The Story
Where it came from, and why it turned out this way.
OriginChina; earliest reliable record in the Zuo Zhuan, 548 BCE
What is documented, and what is legend
The story that Emperor Yao invented Go to instruct his son comes from Zhang Hua's Bowu Zhi, written in the 3rd century CE — a much later retelling, not evidence of origin. The earliest genuinely reliable reference is in the Zuo Zhuan for 548 BCE, where a political manoeuvre is compared to hesitating with a Go stone in hand, implying the game was already familiar to the nobility. So: Go is at least two and a half thousand years old, and that much is documented. Everything earlier remains legend.
The rules are short because the complexity is outsourced
Go's rules fit on a single page: alternate placement, capture when liberties run out, no suicide, no repetition. There are no piece types, no ranks, no promotion. But precisely because the rules specify no particular shapes, concepts like life and death, ko fights, thickness, and efficiency can emerge entirely from the relationships between stones instead of being written down. It is a rare kind of design: the complexity lives not in the rules but in their consequences.
Komi is the youngest rule in an ancient game
That moving first is an advantage was hardly a secret, but for centuries it was handled with handicap stones or pre-placed stones. Compensating with a fixed numerical allowance is a 20th-century idea, and the number has been adjusted repeatedly — because it is fundamentally a statistical question of how much compensation brings the two win rates closest to even. This page uses 7.5 under Chinese rules and 6.5 under Japanese and Korean rules, and the gap comes from the two scoring methods themselves.
One move in 2016
Go was long considered the game machines would struggle with longest, because the branching factor is enormous and positions resist evaluation. In 2016 AlphaGo beat Lee Sedol 4–1, and move 37 of game two — a shoulder hit on the fifth line — was widely judged by professionals as something no human would play, and then turned out to be excellent. Human joseki and opening theory were substantially rewritten afterward, the fastest theoretical upheaval in twenty-five centuries of the game.
- In the famous "ear-reddening game" of 1846, Honinbo Shusaku's opponent is said to have flushed red at the key move, letting a watching physician conclude the game was decided — the most retold anecdote in Go history.
- The Edo shogunate established four hereditary Go houses — Honinbo, Yasui, Inoue, and Hayashi — giving Go the same kind of three-century institutional continuity that shogi enjoyed.
- Chinese rules count area and Japanese and Korean rules count territory; the same game frequently scores a point or two apart under the two methods, which is why endgame habits differ between them.
Complete Rules
Start with what a turn allows and how the game is won.
- Board
- 9 / 13 / 19 lines (19 by default)
- First move
- Black, with komi compensation to White
- Winning conditionGo is not about capturing everything, it is about who encloses more of the board. The game ends the instant both players pass in succession, the position is scored under the current ruleset, komi is added for White, and the higher score wins. Every fight, life-and-death problem, and ko in the game exists only to serve that final number.
- Players alternate placing one stone on an empty intersection, and stones never move afterward. Stones of the same color that are orthogonally adjacent — diagonals do not count — form one group that lives or dies together. The empty points orthogonally adjacent to that group are its liberties.
- A group whose liberties are all occupied is captured and removed entirely. Each move resolves in a fixed three-step order: first your stone goes down, then any adjacent enemy groups with no liberties are removed, and only then is your own group checked for liberties. Because capture is resolved before the suicide check, you may legally play on a point that appears to have no liberties, as long as that very move captures enemy stones.
- If after placement your own group still has no liberties and no enemy stones were captured, the move is suicide. The interface rejects it with the message that suicide is not permitted. Killing your own whole group at once is equally forbidden.
- A ko is the situation where each side has a single stone that can capture the other, and recapturing would restore the board to exactly its previous shape. If immediate recapture were allowed the two players could repeat forever, so the rules must forbid one of the recaptures. The Japanese rules option forbids recreating the board position from two moves ago, which is the simple ko rule; the Chinese and Korean options forbid recreating any board position that has ever occurred, which is positional superko.
