Quick answer
Tsumego (詰碁) are Go life-and-death problems: you look for the exact sequence that lets a group live, kills it, or gets the best available result. A group is unconditionally alive when it has two separate real eyes. A group is dead when it cannot avoid capture against the best defence, even if its stones are still on the board.
Scroll the table sideways to compare all columns.
| Key point | What to remember |
|---|---|
| What makes a group safe | Two real eyes that do not depend on each other |
| Possible results | Alive, dead, ko or seki |
| Skill you are training | Reading the sequence in your head before you play |
| Go rules on ChessPlayX | Area scoring, 7.5 komi, positional superko |
Life, death and two real eyes
Beginners tend to look at territory. As you improve, a better question takes over: is this group actually alive? A big group can still die because the opponent found one key point, and a group that looks completely surrounded can still live with one precise move.
Stones connected horizontally or vertically share their liberties. When every liberty is filled, the whole group comes off the board. But a group that cannot avoid capture even with the best defence is already considered dead. That is why some stones are still on the board at the end of a game and are marked as dead stones.
The word “real” matters: two holes in a shape do not automatically make two eyes.
False eyes: the trap that makes a dead group look alive
“One eye here, one eye there, so I’m alive.” Not necessarily. One of them may be a false eye: the stones around the eye point are not solidly connected, so the opponent can capture or cut them and the point stops working as an eye.
Tsumego is really a reading exercise
Seeing a shape, remembering a pattern and playing the familiar point will help you solve easy problems faster, but it won’t make you a stronger player on its own. If you guess the first move without reading what follows, you haven’t really solved the problem. The reading loop is: find candidate moves, read the opponent’s strongest reply, find your answer, check every form of resistance, and only then decide whether the result is life, death, ko or seki.
1. Pin down the goal
Who plays first, and is the aim to live or to kill? “Find the best move” is too vague. “Live no matter what” or “kill no matter what” is a goal you can actually read towards.
2. Count the eye space
How many empty points are there, how are they connected, which point splits the area in two, can the opponent play inside, and can the group extend outward? The inner shape matters more than the number of stones: a ten-stone group can die and a four-stone group can live.
3. Look for the vital point first
When the attacker and the defender both want the same point, it is usually the vital point. If you want to kill, ask yourself: “If I were trying to live here, where would I play?”
4. Check forcing moves before anything else
Ataris, cutting threats, moves that destroy or make an eye, captures, connections and ko. Forcing moves come first because they shrink the tree: instead of ten possible replies, the opponent usually has one or two sensible ones.
The point your opponent needs in order to live is exactly the point you need in order to kill.
Atari, move order and candidate moves
Atari is not always the best first move
An atari can accidentally help the opponent connect, make an eye or play the move they wanted anyway, and it can cost you the vital point. In many positions the strongest move is a quiet one: you take the point that leaves the group with no way to make two eyes, and the capture follows naturally.
Don’t capture until you need to
Ask yourself: “Do I need to take away liberties, or do I need to destroy the eye shape?” A group with plenty of liberties but no way to make two eyes is still dead; a group short of liberties with two real eyes is still alive.
Move order matters more than any single move
You might see every move in the solution: the atari, the eye point, the cut and the capture. But if the eye point has to come before the atari, swapping them can ruin the whole combination. Tsumego is the best way to train move order.
Always assume the strongest resistance
Tsumego doesn’t ask “is there a line where I win?” It asks “does my opponent have any way to make this fail?”
The candidate-move method
Before reading deeply, pick two to four candidates, roughly in this order of priority: the vital point of the eye shape, a hane, a throw-in, an atari, a connection or cut, then a placement. For each candidate, list the opponent’s replies and read every branch. If even one branch makes candidate A fail, drop A and move on to B.
Life-and-death tesuji to recognise on sight
Scroll the table sideways to compare all columns.
| Tesuji | What it does in a life-and-death problem |
|---|---|
| Atari | Reduces the opponent’s group to one liberty; simple, but the timing is everything. |
| Hane | Bends around the head of the opponent’s stones; in a small space, one hane can change whether eyes are possible at all. |
| Throw-in | Deliberately puts a stone where it will be captured to spoil the eye shape, force a capture, reduce the space or create a ko. |
| Snapback | Lets the opponent capture a few stones, then immediately recaptures a larger group; sometimes being captured is the solution. |
| Placement | Plays straight into the eye space so it cannot be split into two eyes, or so that one eye becomes false. |
| Shortage of liberties | The shape looks alive, but the surrounding stones are short of liberties; in a capturing race, one liberty can flip the result. |
Ko and seki: “not dead” doesn’t always mean alive
A life-and-death problem has more than two possible outcomes. It helps to separate five of them: unconditionally alive, unconditionally dead, ko to live, ko to kill, and seki. If a group can only live by winning a ko, it is not “safely alive”; the result still depends on how many ko threats each side has.
