Principles of 9x9 Go

Many months later as I sit here today, I recognized that no book by me on 9x9 Go is getting written unless I sketch it out here first. Consider this a lesson series on setting up winning positions on the 9x9 board. If you can do this well, you may find a great many transferrable skills to boost your 19x19 game too.

Lesson 1: Attach and pull back

We start from the popular “Three in a Row” opening. This opening typically follows the order given, but you can also reach it by the sequence 3-2-1 without misplaying. The knight’s move to 4 then is the most common response. Some people first attach at H5 as a probing move. The intention is to choose the preferred knight’s move for move 6 once symmetry is broken. In truth, White can play either way. The probe reflects strategic thinking but is tactically no better than playing the knight’s move first.

The maneuver of 5 and 7 is our focus here. It is called “attach and pull back” (or tsuki hiki, in Japanese). It is typically contrasted with “attach and extend” (tsuke nobi), in which move 7 is at C7. But that is a common misplay here.

Suppose Black plays move 7 at C7 to carry out the attach and extend maneuver. White pokes a vital point with 8. And unless Black wants to make an empty triangle, move 9 is inevitable. White can then jump to 10 without worry and secure an advantage in territory, once komi is factored in. Black can’t disconnect 10 through moves 11 to 15 because White can cut it at A.

Even if the cut at A above didn’t work well, White still has an amazing tactic: move 16 here. It threatens to link under at H5. If Black wants to prevent this by kicking at 17, White can springboard to another vital point at 18 and threaten to cut the knight’s move that the attach and extend maneuver had left behind. If Black had attached and simply pulled back afterward, this cutting weakness would not have lingered. The term for this in both English and Japanese is aji. On the 9x9 board, as on the 19x19 board, it is often best to minimize aji and connect solidly in positions like these.

Suppose again that Black understands this lesson and pulls back for move 7. I strongly suspect (having conducted only medium-depth AI analysis) that move 8 now becomes an inaccurate play for White. Although the same general tactic is conceivable for White, the lack of a cutting weakness in the black stones means that Black need not spend a move defending the cut. Instead, Black may attack straight away with move 17. The white stones in the upper right can no longer survive, so long as Black keeps playing accurately. And that last part is important. It’s not enough to attain a winning position if you let your guard down later on. The 9x9 board demands just as much focus to win at the higher ranks as the 19x19 board does, if not more so.

Please comment below if the topic suits you. And see you all the next time.


The topic suits me, It’s amazing. Thanks for sharing.

One very first thing i get from 9x9 is how a group of stones can be quickly threatened even if it looks like it has some breathing space. It makes me more cautious on a 19x19


This is great! I often wonder how it is that I can lose at 9x9 against a group just in the centre with no territory…


Great idea! I would like to read more :grin:

The principles of 9x9 on Sensei’s really helped me to improve at 9x9 when I was DDK level (SDK now): TheArtOf9x9Go/9x9 proverbs at Sensei's Library

To quote just a few to whet your appetite:


Top content, Mark!

Thanks a lot.


Lesson 2: Use the knight to attack

The Go proverb that this lesson is named for applies even more sublimely in 9x9 Go than it does in its original 19x19 context. In fact, the knight is the only reason some openings are viable. And move 1 may be one of these. When Black and White face off in the center like this, each hopes to claim their respective side areas. But if they carry out these objectives peacefully, Black will lose by komi. To overcome the komi disadvantage, Black must attack. And the knight is the only effective attack here (it is unclear to me if a one-point jump is also viable).

Black 3, probing White’s underside, is among the most popular moves in this opening. It is also wrong. White blocks at 4, Black pulls back (recall lesson 1), and White descends to 6, making good shape and making it extremely hard for Black to attack White effectively. For example, Black’s clamp at 7 fizzles because of White’s miai invasion at 12. With either G3 or G7 wide open, Black lacks a good reply, and has fallen behind in territory.