- When the ko rule blocks a move, you will see a message that the ko cannot be recaptured immediately, or that the whole-board position may not repeat. To keep fighting the ko you must first play a move elsewhere that your opponent has to answer; once they answer and the board shape has changed, you may come back and take the ko.
- Superko looks only at stone placement and ignores whose turn it is. Under the Chinese and Korean options, a position identical to any earlier post-move position is rejected even with a different player to move, which cuts off triple-ko and similar endlessly repeating positions outright.
- You may pass instead of playing. Two passes in a row end the game immediately and move straight to scoring, with no additional confirmation step.
- During scoring, clicking any stone on the board marks its entire group as dead or removes that mark, and the score recalculates instantly against the board with dead stones removed. Marking works group by group, so scattered dead stones must each be clicked separately.
- An empty region surrounded by only one color counts as that player's territory. Empty points touching both black and white stones belong to nobody — shared liberties and the neutral points inside a seki fall into this category.
- Chinese rules use area scoring: your score is the number of your stones on the board plus the territory you surround, with 7.5 komi added for White. Japanese and Korean rules use territory scoring: your score is the territory you surround, plus the enemy stones you captured during play, plus the opponent's dead stones removed at scoring, with 6.5 komi added for White.
- The two methods produce different results for exactly one reason: area scoring counts the points your stones occupy and territory scoring does not. Filling in a point inside your own territory leaves your area score unchanged (one more stone, one less empty point) but costs you a full point under territory scoring. So endgame filling, giving away stones, and pure capture exchanges are not worth the same under the two rulesets, and the final margins routinely differ by a point or two.
- Changing the board size or the ruleset immediately starts a new game. Komi is determined by the ruleset alone: 9×9, 13×13, and 19×19 share the same komi value, and it cannot be set separately.
Your First Game
Use this order of thought to get through a complete game quickly.
- For your first game, choose 9 × 9. With only 81 points a game takes twenty or thirty moves, which is enough to experience capture, eye-making, passing, and scoring end to end.
- To count liberties, treat the group as a single unit, trace around its outline, and count every empty point orthogonally adjacent to it, counting each point only once. A group with one liberty left is in atari and can be captured next move; with two liberties left you should already be thinking about running, connecting, or counter-atari. When two groups surround each other, the one with fewer liberties dies first — count both sides and you know who wins the capturing race.
- Open in the order corners, then sides, then center. A corner needs the fewest stones to enclose a definite amount of territory, while enclosing the same area in the center costs more than twice as many stones. On 19×19 the first move is usually a star point or a 3-4 point; on 9×9 you can play directly on the 3-3 point or near the center.
- Here is one joseki idea you can simply follow: when your opponent approaches your star-point stone with a knight's move, if you have few stones on that side, answer with a knight's-move corner enclosure to take the corner territory and let them settle with a two-space extension on the side. If you already have stones on that side, pincer instead and treat the approaching stone as a target. Which one is right depends on how many stones you already have nearby, not on the move being good or bad in isolation.
- Judge life and death in three steps. One: is the group actually surrounded? If there is still an exit, run out first — escaping is easier than living inside. Two: can the space it encloses be divided into two separate eyes? Three: can your opponent play inside and reduce that space to a single eye? A group is genuinely alive only if all three checks pass.
- Real eyes versus false eyes is where beginners stumble hardest. For an empty point to be a real eye, all of its orthogonal neighbors (three on the edge, two in the corner) must be your stones, and you must also control the diagonal points: in the center you need at least three of the four diagonals, and on the edge or in the corner you need all of them. If your opponent holds a key diagonal, they can capture the stone beside the eye and the "eye" collapses — it was a false eye and cannot be used to live.
- Connecting and cutting are usually worth more than capturing. Joining two of your weak groups, or playing the cutting point between two of your opponent's, typically gains far more than capturing one or two isolated stones on the side.