Seki: living without two eyes
Seki happens when two opposing groups both survive because whichever side plays into their shared liberties first puts its own group in atari. Stones in seki are not dead. Seki is one of the most common reasons newer players mark dead stones incorrectly when scoring online.
Solving a tsumego and practising the right way
Say the problem reads “Black to play and kill.” Don’t start clicking.
- Look at the whole shape for five to ten seconds without playing: find the eye space, the weaknesses, the liberties and the cutting points.
- Pick at most three candidates, for example A in the middle of the eye space, B a hane and C an atari.
- Read A all the way to a conclusion. If Black still succeeds against White’s best resistance, keep A.
- If any line lets White live, A is not the solution; move on to B.
- Only play once you have reached a conclusion. Clicking until something works trains guessing, not reading.
One practice session: three kinds of problems, three skills
Scroll the table sideways to compare all columns.
| Problem type | Suggested amount | Skill you train |
|---|---|---|
| Very easy | About 10 problems | Recognising patterns almost instantly |
| At your level | 10–15 problems | Reading full lines in your head without clicking |
| Hard | 2–3 problems | Stretching your limits, with no need to solve them all |
Twenty problems read properly are worth more than a hundred rounds of “click, wrong, click again, right.” When you check an answer, make sure you can explain both why the correct move works and why the move that looked reasonable fails. If you solve everything instantly, step up the difficulty a little; if you have to guess every time, it is too hard. The sweet spot is hard enough that you must read, but still solvable in your head.
When is a group considered dead?
A group is dead when, with best play from both sides, it cannot avoid being captured, even though its stones have not been removed yet. Depending on the ruleset, the players can agree on that without playing it out to the last liberty.
The Black group is currently counted as alive. Tap the group or press the button to mark the whole group dead.
Don’t mark a group dead just because “it’s in my territory”
If the group has two eyes, is in seki or still has a line that lives, it is not dead. On the other hand, an invasion that has completely failed can be marked dead even though its stones are still on the board.
How to finish an online Go game
A game isn’t over just because you can’t see anything worth playing. Go through these seven steps:
- Settle every contested area. Check for invasions, reductions, unsettled groups, unresolved ko, neutral points (dame) and any reinforcing moves you still need. With area scoring, neutral points are handled differently from Japanese territory scoring.
- Pass. If your opponent keeps playing, the game goes on; your pass does not oblige them to pass.
- Both players pass. On ChessPlayX, two consecutive passes move the game into the scoring phase. Many other servers also guess which stones are dead; that estimate is a helper, not a ruling.
- Check each group before you look at the score. Does it have two real eyes, can it run out, is it in seki, is there an unresolved ko, and if play continued, would it definitely be captured?
- Mark the dead stones. Select a group to switch it between alive and dead; dead groups are usually shown faded.
- Review the whole board. Look at Black’s area, White’s area, neutral points, seki, komi and the ruleset. Don’t just read “Black +4.5” and accept: getting one large group wrong can swing the score by dozens of points.
- Both players accept. Only accept when you both agree on which stones are alive, which are dead and how the board ends. On ChessPlayX, any change to the dead-stone marks cancels earlier acceptances.
When players disagree, and why the ruleset matters
You say “this group is dead”; your opponent says it’s alive. Don’t argue in the chat; play it out. You prove your point by capturing the group, and they prove theirs by living when you try to kill it. On ChessPlayX, choosing Resume play clears the dead-stone marks and returns the game to normal play.
Whether adding stones inside your own area to prove a point costs you anything depends on the ruleset. Check the rules before the game starts rather than assuming every server scores the same way.
