Basic Life and Death Scenario

Screenshot 2022-08-10 165716


Sorry if this is a simple question. I thought this was seki but AI says it is not? Can someone explain to me why the white group is considered dead? Thank you for your help.


Let me note that under area scoring rules, white is usually dead as well, if black knows how/when to capture white’s group.
Only when white has non-removable ko threats of the right size at the end of the game, black may need to grant white a seki.


This is a stalling situation during the game since white can’t capture and black isn’t interested in fighting a ko.

So it stays like that until the end of the gane.

It isn’t a seki because black can force white to capture by moving at T18.
Black T18 is atari, white captures, then black can start a ko playing at S19.
This situation is fully described in the page linked by Jon_Ko.
Under Japanese rules white is dead and there’s no need to play it out.


This “simple question” has forced the Japanese to rewrite their official rules, because the life and death status of this group is a bit weird.

So, not so simple. Although, the shape is very renown.

You thought it was a seki; a seki is a situation where no player wants to make a move, because the first who plays dies.

Here it is true that white doesn’t want to play first, otherwise white dies.

But black can play first and start a ko to kill all the white stones. And so this is a bit weird for the following reason: black can wait as as they want before they start the ko, and white cannot do much to prevent that. So there is a strategy for black to win this ko without “paying” for it.

Note that if a similar shape was on the side or the center instead of the corner, it would indeed be a seki as you thought. The ko only arises thanks to the corner.


I would add that in my opinion, and according to pretty much every ruleset in real life today, it only makes sense to say that this shape is completely dead for white if both players are sufficiently strong and know the shape well enough.

If there is any doubt, you should play it out. Relying on an AI to decide for you which groups are dead and which groups are alive only happens on Internet go servers, not in real life go games, and it’s a bit contrary to the spirit of the game.

If during a real life tournament, White claims that this is a seki, the only option for black is to try to actually kill. Black cannot just claim that this is dead “because it’s supposed to be”.


Under Japanese rules, “playing it out” is not the proper procedure in cases like this (group status disagreement in the scoring phase). If white claims a seki in a real life tournament, black should call a referee. If white claims a seki on OGS, black should call a moderator. The referee/moderator should then declare that white is dead.


Well, if black calls it dead because he knows that specific rule, I believe that any referee would agree.


This pattern is widely known as “bent four in the corner”. Even though the pattern does not immediately look like that, it can be reduced to such through optimal hypothetical play (in the context of Japanese rules life and death resolution). It is a prototypical example cited in the Sensei’s Library page linked by @Jon_Ko above.

Under Japanese rules, it would be a mistake for either player to play another move in that corner. It is kind of like a seki (but not actually a seki, since White is simply dead) before the game ends. Clearly, any additional move by White is self-atari that just loses all of the stones anyways. Another move by Black at T18 would turn the status of the corner into a ko fight, since it would be followed by White at R19, then Black at S19, and White at T19, starting a ko fight to settle the status.

However, it is a mistake to actually play this out in the game (since Black would need to be able to win the ko fight and waste an additional moving filling in one point of eventual territory to finally settle the ko fight), and Black is not required to actually play this out during the game to assert that it is dead. Instead, life and death status (under Japanese rules) is settled by analysis of a purely hypothetical playout that applies different ko rules. If the players simply resumed the game, it might not be possible for Black to actually win the ko fight, if White has large unremovable ko threats. However, the curious thing about Japanese rules is that White is still dead in this corner, and Black does not have to prove anything by actually continuing the game, even if there are large unremovable ko threats, since the ko rules are actually different when resolving life and death after the players have passed.

I’ve written about this pattern extensively in other forums posts:

Under Chinese rules, the status of this type of shape can be very different, given the global ko threat situation:

This post was adapted from a previous similar post here: A Nakade Question - #8 by yebellz


Gennan is right. Do not try to prove bent 4 cases when playing under Japanese rules. Calling a referee is the proper procedure.

Proving the kill isn’t proper, because under Japanese rules, playing it out might mean losing several points inside your own territory, because you would have to fix all possible ko threats before starting the ko to prove it is dead. There is a specific rule stating this shape is dead, for that reason.

You can prove without losing points if you’re playing with Chinese rules.


Not necessarily… you have to remove ko threats first, but sometimes there are unremovable ko threats. See Life and Death under Chinese Rules for example


This would be true under the 1949 Japanese rules, but not the 1989 Japanese rules.

Of course you should call a referee when there is a situation such as a disagreement between players. But the story doesn’t immediately end just because the referee is there.
Next thing that happens is that the referee asks you to prove that it is dead.

“I shouldn’t prove that it is dead because I would lose points if I did” it’s not really an argument. Even without going to something as complex as the bent four, all proofs of death usually require that you play stones in your own territory. If I have a single stone with four liberties in your territory, you need four moves in your own territory to capture it, which loses four points in Japanese rules. But that’s fine, because this happens during the “hypothetical play”, as a proof of death, not as actual play, so you don’t actually lose these points.