- Before each move, answer the question "what is this stone doing?" — enclosing territory, expanding, attacking, or securing life. If you cannot answer, the move can almost certainly be played somewhere bigger.
Practical Strategy
Move from knowing the rules to playing with a plan.
- In the opening, look for big points in the efficiency order corner, side, center, and prefer the double-purpose big points that expand your own framework while limiting your opponent's.
- The purpose of attacking a weak group is usually not to kill it, but to chase it while enclosing territory or building thickness alongside. If the group escapes but you gained every tempo and a strong outer wall, the attack succeeded.
- Thickness is a wall with no cutting points, and using it to enclose a small area right beside itself is the most wasteful thing you can do with it. Instead, drive your opponent's weak stones toward the wall, or extend far along the side away from it.
- Invading and reducing suit different situations: invade inside a framework that still has gaps, but switch to reduction once it is solid, riding lightly along the outside to shrink its scale rather than diving into your opponent's strength and dying there.
- The full ko procedure runs like this: after your stone is taken you cannot retake immediately, so you play a threat elsewhere that your opponent must answer, and after they answer you come back and take the ko. If they also want the ko, it is their turn to find a threat. Count both sides' ko threats before starting a ko — the side with fewer threats should not pick the fight, and can instead resolve the ko by filling it in exchange for profit elsewhere.
- Play the endgame in the order double sente, single sente, then gote. Between two points of equal value, play the one that leaves you with the initiative afterward, because a sente endgame move is effectively a free turn.
- Every dozen moves or so, take stock: roughly count the territory each side has secured, add komi, and compare. If you are ahead, simplify and strengthen your weak groups; if you are behind, invade or start a fight now, because the longer you wait the harder it is to turn the game around.
Winning and Special Rules
Details that are easy to miss but can change the result.
- Komi compensates for Black's first-move advantage. On this site Chinese rules use a fixed 7.5 and Japanese and Korean rules a fixed 6.5, changing automatically with the ruleset.
- Because komi carries a half point, the two scores can never be equal, so Go games on this site never end in a draw under either scoring method.
- Seki describes two mutually surrounding groups that neither can make two eyes: whoever fills the shared liberties first is captured first, so neither side moves and both groups live. This page's scoring gives shared points touching both colors to nobody, so the neutral points inside a seki are worth zero under both rulesets.
- Endlessly repeating positions such as eternal life and triple ko are forbidden outright by superko under the Chinese and Korean rulesets. Japanese rules only forbid the position from two moves ago, so in theory the repetition can continue and the players must end the game by passing.
- Scoring depends entirely on dead-stone marking: an unmarked dead group is still counted as alive under area scoring and still holds its territory, which visibly skews the result. Click again to unmark a mistake, or use restore to clear all marks at once.
- A pass counts as a move, and one pass from each side ends the game immediately. Undo rewinds two moves at a time — yours and the AI's — and clears any dead-stone marks you have made.
Common Mistakes
The traps new players fall into most often.
- Treating "all connected" as safe. No matter how large a group is, if it cannot make two real eyes and cannot escape, the whole thing comes off the board — and the bigger it is, the worse the loss.
- Counting a false eye as a real one, especially in the corner or on the edge where the opponent already holds a key diagonal. Discovering you are one eye short at the end of a capturing race is usually unfixable.
- Chasing a local kill so single-mindedly that your opponent plays two or three big points elsewhere — you win a dozen points locally and lose the game on the board.
- Casually filling in your own territory under territory scoring, where each such move throws away a full point. Under area scoring the same move costs nothing, so endgame habits from one ruleset must not be carried into the other.
- Mirroring your opponent's pass before the position is settled and dropping straight into scoring, where unresolved life-and-death groups are scored exactly as they stand.
- Trying to retake a ko repeatedly under Chinese or Korean rules, hitting the whole-board-repetition block, and assuming the program is broken — that is superko working as designed.