3-3 invasion  

• c, 4-4 Point 3-3 Invasion, Extend - (joseki) (common) (beginner) (AI favorite)
• b, 4-4 Point 3-3 Invasion, Hane - (joseki) (common) (beginner) (AI favorite)
• a, 4-4 Point 3-3 Invasion, Knight's Move - (joseki) (common) (intermediate) (AI favorite)
• d, 4-4 Point 3-3 Invasion, Hane Below - (rare/situational)
• e, 4-4 Point 3-3 Invasion, Armpit Hit - (rare/situational)

Overview of Common Joseki

Common Sente Variations

Lee Sedol (9p) vs. Kang Dongyun (9p) (2009)  

In traditional go theory, the 3-3 invasion was not popular since it was considered to yield an inferior result for the invader. Historically, it would occur most often in the late opening. Amateurs would often play it in the late endgame, in which case it would be used as a speculative invasion.

The traditional textbook situation to perform a 3-3 invasion was when one player obtained a double wing formation.

In Lee Sedol vs. Kang Dongyun (9p) (2009), Black obtained a double wing formation (), which makes it difficult for White to approach the bottom right corner from the outside. The stones serve as loose pincers, and they prevent White from developing anything large on either side. Since the sides are not interesting, it makes sense for White to take territory with the 3-3 invasion (), and Black blocks on the larger side () to form a large moyo.

In traditional theory, this outcome was considered acceptable for White because one of Black's double wing formation stones () became inefficient. The stone on the bottom side is in an odd position, and it is not working together with the rest of Black's stones. For this reason, invading the 3-3 was considered proper if Black has a double wing formation. Furthermore, White has sente, which means White has the momentum to play a reduction later. That said, traditional theory would dictate that White should prevent Black from making a double wing formation in the first place.

In 2016, the AI revolution dramatically changed the way that humans view the 3-3 invasion. Strong AI programs often play the 3-3 invasion as one of the first moves in the opening (see: Leela Zero's Opening Gospel), and it is clear that AI considers the 3-3 invasion as one of the most valuable opening moves in the game. Strong AI programs have a territorial playing style, and they have an easy time reducing and invading moyos. For this reason, AI typically considers influence to be less valuable than humans do.

Many professionals and human players do not have predilection for making small territories and subsequently invading opponent's moyos, so humans generally invade at 3-3 point less often than computers do. That said, the 3-3 invasion has now become a basic move in the repertoire of human players, and it is considered a valid move in the early opening at all levels of play. Players should study new 3-3 invasion joseki, and the early 3-3 invasion can quickly favor one player more than the other if the players are not aware of the flaws of the traditional joseki.

Which way to block?

The general rule of thumb is to block on the wider side (i.e. the side with the most potential).

However, this is just an heuristic, and it's not quite a mistake to block on the "wrong" side, particularly since AI joseki often allow the defender to switch sides later if the situation calls for it. Furthermore, the double-hane joseki may have different considerations for which direction to block. Ultimately, the player should block in the direction that supports the joseki that they intend to play.

When to take sente? Knight's move or extend?

Lian Xiao (8p) vs. Chen Zijian (7p) (2020)  

In the AI era, the defender of a 3-3 invasion should generally play for sente by default. The extend (b) and knight's move (a) usually end in sente, so they are typically the main choices. Which one should White play if they both result in sente?

The extend (b) is often preferred by amateurs because it is simple and avoids all complications.

The knight's move (a) is locally a stronger move, but it can potentially lead to complications if Black attempts to resist. The infamous flying knife joseki can arise from the knight's move if Black has the ladder, so many amateurs avoid the knight's move. However, professionals and strong AI programs slightly prefer the knight's move by default.

If we examine the board in Lian Xiao vs. Chen Zijian (2020), should White extend or play the knight's move?

For professionals, the main consideration when choosing between the extend and the knight's move is the likelihood that the defender will get the next local follow-up in the corner. The three-stone hane shape that arises from the knight's move is more unstable, and the defender would like to come back to play the continuation first. Even if White opts to temporarily tenuki, White should calculate whether they will later be able to return to the corner first. In professional games, White gets the first follow-up move 64% of the time^[1], indicating that the follow-up is valuable.

In contrast, when White plays the simple extend and Black slides with a small knight, the local follow-up is not as interesting for White, and the local shape is stable. Often, White will not come back to play in this corner again, and Black will eventually come back to get the first local follow-up move ~50% of the time^[1].

A simple heuristic for understanding this is that if White owes an urgent move elsewhere on the board, White should extend so that White can address that urgent place on the board (usually in gote). That way, when it is Black's turn, Black does not have an interesting follow-up on the corner.

Considering the board position in Lian Xiao vs. Chen Zijian (2020), White does not owe any moves and nothing stands out as particularly interesting. Since there is nothing interesting on the board, the knight's move at a is the choice preferred by professionals and strong AI programs. In fact, the top left corner is the most interesting area of the board, and White immediately plays the follow-up hane at e after forming the three-stone hane shape.

Summary: The follow-ups to the three-stone hane shape are big moves. Ideally, White should not to make the three-stone hane shape if White thinks there is a high chance that Black will play the follow-up first.

When to double hane?

Fan Yunruo (8p) vs. Chen Yaoye (9p) (2020)  

The double-hane joseki is considered an even result on an empty board, but there are many situations when it can become an especially favorable joseki for the defender.

In Fan Yunruo vs. Chen Yaoye (2020), Black has a low position on the upper left (), which makes the upper side of the board uninteresting for White. This area is not easily developable by either player, so it has fairly low value.

The double hane with is favorable for Black because White's ponnuki will not be very useful. The ponnuki is a strong group that projects powerful influence (see: A ponnuki is worth thirty points), but Black already has a group on the top side that neutralizes the value of the influence. Meanwhile, Black managed to take back all of the corner territory, so White's profit is limited.

Identifying good situations to utilize the double hane is a valuable skill. Although this variation is rarely played as the first joseki in the game, it can be a good way to respond a delayed 3-3 invasion when more the board has been explored, and it is frequently used as a strategy to overconcentrate one's opponent.

When to play the hane below?

Cho Chikun (9p) vs. An Kwanwuk (9p) (2018)  

The hane below is a fairly rare move that is not seen in most joseki dictionary, but it is still played by professionals in fairly unique situations.

According to josekipedia, "When black has no real profit to gain at top, e.g if white is really strong around (on the top), he can play like this, building a solid wall without weakness. It can also be played if black has a close double wing formation, where black can stop whites expansion."

Cho Chikun vs. An Kwanwuk (2018) illustrates the latter case, where Black already had an existing set of stones () that block White's expansion. Ordinarily, allowing White to escape along the side is a mistake, but Black utilizes his surrounding strength to launch an effective attack on White's invasion.

It is noteworthy to mention that Black blocks with on the bottom side. Blocking in the other direction would cause Black to be overconcentrated with two narrow walls, which is very unattractive. Playing in this way allows Black to gain strength to launch a second attack on White's stones on the right side, and Black can maintain momentum by staying on the offensive.

Tenuki is bad

Tenuki is bad  

Normally Black should not play tenuki at this point, since allowing is very painful.

See Also

• 3-3 Point Invasion — An introductory article for beginners
• All about 3-3 Point Invasion — An index to the articles on specific 3-3 invasion joseki, their variations and to various queries about them, ee.g:
• (Youtube) Ryan Li (3p): Common 3-3 Invasion Joseki (2019)

References

• [1] Frequency statistics were obtained from Waltheri's Go Pattern Search using the full database restrained to a local search (accessed September 2021).