Is an endgame ko worth more in Chinese rules than in Japanese rules?

[dragon-devourer]

A situation arose in a game that caused me to wonder…

Is an endgame ko worth more in Chinese rules than in Japanese rules?

I think yes. My explanation is below. I would be interested to hear other people’s opinions on this.

The Position

Consider this position:

Assume all surrounding stones are alive, and all endgame plays have been done except one final 1 point gote somewhere else. Black to play.

Variations

1. Black takes gote, white wins ko

2. Black takes gote, black wins ko

3. Black wins ko, white takes gote

Results Under Japanese Rules

1. Black gains gote, white gains ko capture. Even

2. Black gains gote, no difference from ko. Black gains 1 overall

3. No difference from ko, white gains gote. White gains 1 overall

So black cannot lose points by taking the gote move but will lose a point by winning the ko immediately.

Results Under Chinese Rules

1. Black gains gote, white gains two extra area points. So White gains 1 overall

2. Black gains gote and 1 extra area point. So black gains 2 overall.

3. Black gains 1 extra area point, white gains gote. Even

So black may gain or lose if black takes the gote move depending on who is ko-master, but it’s an even exchange if black wins the ko immediately.

Conclusion

Under Japanese rules, black should play the gote move regardless of who is ko-master. But under Chinese rules, black should only play the gote move first if black is also ko-master.

Or, more generally, under Japanese rules, an endgame ko is worth 1 point if the player to capture first is ko-master, zero otherwise. But under Chinese rules, it’s worth 2 points if the player to capture first is ko-master, 1 otherwise. Hence, an endgame ko is worth more in Chinese rules than in Japanese rules.

What do you think of this case? Have I interpreted it correctly? Do you know of any other related situations? Can you think of any counter examples? Cheers

[claire_yang]

This tactic has a name in Chinese Go terminology called - 粘劫收後, filling (by winning) the ko and get the last play/dame.

It’s under the circumstance where one side can start the ko and has more (valid) ko threats and can fill the last dame. Keep the ko going but don’t actually use ko threats, instead playing “neutral points” as if they are ko threats. If the opponent play the ko, keep playing ko. If the opponent fills dame, you also fill dame (treating dame move like a ko). When you get the last dame, and still have more ko threats (the opponents run out of valid ko threats, but only loss-making threats), the opponent would be forced to pass (or risk losing more and more), and you can finally fill the ko to get extra points.

There are even more extreme cases, where the ko are step-ko, and if one side has much more ko threats, when the opponents run out of (valid) ko threats, they would have to pass several steps, and under the Chinese rule, it would be winning step number more points.

[elsantodel90]

If you are comfortable with territory-scoring, you should think of the situation as “scoring a point of territory inside the ko”. That would be, black getting the central “3-3” intersection in your picture as extra territory when winning the ko.

Let’s ignore area scoring for now, and imagine we play by territory-scoring rules (you can think we are using Lasker-Maas rules, or AGA-rules and counting territory for example, or maybe “Japanese rules of a different hypothetical universe”).

Normally, there are many trully neutral non-ko-threat dame in the board, say like 8-12 dame maybe (really depends on the type of game, like big moyo vs many fights etc). The last ko is fought before filling those dame, it is obviously more valuable than just dame. Often, you will not win this final ko-fight having lots and lots of extra ko-threats, but barely win it, maybe you will have just 1-2 extra ko threats compared to your opponent. So, the situation will be that you take the ko, your opponent fills a dame (he might pass but what for? playing dame is at least as good, maybe he gets to win the ko somehow by keeping playing instead of passing. A proper full game includes dame filling ). Now you might fill the ko and that’s it, you win it, you get the “last prisoner” you captured, but nothing else. You might fill a dame just to “show off”, because you know you still have one more ko threat. But if your opponent keeps filling dame, you can’t “show off” forever: you will HAVE to fill the ko, or else lose it (and your opponent might just fill it and complete it, not wanting to “show off”. You now lost the point of that “last prisoner” in the sequence of ko-captures).

Now, imagine the same situation I just described, but there is no dame, or very little dame compared to your ko-threats. Now you make the ko-capture, your opponent has no ko-threat… and there is NO dame! So, he HAS to pass (if he does not, he fills his own territory, losing a point… that is no better). Since he passed, you pass, to finish the game by two-consecutive passes. WOW! The ko is still open! You did not “fill” the ko! But you could do this because you had so many ko threats, that you managed to keep the ko open even until all dame were filled, and your opponent could do nothing, not even dame, other than fill his own territory and pass… so you pass to.

