4-4 Point AI Approach Joseki

    Keywords: Joseki

Return to parent joseki page: 4-4 point low approach low extension, contact, atari down, connect

The 4-4 point low approach, low extension, contact, atari down, connect, connect, more concisely known as the "4-4 Point AI Approach Joseki", is the most common star point side approach sequence in the AI era. It is considered to be the spiritual successor of the 4-4 Point Traditional Slide Joseki, although there are some nuances distinguishing the two. In particular, this AI joseki has a more asymmetrical result, often with one player holding more influence (or potential) while the other player holds more territory.

At this point in the joseki, Black must make a choice between influence and territory. The former is simple, but Black recognizes that there is a chance they may lose the corner. The latter is aggressive, and it is often played with the expectation that it may result in a complicated fight.

Tenuki is uncommon but still possible, particularly when something else on the board is even more urgent.

Table of contents

Quick Navigation

Moves are listed by frequency in professional games[1], which is sensitive to whole-board position. Bolded moves are commonly considered joseki.

4-4 Point AI Approach Joseki
4-4 Point AI Approach Joseki  
  • a, Push? (~47%) - (joseki) (common) (AI favorite)
  • b, Descend (~43%) - (joseki) (common) (AI favorite)
  • c, Small Knight (~5%) - (situational)
  • t, Tenuki (~5%) - (situational)


Take Influence

Influence Variations

Small Knight
Small Knight
  • The B1 push is the most popular continuation (Dia 1) and it is well-regarded by AI on an open board. This move prioritizes influence towards the left side of the board and occupies a vital point that is important for Black's shape. It is also diminishes White's potential to grow on the top side of the board, which makes this position a key focal point when both players have a moyo on their corresponding sides of the board. These three qualities are so important that Black is frequently willing to cede the corner territory in exchange for big-picture benefits, which represents a notable departure from the traditional corners then sides then center go proverb. Given that both Black and White have settled groups after the B1-W2 exchange, the corner territory is no longer crucial to the life and death of either party. While capturing the corner with z (or defending the corner with c) is an enormous move for points, it is gote and considered slow for either player in the early opening.
  • The B1 small knight (Dia 2) is a special-case move that is almost exclusively played when Black has a moyo on the left side and when Black is unwilling to push due to the spacing of stones on the top side of the board. Before playing this move, Black should ask themselves: "Why not push once in sente and then play the small knight for influence?" If White has a loose moyo on the upper side that still has potential for an invasion, perhaps Black does not want to strengthen White's top any further. The immediate small knight encourages White to capture the corner at z, as neither player is settled. However, this typically makes White thinner on the top side. Black can aim to attach at b, which helps Black develop influence towards the center of the board.

Take Territory

The B1 descend is a territorial move that aims to attack White's group. It is most effective when White lacks potential or is weak on the top side of the board. This move demonstrates fighting spirit and can potentially involve very complicated variations. At the highest levels of play, it involves the consideration of at least two ladders, whole board judgements about the direction of play, squeezes, and potential trades. Due to the difficulty of some variations, it is sometimes feared and avoided by weaker amateur players. Similar to the 3-3 point flying knife joseki, there can be a psychological element to this joseki at the amateur level, especially as humans may attempt to guess what their opponent is likely to play or avoid variations that they are less familiar with.

