Skip to the board
Strategy Board Hub

Baduk 바둑 · 囲碁 · 圍棋

Go

Surround territory. The simplest rules in this hub, and by far the deepest game.

go

How to play Go

  1. 1Go is played on the intersections of a grid — traditionally 19×19, with 13×13 and 9×9 used for shorter games. Black plays first, and players alternate placing one stone on any empty intersection.
  2. 2Stones never move once played. A group is a set of stones of the same colour connected along the lines, and its liberties are the empty intersections directly adjacent to it.
  3. 3When a group has no liberties left, it is captured and removed from the board. Captures happen immediately after the opponent’s stone is placed.
  4. 4You may not play a stone that would leave your own group with no liberties — unless that same move captures enemy stones and thereby creates a liberty. This is the suicide rule.
  5. 5The ko rule forbids a move that would recreate the position immediately before your opponent’s last move. You must play elsewhere first; that intervening move is called a ko threat.
  6. 6Either player may pass instead of playing. When both players pass in succession, the game ends.
  7. 7The winner is decided by area: your score is the number of intersections you occupy plus the empty intersections that only you surround. White receives komi — a compensation of 6.5 points here — for moving second, which also prevents draws.
  8. 8Stones that cannot avoid capture are called dead and count as territory for the opponent even if they are still sitting on the board. This site resolves dead stones automatically by simulating the position to the end hundreds of times.

Go strategy guide

Corners first, then sides, then centre

Surrounding territory in a corner takes the fewest stones, because two edges do the work for you. The centre is the most expensive place on the board to make territory. Standard openings occupy corners, extend along sides, and only fight for the middle later.

Two eyes means alive

A group with two separate internal empty points can never be captured, because your opponent would have to fill both and the first fill is suicide. Every life-and-death problem in Go reduces to whether a group can make two eyes. Learn to see this before you learn anything else.

Do not answer every move

The biggest jump in strength for most beginners comes from asking "is there something larger elsewhere?" before responding locally. A move that saves three stones while your opponent takes a twenty-point corner is a losing exchange, however satisfying it feels.

Stay connected, keep them separated

Connected stones share liberties and are far harder to kill; separated groups each need their own eyes. Most fighting in Go is really an argument about connection. If you can cut your opponent into two weak groups, you will usually profit from attacking both.

Play the 9×9 board to learn faster

A full 19×19 game takes 250 moves and hides your mistakes in the noise. On 9×9, every move matters immediately and a game takes ten minutes, so the feedback loop is far tighter. The AI here is also strongest on 9×9, which makes it the best board to train against.

The history of Go

Go originated in China at least 2,500 years ago — it is mentioned in the Analects of Confucius — making it the oldest board game still played in essentially its original form. It reached Korea and then Japan by the 7th century, and each culture developed its own name and tradition: weiqi (圍棋) in China, baduk (바둑) in Korea, igo (囲碁) in Japan.

Japan formalised professional Go under the Tokugawa shogunate, establishing four hereditary houses that competed for the title of Meijin and produced centuries of recorded games. The 20th century saw the centre of gravity move to Korea and then back to China, and today the strongest professionals come from all three countries.

Go was the last classical board game to fall to computers. Its branching factor and the difficulty of evaluating a position defeated alpha-beta search entirely; the breakthrough came from Monte-Carlo tree search in the mid-2000s, which made competitive 9×9 programs possible, and then from AlphaGo’s defeat of Lee Sedol in 2016. The engine on this page uses the earlier approach — MCTS with random playouts, no neural network — which is why it plays best on the smaller boards.

Go FAQ

Should I start on 9×9 or 19×19?
9×9, without question. The rules are identical, a game takes ten minutes instead of two hours, and every move has visible consequences. Move up to 13×13 once capturing races and eye shape feel natural.
What is komi and why is it 6.5?
Komi is a point compensation given to White for playing second. The half point guarantees there can be no draw. 6.5 is a common value for smaller boards; 19×19 tournaments typically use 6.5 or 7.5 depending on the rule set.
How does scoring work on this site?
Chinese area scoring: your score is your stones on the board plus the empty points only you surround, and White adds komi. Rather than asking you to mark dead stones, the site plays the finished position out hundreds of times and lets the statistics decide what was alive.
Why can’t I play in that one spot?
Two possible reasons. Either the move would be suicide — your stone would have no liberties and captures nothing — or it is forbidden by the ko rule, which stops you from immediately recapturing and repeating the position. Play a ko threat elsewhere first.
How strong is the Go AI here?
It uses Monte-Carlo tree search with random playouts, entirely in your browser and without a neural network. On 9×9 at the highest level it is a reasonable mid-kyu opponent. On 19×19 it is much weaker, because the number of positions to sample grows far faster than the time available.

More games on this site

Same engine approach, entirely different kind of thinking.