In this hypothetical situation… Don’t you actually deserve the extra point of territory of the unfilled ko intersection? You clearly controlled it, you had enough ko threats to keep it open!

Well, this is “the actual reason” behind the scenes for this fight: you are fighting for this extra point of territory! Lasker-Maas rules and AGA-rules when counting territory (with pass-stones) give you this extra point: since your opponent passes, he gives you a pass-tone, but THEN you can fill the ko instead of passing, so then your opponent has to pass AGAIN. The extra passing-stone that he gives you effectively works exactly the same as if you could just pass and reclaim this extra territory point.

Under chinese rules, the effect is basically the same because you just “fill” the ko, which is now YOUR territory, and it has no cost! Filling your own territory AFTER all dame are already filled, does not change the score in area-scoring rules (if you do it before ALL dames are filled, you DO lose points: as you are just territory-filling for a net +0 score, while your opponent fills a dame, for a net +1 score. It is just like japanese! you lost one point for prematurely filling your own territory!). So, under chinese rules, it works just as if you could count the ko-intersection as your territory, for one extra point if you manage to not only win the ko, but also leave the ko open until all dames are filled.

So, it seems that these fight makes sense in territory and area scoring rules, this is not a difference of area vs territory, but whether this last “ko intersection” territory can be counted or not. What do the Japanese rules say?

Well, this was a famous rules dispute involving Go Seigen twice. The Japanese ruled against Go Seigen BOTH times, even though Go’s position in the dispute was reversed each time! It seems that the rule at the time was actually “Go Seigen loses the point”. Rule disputes involving Go Seigen at Sensei's Library

Modern Japanese rules says that you CAN’T reclaim this extra territory point, even if you have an infinite number of Ko-threats. You just have to fill it, it is mandatory, sorry. So this is the real reason why there is no “dame-fight” in Japanese rules but there is in Chinese-Rules (and other chinese-like territory definitions, like that of Lasker-Maas territory rules).

[elsantodel90]

Additionally, you might notice that in Chinese rules (or also AGA rules, but not in Lasker-Maas which work as territory rules!), depending on the parity of dame, this final “dame-fight” might be useless to fight because even if you win, the result does not change. In effect, if there is an EVEN number of dame apart from the ko when you can choose to fill it, then no matter how many of these there are, whether you fill the dame before or after the ko, you will always get half of them and your opponent the other half. Looking at it from our “territory-perspective” from the previous post, this is due to the “parity rounding” effect of area-scoring: The area score difference is (normally: flip even/odd if there is an odd number of points as dame in seki) precisely the territory score difference, but adjusted so that black gets one extra point IF the territory-score is even (this is because, the final area-score must be odd, and when the score is not the same as the territory-score, it is because black played one final dame for one extra point). So depending on the parity of the score difference (which will in turn determine the parity of the remaining dame), an extra one-point of territory might change the area score by either +0 or +2.

Again, this situation that “the fight is sometimes relevant and sometimes it is not” can be interpreted by territory-counting by the fact that only ONE territory point is in play (and the dame-parity of filling the ko before or after dame-filling is the same). This reinforces the interpretation that it is actually “the point of territory in the ko” that is under dispute.

[elsantodel90]

As a recap of different rules (assuming positive score is in black favour):

• Japanese Rules: you are never allowed to keep that extra point of territory, so the only relevant fight is for the ko itself, you win or you lose. So the ko has only two relevant results: 0 points for white wins the ko, 1 point for black wins the ko.
• Lasker-Maas rules, or “hypothetical Japanese-like” such that you were allowed to claim the extra territory from the unfilled ko: you are allowed to keep that extra point of territory if you win and there is no dame (your opponent has to pass first). So the ko has four possible results: -1 points for white wins the ko after filling all dame. 0 points for white wins the ko (normal way, fills while there are dame). 1 point for black wins the ko (normal way). 2 points for black wins the ko after filling all dame.
• Most (all?) area scoring rules (Chinese, AGA): Just like Lasker-Maas rules but subject to the normal “parity-rounding”, so some of the four possible results will actually give the same final result in the game, depending on the parity of dame. You can always play for maximum score as if playing by Lasker-Maas rules, and you will be getting the best score in Chinese Rules too (but you might be unnecessarily risking a pointless fight, by trying to get an extra “+1” territory that due to rounding does not change the area-score). More precisely, depending on parity of dame there are two type of final kos: A) Odd number of dames remaining when a player might fill the ko and win it: In this case, who wins the ko “normally” does not matter at all (both options give the same final score). But, if any player manages to delay and win the ko after all dame are filled, then that player gets 2 extra points compared to the result of “normal” ko filling by either player before dame. B) Even number of dames remaining when a player might fill the ko and win it: In this case, fighting for the extra point is useless as I said before, the only thing that matters is actually winning the ko (and it makes a difference who wins it), so this behaves like the Japanese-rules case.