Simple Variations

  • The W2 one-space jump is simple, calm, defensive, and avoids all possible complications. This variation is recommended for double digit kyu players. From a professional perspective, this variation is slightly less satisfactory for White. However, professionals will frequently play this way if they are concerned about the peep. White's group is completely alive, but its potential to develop further is somewhat lackluster.
  • The W2 push is White's most powerful move, as it takes advantage of Black's shape weaknesses and shortage of liberties while occupying the focal point for whole-board potential. If White hopes to develop anything large on the top side of the board, they must occupy this point. The W2 push even helps White on the opposite side of the board, since various moves on the left side such as y and z are more likely to be sente when Black has a shortage of liberties. However, the W2 push risks complications, and White must consider the possibility of a fight. Strong players will almost always peep at a, which can result in very complicated joseki. In some variations, White may even run into the center as a one-eyed dragon. Still, White's group is more resilient than it might appear.
  • The B3 peep at a is not advised for kyu players who have not studied this position, as it can become difficult for weak players to handle if White resists (see below). Instead, players unfamiliar with this position are can be recommended B3 at b or c (Dia 2), which is simple and avoids complications while achieving something similar. Jiang Weijie recommends b in the AI Weiqi Joseki Dictionary if Black is concerned about a potential hane variation (which involves a ladder) (see below), although it is infrequently seen in professional games.
  • Connecting with W4 accepts a fight. This is best when White believes there is a good chance White can counter-attack the peeping stone(s). It can also be considered if White sees value in the left side of the board. Moves such as c, e, and f (Dia 3) are generally sente against the left side. A move like z is defensive and looks to make miai between the left and top; it shown in Jiang Weijie's AI Weiqi Joseki Dictionary, but it hasn't been played very often by professionals, perhaps because it is more passive. White's group is more resilient than it may appear, and it is even fine for White to tenuki.
  • It is important to recognize that Black's B3 and B5 stones are not strong unless Black has existing support on the top side. Black must obtain another move at the marked square (vital point) in order to fix their shape. In the three-space extension (Dia 3), White cannot permit Black to make such beautiful shape, and a move around a and b is most common. Black will usually treat the B3 stone lightly and aim to achieve something else with the B5 stone. The two-space extension (Dia 4) has a similar issue whereby Black owes a move at the marked square (vital point) before they can launch a serious attack on White. However, Black's position is relatively cramped, and professionals will often immediately pincer at a. Black's group is ultimately still weak, and White has the initiative once they jump into the center.

Intermediate Variations

Common Var.
Common Var.
Gote Capture
Gote Capture
Jump-Alt Var.
Jump-Alt Var.
Drawback Var.
Drawback Var.
  • White may resist the peep with the W4 one-space jump (Dia 1) if White has potential towards the top side of the board or if there are additional stones on the board that support a special strategy (see below). It is important to recognize that the spacing of White's potential on the top side of the board is an important consideration, as the common variation (Dia 2) is probably not recommended if White ends up overconcentrated. The common variation is an even result for both players on an open board, but it is not necessarily favorable outcome if White has too many stones on the top side.
  • The basic premise of this position is that Black threatens to capture White's four stones on the inside. There are multiple methods for Black to capture the corner in gote, but some variations can be more or less tolerable for Black depending on the degree of thickness that White obtains on the outside. White can be satisfied with this outcome if White obtains a squeeze on the outside, since Black will have invested one extra move on the inside. The B4 turn threatens to escape (Dia 2), which White can prevent if White plays a tesuji involving the W1 cut in advance (Dia 2).
  • The first branching point in this joseki is shown in Dia 1. The cut with W6 at a is played if White intends to pursue the common variation or hane variation (see below).
  • In the common variation, usually both players will tenuki once they reach this board position (Dia 2). Black can capture the corner at z (Dia 3), which is worth ~20 points in gote, but this is too slow in the opening and gives White a squeeze for impressive thickness. In fact, it is relatively rare in professional games for Black to capture the corner at all. It is more common for professionals to use the aji in other ways. For example, any move on the top side of the board (at w or even further away) can activate the aji at x, which can threaten to resurrect Black's stones. A move like y is also sente, and it is another method for Black to gain thickness on the left side of the board. Consequently, in professional games, it is ultimately more common for White to capture at the three Black stones by the end of the game.
  • Jumping first with W6 at b (Dia 1) often transposes to the common variation (Dia 2), but it can only be played if White has the ladder on the bottom right (Dia 1). In special cases, White can also pursue an alternative variation if White has a supporting stone on the upper right (Dia 4), in which case White forms a wall on the outside with another Black group sandwiched in the middle.
  • Drawing back with W6 at c (Dia 1) is rare, but played when White wishes to make Black heavier before allowing Black to escape out from the side. This is ordinarily not good for White unless White can mount a convincing attack on Black's group. This variation is mainly only seen in special circumstances when White already has another group on the top side, so White aims to take a peep around x.
  • Please note that there are additional variations in addition to these shown in the overview here. Please see the dedicated articles for a further discussion.