Scroll the table sideways to compare all columns.
| Ruleset | How the score is counted | Worth knowing |
|---|---|---|
| Japanese rules | Territory scoring: surrounded territory plus prisoners, including dead stones removed at the end | Living stones on the board are not counted directly; playing inside your own territory costs a point |
| Chinese rules (used on ChessPlayX) | Area scoring: living stones on the board plus surrounded empty points | Prisoners are not added; playing inside your own area usually does not change the score |
In most games both methods produce the same winner, but some endgame situations, such as neutral points or seki, can be counted differently.
A 10-second checklist and common mistakes
Before marking a group as dead, ask yourself:
- Does the group have two real eyes?
- Am I mistaking a false eye for a real one?
- Can the group connect out?
- Can it make an eye in one move?
- Is there a vital point the defender can still take?
- Is this seki?
- Is there an unresolved capturing race?
- Is a ko involved?
- If play continued, could I really capture this group?
- Does my opponent agree with that status?
If you are still unsure about question 9, don’t rush to mark it dead.
Scroll the table sideways to compare all columns.
| Common mistake | What to do instead |
|---|---|
| Reading only your own moves | After every move, read the strongest resistance; your opponent won’t sit still while you finish your plan. |
| Looking only for atari | Look for the vital point; it is harder to see but usually matters more. |
| Not counting liberties | Count the liberties of the inside group, the outside group, and again after every capture. |
| Stopping after a capture | Check for snapback, ko and sacrifices; one capture is not always the end. |
| Not checking for false eyes | For each eye, ask whether the opponent can destroy it. |
| Clicking too early | Finish reading in your head before you play; otherwise you are only finding the interface’s answer. |
Frequently asked questions
What is tsumego?
Tsumego are Go problems about whether a group lives or dies. You have to find the exact sequence that lets the group live, kills it, or reaches the best result against every defence.
Is a group with two eyes always alive?
If they are two separate real eyes, the group is alive in principle. Newer players often mistake a false eye for a real one, so check the stones around each eye.
Can stones still on the board be considered dead?
Yes. If both players agree that the group could not avoid capture if play continued, it is marked as dead when the game is scored.
Do I have to capture every dead stone before the game ends?
Not under every ruleset. Groups that both players agree are dead can be removed during scoring; if you disagree, keep playing and settle it on the board.
Is automatic dead-stone detection always right?
Don’t assume so. Complicated shapes, seki or unfinished positions can lead to a wrong automatic result. On ChessPlayX, the two players mark dead stones themselves and both accept the score.
What if my opponent marks my living stones as dead?
Don’t accept the score. Check the group again, unmark it, or resume play to settle the dispute on the board.
Are stones in seki dead?
No. In seki, opposing groups live together because neither side can capture the other without suffering a decisive loss itself.
Why is the server’s score different from my own count?
Usually because of a different ruleset (territory or area scoring), wrongly marked dead stones, a missed seki, neutral points, komi, or a game that wasn’t really finished.
Should I study tsumego with an AI?
An AI is useful for checking lines after you have practised. To read better, though, solve the problem yourself first: most of the value comes from reading it in your head.
How many problems should I solve each day?
There’s no required number. What matters more is the right difficulty, reading complete lines, understanding why the answer works and reviewing the problems you got wrong.
Key takeaways and how to start practising
Beginners see stones. Stronger players see shapes. Strong players see what a shape can still become over the next few moves. Ask “if my opponent played first here, what would be the most dangerous move?”, then switch sides: “if I had to save this group with one move, where would I play?” When both questions point to the same intersection, you’ve found the vital point.
- Find the vital point, read the strongest resistance and watch the move order.
- Check real and false eyes and count liberties before you conclude anything.
- Recognise throw-ins, snapbacks, ko and seki, and don’t click before you have read.
- At the end of a game: pass, check life and death, mark dead stones, review territory, seki and komi, then both accept. If you disagree, keep playing.
Start with Go puzzles that have a clear live-or-kill goal, review the Go rules on ChessPlayX, then put it into practice in a full game of Go against a friend or the AI.
Sources and further reading
- Sensei's Library: Tsumego: what tsumego are and how to practise them.
- Sensei's Library: Life and death: life, death, ko and seki.
- Sensei's Library: False eye and Vital point: false eyes and vital points.
- Sensei's Library: Seki: positions where groups live without two eyes.
- Chinese Rules of Weiqi (English translation) and Japanese Rules of Go (1989): area scoring, territory scoring and how dead stones are handled.
- British Go Association: Comparison of Some Go Rules: how rulesets differ on scoring and ending the game.
- Read next: Xiangqi vs Go.
