Small board Go

    Keywords: Variant, Equipment

Small board go is go played on a board of size smaller than the standard 19x19 goban. Commonly used sizes are 9x9 or 13x13. These board sizes are suitable for beginners wanting learn the basic rules of go and experienced players wanting to refresh their thinking and exercise specific tactical aspects of play. One good use for small boards is to try to work on your positional judgement with smaller numbers involved.

Capture Go is commonly taught initially on a 5x5 board. Even this tiny board can be amazingly instructive. The book [ext] How to Improve on a 5x5 board actually contains dan-level problems on this size board.

Small board Go has both entertainment and pedagogical value. I (Bill Spight) have often used it for teaching. The 3x3 teaches me ari me nashi, for example. Rules of Go questions are very significant for small boards. Some small boards make no sense under Japanese rules, for instance. And strange superkos can arise.

Small board Go researchers include Ted Drange, [ext] John Tromp, Bill Taylor, Robert Jasiek, and many others (see also Tromp Taylor Rules).

Mane go strategy  

Here is a cute 3x4 problem:
Black to play and stomp.


-- BillSpight

ferdi: There a lot of interesting 3x4 problems. You can find some of them here: 3x4Problems

See solutions for small boards.

Go has been solved[1] for boards up to and including 6x7 with area scoring (Ref. [ext] Mini-Go by Ted Drange with assistance from Bill Spight). With ideal play (and no komi), Black wins 3x3 by 9, 4x4 by 2, 5x5 by 25, and 6x6 by 4.

  • 7x7 may[2] be the smallest undecided Go board. (This may be best play)(See 7x7 game for an example)
  • 13x13 has handicap (hoshi) points spaced the same as 19x19.
  • 15x15 was popular in China 1000 years ago.
  • 17x17 is used in Tibetan Go and was used in China 500 years ago.

Small boards (and large boards) are easily played on KGS and other servers. KGS will handle all from 2x to 38x.

What are the suggested komi for non standard board sizes (7x, 9x, 11x, 13x, 15x, 17x, and 21x)? (see Komi Go / Discussion) --Hu of KGS

Neil: I wonder, does Black win 7x7 by 49 and 8x8 by 6 or 8? How likely do the experts figure that there is a pattern there?

Winning 5x5 by 25 means white cannot live. This is probably not true on 7x7 and is clearly false on large boards, so no such pattern here.

What is the contents of the Mini-Go site, I cannot reach it...

Where are the proofs for most of the boards of smaller sizes?! I have seen 0x0 various rulesets, 1x1 various rulesets, 1x2 a few rulesets, 2x2 some rulesets, 3x3. --Robert Jasiek

Jasonred This is actually really useful for me, as it gives ways to give life other than the old "two eyes" bit, which some amateurs (me), get too used to. We forget there's also shortage of liberties and seki... I always forget seki.

Could some people come up with some question and answer problems on how to give a group life through seki? Thanks.


Bill: I do not think that Ted, John, or Bill would say that all of the small boards they investigated, up to size 6x7, have been solved.

Dieter: Erik van der Werf A dutch Go player and programmer has his Go program solve the 5x5 board. It is claimed to be invincible with Black even by a professional opponent. An article appeared in "De Standaard", where van der Werf is interviewed on his project.

ilan: Hello!? By definition, a position being solved by computer means that the program will play optimally in every position, so the strength of the opponent is irrelevant. Once again the aura of professional strength raises its head...

Dieter: Is it forbidden then to explain an accomplishment by adding a weaker, but still formidable statement?

OK: 5x5 go is solved by the program. A fortiori, a professional player will not be able to beat this program on a 5x5 board.

ilan: Well a question of independent interest is whether a professional could match the computer if proper komi (computable by the program) were given for various opening moves, e.g., Black starts at the 1-2 point. In other words, the professional "loses" if he cannot get a tie with the correct integer komi (playing either side).

[2] Erik van der Werf in his Ph.D. thesis, [ext] AI techniques for the game of Go (page 54) reports that a group of Japanese amateurs with some support from Kudo Norio and Nakayama Noriyuki claim to have solved 7x7 go. The result is a black win by 9 points. Erik references James Davies article 7x7 Go in the American Go Journal 29(3):13, 1995.

No superko rules mention board symmetry. So, applied to the letter, they all consider the following two positions different. However an observer walking around Board 1 will eventually see Board 2.

If you believe they really are different, try retaking a ko 'illegally', then rotating the board 90 degrees and claiming that its a new board position.

In practice then, the 8 symmetries of the square are applied to a board position for the purposes of superko. However I do not know whether this is the interpretation used by any of the analyses above.

Board 1  
Board 2  

axd: small boards are great (pun intended) for initiations, but they might come with an inconvenience: just as capture go risks putting focus on capturing stones, small boards risk to give beginners the false impression that the game has to do with total win or loss, an inherent property of small boards.

Slarty: in my experience 9x9 has a beginning, middle, and endgame, and you get to play a complete game of Go for a fraction of time investment. People often get the impression that one or the other of these phases is all-important on the normal size board, because that's all they end up being comfortable playing. Also, the basest tactics are all-or-nothing, so 5x5 could be very instructive.

See Also

Small board Go last edited by on November 16, 2015 - 06:12
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library