Advanced Variations

Hane Var.
Hane Var.
Black Collapses
Black Collapses
Hane-Atari Var.
Hane-Atari Var.
Hane-Atari/Turn Var.
Hane-Atari/Turn Var.
Lost Ladder
Revert Var.
(Bad Ladder)[1]
  • The most complicated variations involve the W6 hane (Dia 1), which fierce and territorial. This move is most effective when White has an existing ladder breaker in the bottom right. Joseki surrounding this position are extremely complex and involve wild unstable fluctuations in katago's evaluations (c. 2023), so it can be considered an area of active research. Players reviewing this position with AI should be very cautious and skeptical about AI recommendations made from low playouts.
  • According to Jiang Weijie in the AI Weiqi Joseki Dictionary (2021), if White has a ladder breaker in the bottom right, Black's best option is to tenuki from the position entirely (Dia 1). With an unfavorable ladder, Jiang Weijie calls the B9 resistance at w an "overplay". It is better for Black to play elsewhere until the status of the ladder changes.
  • The relevant ladder is illustrated in Dia 2. Realistically, this position is never seen in actual games because it is the worst possible outcome for Black. Typically, Black will cut their losses early and accept a trade before reaching this point. While there are many potential variations, White can generally capture something, whether that is Black's peeping stones the middle or Black's three stones on the left.
  • The W1 hane is also reasonable and played by professionals when White has a supporting stone on the top right (Dia 3). In this case, White plays the W3 atari and White is able to crawl out on the second or third line.
  • Black has the option to turn with B6 at w (Dia 3). This variation sets up a large complicated sacrifice and squeeze whereby Black obtains a large thick wall facing the left/bottom side (Dia 4).
  • If White loses the ladder breaker (Dia 1), White can transpose to the atari variation with W5 in Dia 5. If White crawls three times on the second line, white has enough liberties to capture the two stones in the inside. However, Black can squeeze from the outside.


When to push?

Chen Xinyang? (2p) (B) vs. Wu Junyi (2p) (W) (2023)  

yuzukitea: The push for influence is a general purpose move that is well-regarded by AI and professionals in the early opening. It is playable whenever the defender has potential facing in the direction of their wall. Even if the defender lacks potential, the push can still be played to reduce the approacher's potential on the opposite side of the board.

Accordingly, when both players have potential on their respective sides of the board, the push is even more valuable.

Chen Xinyang? vs. Wu Junyi (2023) is an example that illustrates this concept. In this game, Black has potential towards the bottom of the board with a one-space high enclosure. Meanwhile, White has potential on the left side, particularly with an attractive pincer around z. It is fairly intuitive to see how the push can expand White's potential while simultaneously shrinking Black's potential.

It is important to note that the bottom side of the board is fairly open. Even after the push, there is still plenty of space for a future reduction or invasion during the middle game. As a result, White is not concerned about strengthening Black's position on the bottom left. If the spacing of stones is more of a concern, sometimes the immediate small knight is an alternative to the push, especially if both players are pursuing a large moyo strategy.

When to descend?

Nakane Naoyuki (9p) (B) vs. Kim Chong-su (9p) (W) (2023)  

yuzukitea: The territorial descend is another general purpose move that is well-regarded by AI in the early opening, although its popularity seems to have have declined slightly among professionals in recent years (c. 2023). It is playable in most circumstances, although several conditions make the descend even more favorable:

1. The approacher lacks potential on their side of the board.

2. Influence isn't as valuable for the defender.

3. The defender has supporting stones on the opposite side of the board.

4. The approacher isn't likely to resist a peep (e.g. lack of a ladder breaker) (minor factor).

Nakane Naoyuki vs. Kim Chong-su (2023) is an example of a game that fulfills all of these features. Black's approach is facing the "wrong side", so it is unlikely that Black will develop anything large on the left side the board, especially since White has a white+circle enclosure blocking that direction. Meanwhile, White's corner is also facing Black's enclosure, so neither players have much potential.

