This is a stable seki, right?

Just checking - this is a stable seki, right? Despite it being interestingly asymmetric in the sizes of the big eyes?

And neither player can play any further in the seki, right? All three dame must remain open exactly as they are?


Oh, maybe white has some way to sacrifice the smaller group to make the larger one live indepenently? (I guess that’s not profitable though).


You can fill E1 and E2 and white still is unable to fill F7, but black now captures at H2.

Also black can capture directly at F7, H8.

Black has 1 eye, while’s groups don’t.


If black plays E1, then white captures at H8 to start a race, and then what does black do?

And if black instead plays F7, white captures, and black plays the critical point J7 to try to kill, white plays E1 and races to fill the bottom liberties, and black is also in trouble, right?


I guess you are right. I missed that the captures free up liberties for white.

1 Like

You had me nerdsniped for a bit, but I agree. Stable seki. :blush:

1 Like

There are some similar seki positions listed here:

(figure 15 only differs in the shape of the four stone nakade)


Nice, is this already included in Other Go Resources?

1 Like

Something like this you’re saying?

Yeah, stuff like that.

This feels like the kind of position where if it ever turned up in a game, it would be incredibly easy to lose by not realizing that this position was the terminal position.


don’t finish the game if seki is not clear enough. I mean, you should capture all capturable stones. looks what happen in this game, seki on whole board
J7 stone should be captured by white before both player start scoring. it affect the point.

That’s exactly the point, there are no safely capturable stones in this seki. The J7 in your linked game is capturable for an extra point, but here it will cause the group to die. The advice to “don’t finish the game if seki is not clear enough” is exactly the thing that would cause someone to accidentally collapse in this position. :slight_smile:


are you kidding… white is dead.
Demo Board (see the variation in the comment)

in that game, even if it seems seki like you said, black should keep trying.

if both player pass and assume seki = black lose (no territory)
black try to kill but failed = black still lose (all black stones captured)
black try to kill, fortunately white made mistake = black win
black try to kill and success, found divine move = black win

you got my point? :slightly_smiling_face:

are you kidding… white is dead.

Based on your variation, did you mean to say “black is dead”?

Ah, cool, I understand what you were trying to say now. Yes, you are right that if you’ve accurately counted the whole board and determined that you are definitely losing, then because you have nothing to lose, you are free to try more moves to win. Even if those moves result in the death of your own groups that should be alive, you can still try if you aren’t sure, since there isn’t a penalty for losing by even more.

Still, this position is interesting because it is a finished position but is very hard to recognize as a finished position. Your own variation where you thought that black (or white?) is dead proves that how easy the position can fool people. You might have gotten a little fooled here! Neither side is dead, your very first move for black is wrong, black lives if they pass instead.

Anyways, you’re right, in that way it doesn’t matter then for finishing the game (so long as you trust yourself to count the winner accurately, and you have time to do such a count to determine if you should escalate and risk more). It would matter more prior to finishing the game - if in a once-in-a-lifetime miracle such a position was potentially going to occur in a fight in your game, it would of course matter if you could judge it correctly as a seki by the tricky realization that both players should stop playing at that point, rather than misevaluating one side or the other to be dead. So that you could judge the result in comparison to other moves in the fight that might lead to different results.


Ah, I see, you put the variation in the comments in the game rather than into the actual demo board variation tree.

I see where you might be making a mistake now. Your move for White 4 is wrong. White should play E1 or E2 instead of capturing the 3 black stones.

This E2 is where white should play instead of H2. Now black will die. This means black’s original move was killing themselves, black should instead pass at the very start, leaving the position as a seki.


ah ok. now I can see that… thanks.

thats explain my third option “black try to kill, fortunately white made mistake = black win”
in this case, I (play white) made mistake by capturing black 3 stones. then black win, so lucky


Open question for those enthusiasts out there who love to collect pathological positions and tactical beasts and such:

As a rule of thumb it seems that in capturing races involving almost-filled big eyes, one should delay capturing the stones that are almost-filling the eye as long as possible. For example as shown above, white avoids prematurely capturing the stones in their big eyes - only doing so unless forced to by black’s atari, or until there is only one shared liberty so that now capturing is required before filling it.

Is there any case where capturing the almost-filling-stones “early” is strictly better than doing so later? Maybe you can get this to happen if there’s some sort of ko possibilities involved as well, such as using it as a ko threat in an active ko or prematurely gaining liberties to remove possible threats by the opponent in a ko that is about to form? If so, are there any cases where early capture is strictly optimal for the capturing race tactics that don’t involve ko?


I haven’t found any way of achieving this, but I think there should be such a case.
I use the concept of ​ protected liberty ​ from ​ .


It would suffice to have a region with [exactly one sufficiently large white group inside] and
[a black chain C which forms at least part of the region’s boundary
and has at least two outside liberties], such that on the inside,

Black moves to gote local seki of the inside, where [Black can simultaneously keep the
local seki and have the region give chain C sufficiently many liberties] but [White can
simultaneously keep the local seki and have the region give chain C at most 1 eye]
White has 2 options, one of which is [sente vs gote doesn’t matter, but otherwise local seki as above].
White’s other option is a fight where [if one of C’s outside liberties is protected
then White’s sufficiently large inside group dies] but [if all C’s outside liberties are
sufficiently far from being protected then [White’s sufficiently large inside group lives
with sufficiently many points, even if a Black tenuki would give chain C two eyes]].

Suppose complex_region is a region satisfying the above conditions,
and let large_white be the sufficiently large white group inside it.
The almost-filling stones would be black stones which [form an unsettled eyeshape,
rather than a dead eyeshape] and [are inside a white chain that is in
turn completely enclosed by chain C, but is not part of complex_region.
I call that white chain small_white, and claim in the combination of the above regions,
White-to-move achieves at least Black’s choice between
[large_white lives with sufficiently many points] and [large_white and small_white both live].

To achieve at least that, White can capture the almost-filling stones inside small_white.
If Black doesn’t bring that back to 1 eye, then White follows up by making
making it 2 eyes, resulting in small_white and large_white both living.
Thus, Black responds inside small_white. ​ This leaves small_white with
at least 2 eye-liberties, so it filling its outside liberties won’t be self-atari.
In particular, with respect to complex_region, C’s outside liberties are not quite protected.
How far they are from being protected is given by how big small_white’s eye is, so
assuming that eye is big-enough, White starts and wins the fight inside complex_region.
This results in large_white living with sufficiently many points.

Suppose White’s first move was not capturing the almost-filling stones inside small_white.
If it was starting the fight, then Black will ignore when White captures those almost-filling stones.
This makes one of small_white’s outside liberties into a protected
liberty for chain C, so Black wins the fight and large_white dies.
If White’s first move way anything else, then Black makes local_seki in complex_region.
That simultaneously gives chain C sufficiently many liberties which are not
shared with small_white, so small_white dies and large_white gets at most seki.



(However, I do not know of any examples for complex_region.)


There are a few other weird sekis in the same vein, on that page:


any capture race is seki if both players are afraid to continue