In this particular position, you made it even better: Black can kill in N+1 moves, or live in N moves!
5 Likes
This common seki shape has some interesting properties, both in Chinese rules and in Japanese rules.
Under Japanese rules, it’s a mistake for black to play B - white gains a point by capturing that stone. Black should therefore just leave the situation as it is. White can choose to also leave it, or to play at A and initiate a trade, which would not affect the score: white gains 4 points in the corner (2 territory + 2 captures), while black gains 4 points around B (2 territory + 2 captures).
Under Chinese rules, it would be a mistake for white to play A. In the trade, black’s area becomes 1 point bigger (from 3 points in the corner to 4 points around B), whereas white’s total area stays the same. Therefore, white should just leave the situation as it. Black can choose to also leave it, or to play at B, which is forcing and does not lose any points.
Keeping all that in mind, consider this position:
Black to play, no komi, what’s the result?
Does the answer depend on the ruleset?
5 Likes
Chinese rules
Variation 1 (suboptimal for White): Black takes the ko, White can play G2 like a ko threat; Black plays B6 as a ko threat, which loses one point locally (but the ko is big so it’s worth it); Black captures the A6 and B7 stones while White captures the E6-E7 stones; Black wins the ko.
Finally White has 24 points and Black 25 (there is no seki in this variation so it sums up to 7x7=49).
Black wins by one point.
Variation 2: Black takes the ko; White plays D7 as a ko threat; Black has to answer the ko threat and has no other ko threat. White wins the ko.
Finally Black has 23 points, White 25 (there is seki in this variation so it sums up to only 48).
White wins by 2 points.
Japanese rules
First let’s count prisoners, which we can’t see on the board. We notice that there are 16 White stones and 16 Black stones and White played last, so Black and White have the same number of prisoners.
Variation 1 (suboptimal for White): Black takes the ko, White plays G2 as a ko threat, Black answers, White takes the ko, Black plays B6 as a ko threat, White takes E6-E7 while Black takes A6-B7, Black wins the ko.
Score: Black 4 prisoners + 7 territories; White 3 prisoners + 6 territories; Black wins by 2 points.
Variation 2: Black takes the ko, White plays D7 as a ko threat, Black answers the threat; the top remains seki; White wins the ko.
Score: Black 2 prisoners + 5 territories; White 1 prisoner + 7 territories; White wins by 1 point.
Conclusion
White always wins the ko and the game.
1 Like
Good analysis, but black can actually do better! (under both rulesets)
Chinese rules
Variation 3: Black pre-emptively plays at B6 before starting the ko; White D7, Black captures the two stones A6-B7, White captures the two stones E7-E6. Then Black takes the ko. White cannot play G2 as a ko threat (otherwise Black would connect the ko at D1 and if White tries to kill at F2, Black cuts at B2 and wins the semeai). White loses the ko.
Score: 24 for White, 25 for Black, Black wins by 1 point.
Japanese rules
Variation 3: Same variation as variation 3 of Chinese rules.
Score: Black 3 prisoners + 7 territories, White 2 prisoners + 8 territories. The game is jigo!
3 Likes
Correct!
Except...
White can play the G2 ko threat. Note that white already has two eyes at that point so there is no semeai to lose! However, black still wins the ko using the (newly created) ko threat at E6, so the final scores are exactly as you said.
The thing I found interesting is that under area scoring it can be worthwhile for black to give up a point in the upper left to win the ko - it seems like this could easily happen in a real game if this seki shape occurs. White should try to prevent this by throwing in earlier - in the position showed here, white’s last move at G5 should actually have been at D6, which would have won the game for white.
That even holds true for Japanese rules, where white normally shouldn’t play D6, but here it’s worth it. (although in this particular position another way to win the game for white under Japanese rules would have been to connect the ko - we can make a similar position where throwing in is the only correct play, for instance by removing the open endgame on the right and having the ko be already captured by black - then white must throw in before taking the ko).
3 Likes
Variation 3: Black pre-emptively plays at B6 before starting the ko; White D7, Black captures the two stones A6-B7, White captures the two stones E7-E6. Then Black takes the ko. White cannot play G2 as a ko threat (otherwise Black would connect the ko at D1 and if White tries to kill at F2, Black cuts at B2 and wins the semeai). White loses the ko.
Not sure I understand. When white plays G2 as a ko threat, if black ignores it and white follows up at F2, there’s no semeai to win because Black is dead with only one eye, while white has two eyes (their original eye in the upper right, plus the eye from capturing E7-E6).
So G2 is a good ko threat, black must answer. It’s just that black has their own ko threat by throwing back into E7-E6.
2 Likes
Oops, you are entirely right.
2 Likes
When I was reading it in my head, I totally missed the fact that black did have an under-the-stones ko threat, by the way. Under the stones is hard. 
4 Likes
I hope this is not too OT here, since passes (or their lack) seem a key point for this topic. From a recent L19 thread - B to move, no komi no prisoners, W normally wins with his +1 territory:
I think this example shows why such “simple” rules (ie. without ban lifting passes) are only considerable under area scoring. There the players usually have substitute - plenty of 0 pt ban lifting moves, plays into territory. So ko play can still follow a natural path, and passes only matter on special, crowded and small boards, not in daily practice. (Same with territory + area encore hybrids.)
Under true territory scoring, however, only dame could substitute, which are not always available (or enough). The players often have no means to change the board without losing points. So without real ban lifting passes, ko play could pathologically diverge from both area scoring and common sense - even on 19x19 board, and even in everyday positions like above.
B could gain a point / prisoner by taking the unwinnable outer ko and freeze the board - which makes little sense under area scoring. So the territory and area game can only remain coherent if territory scoring has ban lifting passes - it doesn’t work without them.
Applying that on its own would let an otherwise-loser
churn a double ko to keep the game going forever.
To avoid such churning, there are two good fixes I see:
(S) Spights superko rules Spight Rules at Sensei's Library , or something similar
or
(L) a long-cycle rule, in addition to (not instead of) the basic ko rule:
Repetitions of the 3-tuple
board coloring , player who’s turn it is , the 0 or 1
intersections that are banned by the basic ko rule
such that in the cycle [[the players passed different number of times]
or [the players [played captures that would’ve been illegal if
pass-pass didn’t lift basic ko bans] different numbers of times]]
end the game: If the players passed different numbers of times in the cycle, then
whoever passed more in it wins. If in the cycle, the players passed the same number of times
but [played captures that would’ve been illegal if pass-pass didn’t lift basic ko bans]
different numbers of times, then whoever played fewer such captures in the cycle wins.
(The latter is at least an approximation of, and maybe even equivalent to,
there are N 0 pt ban lifting moves, where N is finite but otherwise sufficiently large.)
I think a lone double ko is not really usable for repetition since it includes periodic two passes and stops. And whatever the rules are, they unlikely allow infinite resumptions (esp. from the same position and without score change).
I now see I was assuming the sequence would be just
take ko , pass , pass , take ko
, rather than
take ko , pass , pass , pause , resume , take ko
or
take ko , pass , pass , pause , resume , pass , take ko
.
However, even if resumptions are used to enable
the ko recaptures, there would still be the issue of
What if each player would be seeking a resumption infinitely often,
but one is doing so more times per cycle than the other?
.