If Black plays the knight for move 3, White commonly responds in kind at 4. Black has about four good choices for move 5. Among them is another knight, this time cutting the white stones. Although White can capture the cutting stone, Black can get benefits on the outside with 9 and 11 that make up for it. The opening is still balanced.

By the way, if Black opens on a nonsymmetrical point like F5, White can often use a mirroring strategy. But the knights at 3 and 5 counter a mirroring strategy. White can’t mirror move 5 because Black can take the center point right afterward and disintegrate White’s position.

Move 5 can also be at F3, as shown above. White mirroring this move makes an opening that on this server has been called “Andromeda.” Black can attack with the knight after this, too, although the pattern differs. If White moves to capture the cutting stone, Black must respond at 11 to keep a balanced position. This critical move makes miai of plays on the top and lower sides.

Things won’t end well if Black instead plays move 11 here. This hasty play does not respect the exchange that has occurred. This time, White would capture the cutting stone at 12 and make miai of A and B. The situation now favors White. So make sure to play each position according to the stones on the board. A single exchange can make all the difference.


My principle of 9×9 Go is:

To decide if a move is good, read to the end of the game and see if it makes you win.



Lesson 3: Opening at 3-3 is bad

The 3–3 point is a playable opening at the kyu level, but it is rarely played by dan players and almost never played by professionals. There is a reason for this. Beginners often worry about losing a group, and so they tend to mark solid territories and play close to the corners to keep their positions secure. But if the beginner sticks with Go for an appreciable time, they will eventually learn that active play is good and passive play is bad. Active play, of course, refers to taking the initiative (sente), attacking the opponent’s positions, and building big potential territories. Active play is especially desirable for Black, which needs to overcome the komi disadvantage. Passive play, on the other hand, refers to “slow” play, giving up the initiative, and building smaller territories than a player may be entitled to. Knowing these two concepts often leads to an attitude adjustment; the player begins to feel differently about the board and play more boldly. And so the 3–3 point falls by the wayside. And it should. AI analysis confirms what the professionals suspected all along: it is a poor opening move for Black.

How then should White answer the 3–3 point opening? The strong tendency is to respond in kind and take either the opposing 3–3 point or a similarly territorial one, like an opposing 3–4 point. But White is better served by occupying a high position, such as the opposing 4–4 point or a shoulder hit.

Let’s consider the opposing 4–4 point first. Black commonly responds with another territorial move like 3 here. In response, White can simply enclose the corner, either high or low. The positions of the black and white stones may look similar at first glance, but don’t be fooled. If you count the points, White has encircled a few more than Black has.

Black naturally wants to keep expanding the reach of Black’s group, since its reach was so slight to begin with. White then should stand up at 6. If this seems strange to you, then remember the concepts of active and passive. A White play at 7 would have been gote and somewhat passive. White 6, by contrast, demands a reply from Black, keeps the initiative, and allows a double enclosure / shoulder hit at 8. Again, looks can be deceiving. Despite occupying three corners, Black is behind in territory once komi is factored in.

The other really good response to Black’s opening at 3–3 is the shoulder hit. Black typically hits the opposite 3–3 point at A or replies locally at B. But in either case, White is better served by occupying a high perch from which to bully Black. I’ll leave the specifics as an exercise for the reader. I should also add that the 3–3 point is often a perfectly viable choice for both players after move 1, including for White’s first move. It’s just slow enough, and the pace of 9×9 Go is fast enough, that playing it for move 1 is less than ideal.


Lesson 4: Give way and take sente

The “pendulum” opening is the name sometimes given to the unmarked moves above. When these stones are played, White’s descent to move 4 is a common but inaccurate idea. Black can gain a slight advantage by using the knight to attack (recall lesson two). In fact, this was the second time in a row that Black used the knight to attack. But it can be hard to convert a slight advantage into a win, and the obstacle to winning here is correctly judging the value of the initiative. The value of initiative is highly variable and depends on the characteristics of each position. Here, the initiative is critical to White, which cannot win without forming a second group. And it is critical to Black for defending the upper-right area, which is wide open to invasion by White. Because the initiative is more important than usual, Black must be willing to accept a small local loss in pursuit of the initiative.