[Feijoa]

@dragon-devourer

• Black gains gote, white gains two extra area points. So White gains 1 overall
• Black gains gote and 1 extra area point. So black gains 2 overall.
• Black gains 1 extra area point, white gains gote. Even

I think something about your counting is off, since under Chinese rules the result changes only in multiples of two. Here, I made a review, is this the kind of thing you were talking about?

Endgame ko

For this example I counted:

Variation	Japanese	Chinese
1. Black takes gote, white wins ko	B+2	B+1
2. Black takes gote, black wins ko	B+3	B+5
3. Black wins ko, white takes gote	B+1	B+1

So it seems okay (but maybe neutral) to play gote first under Chinese rules, too.

[claire_yang]

@Feijoa

Variation	Japanese	Chinese
1. Black takes gote, white wins ko	B+2	B+1
2. Black takes gote, black wins ko	B+3	B+5
3. Black wins ko, white takes gote	B+1	B+1

So it seems okay (but maybe neutral) to play gote first under Chinese rules, too.

You forgot that in Chinese rule, there is one more komi. The trick only works by forcing your opponent to pass, so you can play one extra move on the board. (think about in your case if in Japanese rule it is 3.5 komi, and in Chinese rule 4.5 komi, and black actually has one more ko threat at E2)

[Feijoa]

I was using zero komi for simplicity, since it’s just to compare the relative value of moves within a given rule system.

Are you saying something different from my #2 above? Here’s what that branch looks like:

Move 4 is a pass for White. This does seem to show that the ko has more value under Chinese rules, but I didn’t agree with this statement:

@dragon-devourer

But under Chinese rules, black should only play the gote move first if black is also ko-master.

I’m kind of fuzzy on this so wondering if there’s a way to set it up for a different outcome, but in this case isn’t it always okay for Black to play the gote move, under any ruleset?

[claire_yang]

@Feijoa

I was using zero komi for simplicity, since it’s just to compare the relative value of moves within a given rule system.

Since the board has an odd number of intersections, without ko and seki, black will play the last dame. If two systems would to give the same result, either black needs to give 1 more komi, or force both sides to play equal amount of moves. Without the adjustment, they naturally give different results. In a sense, black always has “sente” for the final move. Ko changes the sente/gote is the reason why when one side cannot win and run out of ko threats, naturally has to pass.

In general, this technique it’s not about black gote move, but winning the ko throught the use of gote/dame move as ko threats (in Chinese rule, most dame is 1 point gote yose). If it is a larger gote value yose, then the other ko value need to be considered. In your example, winning the ko is only worth 1 points in Chinese rule, but C1/D1 is 2 points gote yose. In that case, black’s option is better to win the ko using that as a threat. Even if there are only dame moves left, as long as there is an odd number of dame moves left, they can be used as threats (since the last one is worth one more, think about the whole dame exchange process as one giant 2 point gote yose in Chinese rule).

This is where this technique gets its name from “(at the end) connect the ko (by being able to) play (the last) gote dame (as ko threats, when you have other larger and more ko threats left)” and get one more point as a result (under Chinese rule). In your first variation, black has one more ko threat at E2, so the result will be in Japanese rule B+3 (I originally type B+2, it is a typo), and in Chinese rule B+5 (black’s best action in Chinese rule is not to pass and let white win the ko and play the last move, but to keep the ko, as long as there is at least one larger threat left)

[elsantodel90]

@claire_yang

Since the board has an odd number of intersections, without ko and seki, black will play the last dame.

This is false as written. Maybe you meant that “sometimes” black will play the last dame. Consider these two games (these are academical nonsense games, but check random real games and you will find real examples of both kinds):

In the first one, black plays the last dame. In the second one, white does (well it is not a dame, but it is the last meaningful move. It has the same effect. Just add a few dames in the middle to fill at the end and the result is identical with “actual dames”).

