The 9x9 Amateur-Hour Strategy Thread

I’m by no means an authority on the 9x9 board, or any board for that matter. But I won’t wait around for permission to delve into a flighty dissection of the 9x9 board, which, on closer inspection, begins to unravel an inspiring level of complexity distinct from that of the larger boards. I should know; I’m giving a lecture on 9x9 for Learn Go Week. And like many overly anxious teachers, I’ve lately felt the motivation to impart my own brand of “hype” on some of the material I hope to discuss on this transcendental lecture. This forum post is meant as my own opinionated primer—a navigational brochure, per se—on the 9x9 board.

First things first: 9x9 is about fighting, and reading is key. (If 19x19 is a battlefield, 9x9 is a knife fight in a phone booth.) Sometimes the game is like one big go problem. A unique feature of 9x9 is that after a few moves it becomes possible to read out lines of play to the very end. Quite often I realize that the “natural” line leaves me behind by half a point, and then decide to change my plan. Generally speaking, a calm game to the very end means one player didn’t realize they were losing and failed to complicate the game in looking for a chance to catch up.

The opening is not complicated. Pros play anywhere on the third line or above, and the opening is usually over by move 5.

If you’ve got a taste for tengen, check out senseis’ take on tengen continuations. A box of rocks, if you ask me.



For those who haven’t seen professional 9x9 play, I’ve commented a game played last year between two of Japan’s living masters. Happy viewing, friendly people of OGS!

review-31900.sgf (11.6 KB)


A perfect analogy. +1



Perhaps you have heard the proverb Five Groups Might Live But the Sixth Will Die. In 9x9, the proverb is Two Groups Might Live But the Third Will Die. I do not know of any dan-level game where white lived with three groups. Therefore, white may employ one of two strategic choices: (1) live with two groups or (2) make one group as big as black’s group (and komi is your friend). White may switch between these objectives based on the state of the board.

Note: two living groups is not always enough for the win. (If your territory is only two sets of two eyes, you’re not doing very well, obviously.) But the general rule is good enough because in practice it’s hard for black to reduce the territory of both groups at once.


White’s strategy comes first because black’s strategy is mostly a reaction against the other. Black’s strategy is about balance. On one hand, black strives to stop white from living with the second group. On the other hand, black has to watch that the former group doesn’t wrap around black’s group or grow in size to rival it. Black should win if both objectives are met—white has one living group that is smaller than black’s group.



Few things in life are as fun as statistics. And luckily I’ve got some right here for the 9x9 board. Huzzah!

Introducing: MARK5000’S 9x9 OPENING EXPLORER

In celebration of this release, I’ve put together a brief interview.

mark5000: Tell us a little bit about the Opening Explorer.

mark5000: Well, it’s built on the new puzzles feature and contains a modest library of pro games (266 to be exact) with detailed statistics of each move, such as players and their ranks, winning percentages, dates, even bot evaluation. It’s an ongoing project, but I hope to finish in the next couple of weeks as I find the time.

mark5000: Where did you get all these professional 9x9 games?

mark5000: A few places. GoBase was a very good start, though sadly lacking in metadata. I also got several games from World Baduk and an archive of RICOH Pair Go Cup games.

mark5000: What went into mark5000’s 9x9 Opening Explorer?

mark5000: My heart. All of it. My guts, my intestines, my diaphragm. I pulled a few muscles. I had a few breakdowns. If you saw what I did to make this database you would have called 9-1-1.


Super old topic, but I want to share a second opening explorer I made using only drawn games from >2800 ELO bots on the Computer Go Server. I figure it might interest someone other than myself.

The games are drawn because komi on CGOS is 7.0 under area scoring rules, which the programmers there consider balanced for computers playing 9x9. I figure these games are probably as close to perfect play as exists right now for 9x9.


Nice self necro :wink:

1 Like

Great topic though, thank you! :slight_smile:

1 Like

Neat. I made something similar years ago… my master 9x9 compilation. Sadly lost it to a fried HDD. Ever since studying Go Seigen’s 9x9 games on boosted my KGS2 rank by 4 stones or something, I’ve been hooked.

Slightly OT: I still remember that one 9x9 I just had to win (after a weak move in the opening, mind you). My study sgf reached upward of 80 kb…

1 Like

Hello Mark,

I do appreciate this new opening tree… I only play 9x9 and my favourite part is openings so I will enjoy this a lot.
I wonder, however, if it could be slightly biased towards black when playing with a 5.5 komi. In fact, all these games would have been lost by white with a smaller komi…

I have heard before about the comparison: “9x9 is a knife fight in a phone booth” but I do not totally agree with it. Of course, tactics is paramount in 9x9 but it is not something chaotical… I would compare rather with a 100 meter hurdles race.
It is very short. Starting (opening) is essential, it is difficult to recover when you are behind. A single mistake and you are out. Very technical (tactics). Short end part. Strategics has a role… (for a computer is all about reading but for humans It is impossible to accurately read to the end. Very often you must choose between moves from strategical (intuition) considerations only). I would not say 9x9 is easier than 19x19. It is obviously simpler, but It is much “easier” for a dan player to loose against a kyu player in the small board. (and It is supposed dan players excel at tactics…).

It is a pity there is so little bibliography (books, whatever, on 9x9 go). That is the reason it is so undervalued against its big brother. Thanks again, Mark, for you divulgative work.

Anyway, I am feeling a bit sad these days… bots including neural networks are improving more and more and I do not know what future holds for go…



Great analysis, Mark!

I notice Murakawa and Iyama playing with 6.5 komi whereas we on OGS are using 5.5. Have we ever actually reviewed White’s winning percentage in 9x9 here?

1 Like

The “test” is much more difficult than just comparing b/w winrate. You would have to factor out rank disparity (by default, the ostensibly stronger player plays white) and anomalies (stronger player playing black). I’m quite convinced that 7 komi with Chinese rules / 7.5 komi to break ties is the proper setting. This equals 6.5 komi under Japanese rules (unless there are 2 seki on the board). That said, I like to play white on 9x9. :slight_smile:

1 Like

I’m new to 9x9. How can the Komi be 5.5 or 6.5 on an 81 stone board when it’s 6.5 on a 361 stone board?

In the 19x19, the Komi is 1/55th of the total board.

In the 9x9, the Komi is 1/13th of the total board. :scream_cat:

To put it in perspective, that’s equal to a 27 stone Komi for a 19x19 game.

Shouldn’t the Komi on a 9x9 board be 1/55th of 81, or 1.5 stones?

(sorry, I’m sure this has been discussed ad infinitum here in the past, but then again, what subject hasn’t…lol) Thx

1 Like

Well, consider that the first move in 19x19 claims 1/361 of the board while in 9x9 it claims 1/81. Komi exists to offset the first move advantage, which is much larger in 9x9 than in 19x19.


What would be the max (most likely) amount of groups to live on 13x13?

1 Like

The thing is there are even much fewer study on 13x13 as on 9x9. Maybe phone size (and servers like Goquest oriented to 13x13) could make this size more popular and then more studied. But until then…
To answer your question it should be between 3 and 6 excluded so 4 or 5.
13x13 is much more near 9x9 as 19x19 so tend to think 4 as the upper limit of groups alive


Definitely does not work that way consider arbitrary board sizes.