If you need to claim “this is obviously dead because a referee or AI or player who is stronger than me would know how to kill it, even if I don’t” then you’re not playing go, you’re just haggling.


OGS does not support hypothetical play. The next best thing is having a moderator declare the group status in case of a status disagreement.
If you feel that “playing it out” is the proper way to deal with status disagreements during scoring, I suggest you only play under a variant of area scoring rules.


Well, the moderator could ask the player to show how they would kill the group on a review board.

That suggestion is not realistic. The important thing is the opportunity to play games of go, not the precise rules used to play those games. The default setting on OGS is Japanese, so most games are played with Japanese rules, if only because most people don’t bother to change the settings, even if they are used to playing with AGA rules in real life. If I can choose the rules, I usually choose Chinese rules, but suggesting that I should refrain from playing a game of go just because I can’t choose my preferred rules is completely disconnected from reality.

Actually, this would be a very good argument to switch OGS 's default from Japanese to AGA!


My impression is that hypothetical play intends to determine the objective group status, i.e. the status under optimal play, so a referee may need to step in when status disagreement occurs during scoring in a weaker players game. But I may be wrong in my interpretation of the Japanese rules. The wording and details of the Japanese rules are not really tailored toward hypothetical play procedures. Perhaps because it doesn’t really happen in pro games.


Yeah. I’ve run into difficulties with that.

I’m used to introducing the games to players using French rules (which are mostly the same as AGA, with area-scoring, but the ease of a pass-stone to emulate area-scoring with a territory-counting method).

I’ve never had any issues explaining to beginners that dead groups could be removed “for free” at the end of the game, because that’s just a convenience, and we could play it out if we wanted to.

But I’ve had to introduce the game over the internet using Japanese rules on some occasions, and I’m a bit struggling as to how to explain that in a convincing manner.

“So, you see that stone that you played into my territory? Well, I’m telling you that if I wanted to capture it you couldn’t stop me, and so I’m going to remove it from the board for free without actually capturing it, even though it would cost me points to actually capture it.”


I have had a few cases of group status disagreement between beginners under Japanese rules IRL. In some cases I let them resume with pass stones (requiring an even number of moves in the encore) until they agree. But in almost all cases it is sufficient to ask about eyes or seki (if not, it is dead).
I don’t think I ever had them revert to hypothetical play.
So when I get called to a casual IRL game with a status disagreement, I don’t follow Japanese rules strictly.
But these cases are quite rare IME, and usually beginners are not very inquisitive about formal rules on handling these rare situations.

I can understand it is a bigger problem online, where some players try to cheat when scoring a game against a beginner. I’d be OK with having an AI adjudicate group status in case of a disagreement when scoring an online game.


That’s good insight regarding refereeing a tournament.

Do you have also insight regarding explaining the rules during an initiation? Imagine the beginner has already played capture-go, but is playing their first game of go, or has played at most three games of go.

I’m not talking about a rare and complicated situation such as bent-four or even seki. I’m talking about the very common and very simple situation where there is just one opponent stone in your territory, and you have to explain that you can remove that stone “for free” because it is “dead”, even though it would cost several points to actually capture it.

With Chinese or AGA rules you don’t even need to define the concept of “dead” stones. There are stones which are on the board, and stones which have been captured. And you can easily explain that if both you and your opponent agree that they can’t prevent you from capturing their stone in your territory, you don’t have to bother adding the stones to capture it. And even if your opponent disagree, which they might because they’re a beginner so it’s hard for them to be sure that there is no way to save their stone, then there is still no issue, because you can just play it and actually capture them. (And if you fail to capture them, well, then they were right to disagree)

But with Japanese rules it looks like you have to define what it means for a group of stones to be “dead”, so that you can remove dead groups without capturing them, and I find that it’s a very complicated concept, compared to the otherwise extremely simple rules of go.

And I have only very little experience introducing the game with Japanese rules, so I’d be very thankful to anyone who has more experience and could explain how they go about it.


I was talking about casual games. I’m not a tournament referee.

I don’t focus on formal rules much.
I don’t start with capture go.
I start with the objective of the game: score more points than your opponent.
I then proceeed to show the 2 ways to make points: walling off parts of the board (surrounding territory), capturing stones (by taking all their breath away, smothering).

Other things come only later, such as ko, eyes, seki. Redundancy of capturing dead stones comes later still. Neccessity and efficiency of moves are topics later again. At first, novices will just capture if they can, regardless if it loses points, and I would let them do it without me criticising it. Only when they reach about 20 kyu (not really novice anymore), I may tell them, if they haven’t figured it out by themselves at that point.