Although White could seize the initiative right away, this would come at probably too high a cost. Black would descend to 7 in sente and then pivot to a dual-purpose attack and defense at 9. This maneuver threatens the stability of both white groups—neither of which is alive yet—and makes good territory for Black at the same time. White can’t be satisfied with this result. For this reason, White typically replies locally for move 6, even though the initiative has above average value.

The intuitive and most common move for White is the hane at 6. Answering this move can be tricky, and even AI can’t find the ideal path forward at lower playouts. Playing elsewhere isn’t the answer for reasons that should be clear enough. But Black can’t play softly at 8 either, because White will have secured the lower left group in sente. In other words, White got everything White wanted, which is really unacceptable. So Black must counter hane at 7 and block on the outside at 9. This tactic can be hard to find because it involves a local loss, but this loss is worth the initiative here. Black can defend against the most severe invasions while loosely sealing White’s entrance into Black’s lower area. That’s everything Black could hope for.

The second most common move for White 6 is the clamp. This clamp secures the lower left group by targeting the black stone clamped. Trying to save this stone would be misguided. Instead, Black should counter clamp (you may start to see a pattern here) and seize the initiative. Again, the local loss is worth taking the initiative here. To be clear, the initiative is not always worth taking a local loss. It is up to your reading and judgment based on each position. But in general, wide open areas increase the value of the initiative, while closed areas decrease the value.


Lesson 5: Passive plays need aggressive follow ups

“Active” and “passive” aren’t binary concepts. Most moves that you could play lie on a spectrum from extremely active to extremely passive. Moves on the passive side of the spectrum aren’t always poor choices, but to make them good you need to balance them with active, even aggressive plays. This goes against how players playing passively often want to play, but compounding passivity upon passivity brings sure defeat. Some great examples occur right out of the opening from third-line moves, which are somewhat passive compared to plays on the fourth line or the center point. For these examples, I’m going to refer to OGS stats from Yannikkeller’s Go 9x9 Opening Explorer.

White 2 is the most common response to Black’s opening at the 3-4 point. But OGS players find the best choice for move 3 just 1% of the time. In a sample of 18,464 OGS games, Black attached at E4 221 times (and several of those were me), and it was the 16th most popular move. If you miss this move, White’s odds of winning are 2:1, but if you play it, then the odds favor Black. It’s a big difference. The reason for this has to do with the somewhat passive character of opening at the 3-4 point. After playing somewhat passively, Black needs to press aggressively against White’s position to maintain the balance of territory.

The need for aggression doesn’t end here. If White then stands up for move 4, then Black must hane (the third most popular choice). And if White follows at 6, Black must connect. This opening favors Black when komi is 5.5, but the correct sequence is elusive if you never saw it before. Although I lack the comprehensive AI analysis to back this up, I believe based on my study with AI that this line is the only “perfect” play when White responds as shown. If you know of another non-lossy line, please inform me! At the very least, this is a great line for Black, and very underplayed.

Another example arises from the 3-5 point and White 2 above. The best and possibly only viable choice for move 3, according to my AI analysis, is attaching at the center point. But out of 3,631 games, OGS players chose this move 260 times, or 7% of the time, and it was only the fifth most popular move. As before, Black must balance a somewhat passive opening move with aggressive follow ups.

Suppose White stretches for move 4. To preserve the territorial balance, Black should hane at 5. Playing 5 at 7 would let White play 6 at 5, and White will enjoy a territorial advantage and likely sail to victory. Only after exchanging 5 for 6 can Black play 7. After Black 7, White can put the stone at 5 into atari in two ways, but Black can manage either one. More importantly, Black kept the balance of territory by following up the somewhat passive play at 1 with pressing firmly against the white stones.