The white+circle stone is conveniently positioned in a location where it would be able to assist White with a future peep at z. If we imagine that Black connects the peep, White can easily obtain a nice loose framework of stones on the left side. It is unlikely that Black will resist the peep, because Black lacks potential to the bottom left, supporting stones, or an existing ladder breaker in the bottom right. (Note: the status of the ladder is only a minor factor for decision making; the W8 descend is still good even if White avoids the W10 peep and plays a more conservative move such as the W10 pincer at n).

Taken together, the W8 descend at a paints a target on Black's back. It announces an intention to attack, and there is a significant possibility that Black's group will end up as a one-eye dragon in the center of the board with little to gain.

Unwilling to risk this outcome, Black plays submissively with B9 at n instead of strongly at m.

Hong Seong-ji (9p) (B) vs. Kim Seongjae (8p) (2023)  

yuzukitea: There are numerous other fuseki patterns where professionals commonly play the descend. Hong Seong-ji vs. Kim Seongjae (2023) is a game that illustrates the general idea.

In this board position, we can see that the black+square stone is weak and White would like to obtain a move around y or m. If Black connects in response to the W10 peep, White can make an extension on the board of the board that serves multiple functions. Multi-functional moves are often strong moves, and situations like this can make the W8 descend more appealing for professional players. Moreover, the left side of the board is uninteresting due to the presence of the black+circle marked stones, and White's potential is limited.

However, do you think Black will connect in response to the W10 peep?

Try to imagine plausible resistances that Black could play:

  • Jump-Alt Variation more likely (due to presence of black+square stone)
  • Hane-Atari Variation more likely (due to presence of black+square stone)
  • Common Variation less likely (White can obtain a free move on the bottom due to aji of the three peeping stones)

Ultimately, in this game, White chose to play W10 at m, presumably because he was concerned about a possible resistance sequence. Black protected the left, and White achieved his desired profit by capturing the black+square stone at n.

When to small knight?

Choi Kwangho? (B) vs. Kim Kiyoung (W) (2018)  

yuzukitea: The small knight is a situational move that strongly emphasizes influence in the center of the board. It is most commonly played when both players are pursuing a moyo strategy and the defender wishes to keep the attacker as thin as possible.

Choi Kwangho? vs. Kim Kiyoung (2018) is a game that illustrates these concepts. In this board position, both White and Black have the potential to make large a moyo. In particular, the white+circle stones make the center even more appealing than usual. Since influence is clearly valuable, White's best option is to play W8 at either a or b.

Pushing once with W8 at b and then playing the small knight's move is an idea, but the W8 push strengthens Black on the bottom of the board. It induces Black to fix the spacing of their stones, as the relationship between the B9 and black+circle stone is too convenient. It will likely be more challenging for White to play on the bottom side in the future.

Since the W8 push helps Black, White chooses to play W8 directly at a instead. The immediate small knight encourages Black to capture the corner with B9 at at y, but in compensation White aims to attach with W10 at z. This greatly weakens Black's potential towards the bottom and provides White the outline of a large moyo towards the center of the board.

When to tenuki?

Zhou Hongyu (B) vs. Yan Ximo (W) (2023)  

Tenuki, while not very common, is possible if there is something else on the board that is more urgent than this corner.

In Zhou Hongyu vs. Yan Ximo (2023), White was afraid that Black wouldn't respond if she pushed with W8 at x. Black already has a strong group (black+square stones) on the left side, so there is much less of a compelling reason for Black to create a second strong group on the left side.

Black's two black+circle stones are more urgent, and White assessed that there was a significant risk that Black would tenuki if she pushed at x.

As a result, White chose to play W8 at a directly.

See also


[1] Frequency statistics were obtained from professional games downloaded from kifudepot? (Jan 2016 - Nov 2023) constrained to a local search.

[2] AI Weiqi Joseki Dictionary (2021) by Jiang Weijie

[3] An earlier variation (the tiger's mouth) transposes to this position, so the percentages are inflated over 100%.

4-4 Point AI Approach Joseki last edited by yuzukitea on December 12, 2023 - 22:30
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