The second one should give identical results for area and territory, because white played last. And indeed, it is +9 for black both counting territory and area. Notice that it is odd: when the territory-score is odd, then the area scoring is the same (unless there is odd seki: then the property flips, and an even territory-score gives the same area score).

The first one should give one more point to black in area scoring, because black played the last move. And indeed, it is +6 for black when counting territory but +7 for black when counting area. Notice that the area score is odd again: when the territory-score is even, then the area scoring is +1 for black, so that it becomes odd again (unless there is odd seki, then the property flips, and an odd territory-score gives +1 area score).

So unless odd seki, area scoring with 7,5 komi games will give the exact same result as territory scoring with the same 7,5 komi. The reason to use 7,5 instead of 6,5 is that due to the rounding explained above, when territory score gives +6 or +7 for black, in both cases the area score becomes +7. So a komi of +6,5 would have almost no difference to 5,5 due to this rounding effect (the difference kicking in only in games with odd seki). In this sense it is that territory-scoring rules are more “precise” or more “fine grained” in the score, without changing strategy, than equivalent area rules (equivalent means that territory/life etc is defined the same. Japanese is not equivalent to Chinese in this sense, but AGA and Lasker-Maas are practically equivalent in this sense, with Lasker-Maas being a territory-scoring version and AGA being an area-scoring version).

[elsantodel90]

To emphasize: “the parity of dame” and “the parity of the final territory score” are linked together, and one determines the other (unless players pass early). That is why you can use the rule: “count territory-score, say it is X, then if X happens to be even, final area-score result is actually X+1, always in favour of black, but if X is odd then the area score is the same X”. And you do not need to “check” who played last, it has to agree with this rule (unless players pass early, odd seki, etc).

One critical property of that is that better territory-score implies better or same area score, so playing for maximum territory is optimal for area score too, but not vice-versa (as for the same area score, there are two different territory-scores possible. Area score has “less resolution”, so to speak). (Always assuming that you count territory with the same criteria, of course. In these final dame-kos, this means counting the point of territory inside the last ko, as explained before).

[elsantodel90]

@Feijoa

|Variation|Japanese|Chinese|

|1. Black takes gote, white wins ko|B+2|B+1|
|2. Black takes gote, black wins ko|B+3|B+5|
|3. Black wins ko, white takes gote|B+1|B+1|

When also counting the extra-point-of-territory inside the ko in the territory count (the one modern Japanese rules do not allow one to keep) the above table for the variations would become:

Variation|Territory|Area

1. Black takes gote, white wins ko | B+1 | B+1
2. Black takes gote, black wins ko | B+4 | B+5
3. Black wins ko, white takes gote | B+1 | B+1

Which matches the rule

[claire_yang]

@elsantodel90

This is false as written. Maybe you meant that “sometimes” black will play the last dame. Consider these two games (these are academical nonsense games, but check random real games and you will find real examples of both kinds):

Ya, I should say, “and when black plays the last dame”, not “will play”. Grammar is hard when my mother tongue is tenseless. In general, when this technique is applied, the situation has more to do with the number of dame left to be even or odd numbers to determine who could initiate it, and it’s not always a benefit. Most times, you need to check if additional ko threats in the followup of several yose existed for this to actually work, even ko threats that lose points (損劫).

@elsantodel90

One critical property of that is that better territory-score implies better or same area score, so playing for maximum territory is optimal for area score too, but not vice-versa

This is not true, as in the tactic of 粘劫收後, it is possible to play point losing ko threats, that definitely make territory score worse, but not in area score (since no matter how many captures they don’t count), and still able to win the game that get 1 extra with the price of none. But if playing for maximum territory score, it could not be possible. The issue will be even more obvious when step ko is involved for this tactic.

[claire_yang]

@elsantodel90

When also counting the extra-point-of-territory inside the ko in the territory count (the one modern Japanese rules do not allow one to keep) the above table for the variations would become:

This I don’t understand, those variations have already resolved the ko, Can you point out where in the modern Japanese rules that discount the capture in ko not as part of the territory score?

[elsantodel90]

Sure. In this variation, after 1-2-3-4-blackpass-whitepass (note that black would not be forced to pass if there were still dame, could play dame still, creating a fight-not-to-pass-until-after-dame in territory-scoring), the marked square is the point in dispute (modern Japanese rules never allow to keep this point, even if you have infinite ko threats, but Go Seigen was twice involved in a dispute about it).

Note that area scoring definitely counts that point as territory: you can actually play it instead of passing, to avoid a dispute altogether, but because “a play inside your territory at the very end without dame does not change area score”, it is clear that the area score is the same as if counting the above position and considering that point as territory.

