incorrect.
If you want to have some kind of strategic middle ground, then you will have to make compromises and introduce novel end game sequences in the process. I think this is undesirable. No one said that we should care about these strategic differences, the argument was that area lost 1 bit of information. So I used math to restore that 1 bit because it is still there on the board, it’s just that the scoring method ignores it. We can not ignore it and have normal area rules, with all of the same strategic ramifications while still factoring in that 1 bit of parity information.
If you wanted to avoid this “same strategy” thing for some reason, then you have to correct for the full difference in the number of board plays. But this is what territory scoring does, and it requires multiple phases so that there’s a definitive point where that difference is counted somehow. I suggested this above as an alternative. But I do not know if it leads to pass fights or other issues. It’s certainly plausible that this works fine. But there is always the possibility that the slightly different rules and scoring in the second phase will cause people to delay resolving a ko or some other thing until the second phase, rendering it strategically meaningful instead of existing solely to resolve disputes. (This happens with normal Ikeda for example.)
And at that point, I think you are better off with simplified territory rules instead of making your own multiphase thing that will either introduce novel end game behavior or adjustments like 3 passes with passes clearing ko.
To further motivate the idea behind what I am doing, let us suppose that we still used stone scoring, but we gave a 2-point bonus for each group’s 2 eye points. If we add the restriction that players get an equal number of scoring opportunities, then in a tied game, the entire board will end up filled. In this instance, your “last competitive play” is obvious because then black would be winning an otherwise tied game by virtue of the board being odd. So you would remove that final black stone, the board would have a single empty point, and without fractional komi, it would be a tie.
The math I did preserves this relationship at every step of the game regardless of what players end up doing. So the strategy stays identical to both your idealized version of Taiwanese rules and is obviously close to the very oldest versions of the game from which area scoring arose. All I did was preserve that extra bit of information by making the scoring still care about it. By design, no strategy was changed along the way.
If you think that there is a problem with the strategic implications of area scoring (as opposed to just the “rounding” that it does), this will not satisfy you. But if you otherwise like area scoring, and are solely concerned about the scoring granularity, then what I did fixes the issue.
Edit: NB: If you are solely concerned about the whole “players must fill in dame” thing and don’t care about the more meaningful strategic differences vs territory, and you are fine also correcting for odd seki along the way (or if you want to correct for odd seki and are willing to change dame behavior), then I think you can modify things track this and propagate that parity into the rest of the conditions such that this dame thing will not matter no matter what white does.
I’m unsure of whether you need to say that say that all neutral points (dame and unplayable seki intersections) are split evenly between the players with any odd point being counted for black. (Forcing the board into odd parity, and then correcting it back if necessary using our usual formula.) Or if it is better to distinguish the two cases with further rules.
I don’t think the extra bookkeeping to essentially fill dame hypothetically is beneficial. And I haven’t worked it our rigorously. Though I will attempt to do so and edit this if time allows and you haven’t already responded.
At first glance, to deal with dame, I think you end up needing to track who has passed more and swap the compensation rules accordingly. But I am unsure. Tracking 4 variables is reasonable. Tracking 8 is quickly getting out of hand. This “swap it around” idea works in your example, but I would need to work through the math. And I worry that it goes wrong in more complex situations without having the math to confirm my hunch.
Though, again, odd seki aside, all you are accomplishing here is to play out the normal area scoring dame filling sequence virtually. Done correctly, the only change you get in all of this complexity is that when there is an odd seki on the board, the rules flip to be as-if there was an even board. But you could just do that directly without all of the extra book-keeping complexity of trying to also make the dame work out no matter what. I don’t think expecting players to fill dame is terribly unreasonable.
Again, I think the goal to have area scoring that doesn’t round the score to odd/even (depending on board parity). I don’t think trying to achieve specific aesthetic outcomes of no strategic import should matter. Is there some good reason to care about “fixing” this that I’m unaware of?
Regarding odd seki specifically, I’d appreciate your help. I want to make sure that I’m undoing the rounding in the correct direction and the pages on sensies don’t do a good job of illustrating how the area score rounding works on an even board.
So, ignoring normal dame and sticking with “same strategy as area, we are just going to do a calculation to avoid the rounding”, is it correct to say, that if there is an even seki at the end then my proposed rules (or some suitable mathematically equivalent version apply. But if there is an odd seki, then it is black who should be compensated when play parity is odd (cases 2 & 3)? Or do we instead need to compensate white still, but now in cases 1 & 4?
As I cited above, there are clean version of actual territory rules that do not have the complexities of the Japanese and Korean professional rules. In the worst case, they require 2 playouts to resolve a complex dispute, but they are easily implemented and they do not require that players learn a 3rd set of end game considerations like the various hybrid Ikeda-like variations.
Per above, to avoid aberrant behavior with Ikeda-like rules, either you have to use a different ko rule and have 3 passes to end a phase, or you have to use multiple phases with different rules and special ko-pass rules like Kata Go. This cure seems worse than the disease.
I think that practically speaking, given that there is no good online implementation of territory scoring, that area is better for online play, at least until a high enough level where that marginal point of difference in an otherwise tied game starts mattering somewhat regularly.
Regardless, the messages much further up in the thread, do make a strong case for why we care about the parity difference that territory accounts for and area scoring does not. I was initially skeptical, but was ultimately convinced.
Which specific territory rules would you recommend for online play (assuming any necessary modification to the software can and was made)? This is mostly an hypothetical question because I do not see the will in OGS maintainers to implement an encore phase or similar but let us set that aside for the sake of discussion.