In the variants where black manages to keep the ko open until after all dame, the extra point in question is D6.

[elsantodel90]

@claire_yang

This is not true, as in the tactic of 粘劫收後, it is possible to play point losing ko threats, that definitely make territory score worse, but not in area score (since no matter how many captures they don’t count), and still able to win the game that get 1 extra with the price of none. But if playing for maximum territory score, it could not be possible. The issue will be even more obvious when step ko is involved for this tactic.

I am interested in an example of this. My claim is that there should be no difference, IF territory were counted using the same criteria, and players do not pass early. For example in a step ko that a player manages to win with extra ko threats after dame, the “extra points of territory” just like the above that he “gets to keep without needing to defend” should be counted for the equivalence to work. And in the same way, points inside a seki and even one-sided dame must count as territory for the equivalence (any point that a player gets to play for free while the other is forced to pass and can do nothing better should count as territory for this).

[claire_yang]

E2 is a ko threat white cannot ignore, and white will have no valid ko threat after that and black can connect win the ko. Black passing is not a good resolve for black using the technique, that’s the opposite of what black should do in Chinese rule.

[claire_yang]

@elsantodel90

I am interested in an example of this.

The variation I showed in the previous reply is also an example, black can play a point losing ko threats (in Japanese rule) to win the ko (in Chinese rule)

And here is a real game example, a game between Peng Quan (3p from China) and Yoo Changhyuk (9p from Korea). The last ko, black has one more point losing ko threats as marked in triangle, but white has none. In territory rule, this ko threat isn’t going to work at all, white can just answer it, and when black take, just pass, it doesn’t matter if white doesn’t have that point as a territory or not, it’s W+0.5, but in Chinese rule, when white pass, it is B+1.5, as the actual game result.

Peng Quan vs. Yoo Changhyuk

[elsantodel90]

But you are not counting the B19 point of territory for white when winning the ko, nor the C19 point of territory for black when winning the ko in this analysis. This point is not allowed to keep in modern Japanese rules, but as I say, since defending is effectively free after dame in Chinese rules, those count it as territory implicitly. If that point is counted, we have territory and area rules equivalence (with the corresponding rounding effect) again. For example, under lasker maas rules which are territory scoring rules, it would definitely be important for black to fight and win that ko, as that would gain him 3 points by comparison and not just 1 as in Japanese.

Let’s check: if the shown position was actually the final position and everybody passed there (or white managed to play B19 under area rules, that would not change) the result is W+2,5 both counting area and counting territory (as long as you count the B19 point as territory: that is why Japanese score here is W+1,5 instead, as in Japanese you will not be allowed to finish with B19 still open and count it). Without Komi it is B+5 which is odd, so that is why territory score is the same as area score.

Now in case black uses ko threat and wins the ko, since black is winning the ko after all dame, the final position can be taken to be that after B19 is taken, but C19 is still not filled, as filling it is free when there are no dame remaining and thus can be counted as black territory. Then in such position, the territory count must be 3 point better for black than before: one because of the B19 intersection that white loses, one because of the captured ko stone, and one because of the C19 intersection that black wins. So the territory score is now B+0,5, which is B+8 without Komi. As it is even, because of the adjusting-to-even effect (which is always in favour of black, as explained before) the actual area score must be B+9 which corresponds to B+1,5 after Komi. And indeed, that is the result we get after counting area directly.

The critical difference is counting C19 as territory for black to get to B+0,5 when counting territory. That point is not allowed to keep in modern Japanese rules, so the Japanese territory count is actually W+0,5, and so black still loses. But the area and territory scoring actually agree here, it is just a matter of which intersections count and not count as territory in each ruleset (The chinese definition of territory is sort of implicit: any intersection where you play safely after dame were filled can be counted as your territory, thus effectively “the point inside the final ko” can always be counted).

[elsantodel90]

@claire_yang

E2 is a ko threat white cannot ignore, and white will have no valid ko threat after that and black can connect win the ko. Black passing is not a good resolve for black using the technique, that’s the opposite of what black should do in Chinese rule.

In this example position, the extra point of territory for black that modern Japanese rules do not grant him is D6, because white has to pass now and if black can pass then we have D6 still open and the dispute on whether black can keep it or not as he had enough ko threats. Modern rules says no, but Chinese rules (implicitly, by giving zero penalty to that territory-fill after dame are played) says yes.