How bad is the attach and block joseki?
It has always made me feel uneasy, because it accepts a "bad" shape.
I am planning to show some pro examples from the Elf GoGoD commentaries, but feel free to join in with your own examples and comments.
| Life In 19x19 http://prod.lifein19x19.com/ |
|
| Attach and block joseki http://prod.lifein19x19.com/viewtopic.php?f=15&t=17481 |
Page 1 of 1 |
| Author: | Bill Spight [ Tue May 12, 2020 9:07 am ] |
| Post subject: | Attach and block joseki |
How bad is the attach and block joseki? It has always made me feel uneasy, because it accepts a "bad" shape. I am planning to show some pro examples from the Elf GoGoD commentaries, but feel free to join in with your own examples and comments.
|
|
| Author: | Bill Spight [ Tue May 12, 2020 9:27 am ] |
| Post subject: | Re: Attach and block joseki |
Example 1 Ding Lie, 5 dan (W) vs. Zhang Yingting, 5 dan GoGoD 2008-02-25q Elf says that  loses 6% to par, but then  returns the favor, losing 6½% to par; minor errors.  , however, loses 9% to the kick at  . (See next diagram.) What about the block, ? Elf prefers the nobi at 10, but only by 1%. The block is surely playable.  In its mainline variation Elf chooses the kick, and after White stands, the one space jump. Then White invades the top right corner. In this case the problem lies with the attachment, not the block, according to Elf. |
|
| Author: | Bill Spight [ Tue May 12, 2020 10:20 am ] |
| Post subject: | Re: Attach and block joseki |
Example 2. Yamashita Keigo, 9 dan (W) vs. Yamada Takuji, 7 dan GoGoD 2011-03-31i Elf thinks that loses 6% to the invasion of the top left corner at a, and that  loses 5% to the invasion of the bottom right corner at b; minor errors. This time the jump attachment, , is Elf's top choice. But the block,  loses 5% to the nobi at 12, another minor error. Elf's mainline variation is of some interest. It is fairly long, based upon 77.3k rollouts.The kosumi,  , avoids strengthening Black. Ditto  , but the cut is playable. Black continues strongly with  .Now  cuts. Then White suffers being forced to crawl along the 5th line.   stakes out a large territorial framework.  is the last play Elf shows, with 2.3k rollouts. The next play, if any, gets fewer than 1500 rollouts. You can't be sure of plays that get few rollouts at the tail end of variations. Elf's verdict in this case. The jump attachment is fine, but the block is a minor error. |
|
| Author: | Bill Spight [ Tue May 12, 2020 11:37 am ] |
| Post subject: | Re: Attach and block joseki |
Example 3. Yi Chuyong, 6 dan (W) vs. Seo Pongsu, 9 dan GoGoD 1991-03-28c Here is a bread and butter example of the attach and block joseki, if there ever was one. Elf says that the jump attachment,  , loses 5% to the kick at  , a minor error. The block, , loses 3% to the nobi at . For other bots a 3% difference might indicate a minor error, but Elf is fairly opinionated. IMO, the block is playable. Elf's variations look interesting.White plays the kick, (30k rollouts), followed by the one space jump. approaches the top left corner and is a mild pincer, looking both ways.Black plays a double approach with  (6.3k rollouts). Then White attaches and extends. Black now makes a small base with . cuts and the fight continues. Surely  can't be the end, but now we are down to 1.6k rollouts, speculative territory. Next, Elf's mainline variation for the nobi. Elf prefers the nobi, , with 57.7k rollouts. and are usual, then approaches the top left corner.  attacks the Black stones in the bottom left. Both players continue strongly.  and block, and and make a 2 step hane.  extends. makes a solid connection (9.8k rollouts). Black bolsters his position with - . White presses with . Then Black makes a half eye in the corner with sente. The turn,  , is a strong, large scale play. Elf leaves us in media res again. But we are down to 2k rollouts for .In this case, it appears that the jump attachment is a minor error, but the block is probably playable. |
|
| Author: | John Fairbairn [ Tue May 12, 2020 11:40 am ] |
| Post subject: | Re: Attach and block joseki |
Quote: How bad is the attach and block joseki? It has always made me feel uneasy, because it accepts a "bad" shape. For the record, I said that Shibano wrote about tsukeosae having gone right out of fashion in the AI era, but he didn't quite say it had gone out of fashion because of AI. He said it enjoyed two waves of popularity, both before he became a pro. But humans had already concluded that it was usually best avoided. The context he is talking in is not AI research as such, but rather "why have certain fuseki or joseki moves disappeared." AI is just one tool to help with the answers. But it appears that, with AI help, pros can be a little more certain of conclusions they had already reached. There is an interesting example in this case. He starts off by saying that the reason tsukeosae disappeared was, in a nutshell, because of the empty triangle. However, he also gives another way of looking at it (which is more attuned to AI insights, perhaps). And that is that the tsukeosae side should feel uncomfortable about giving the opponent the atari that creates the empty triangle. With this important difference: he does not call it an atari. he calls a tataki (a really strong atari, if you like - a strike rather than a hit). I think everyone can see how that might play into the typical AI centre-focused view of the go board, even if, like me, they don't understand the subtleties of how that tataki may reverberate. But not every position is purely centre-focused, Shibano points out, and there are cases where the "greedy" joseki (jikara joseki) can work out. He summarises all that by saying that rather than regarding the tsukeosae as a go-to joseki, it should be regarded as a baai no te - a horses for courses move. |
|
| Author: | Bill Spight [ Tue May 12, 2020 3:27 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 4. Chang Suyeong, 8 dan (W) vs. Seo Pongsu, 9 dan GoGod 1991-11-14b This time Elf says that the jump attachment is correct, but the block loses 6½% to the nobi. Here is Elf's mainline for the nobi. and are usual. Then approaches the top right corner. Black ignores the approach and makes an approach to the top left corner. - bolsters the White stones in the bottom right. Then makes a double approach in the top left.In this case the jump attachment is fine, but the block is a minor mistake. Note that this is essentially the same verdict as for Example 2, where the bottom side stone is on the 3d line instead of the 4th line. And the recommended play after the nobi is quite different. |
|
| Author: | Bill Spight [ Tue May 12, 2020 4:36 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 5. Seo Pongsu, 9 dan (W) vs. Yang Chaeho, 7 dan GoGoD 1991-95-92b It's Seo Pongsu night! This time there are enclosures in the two adjacent corners, which affect the play. For Elf prefers the 3d line attachment at a, but only by 2½%. The jump attachment is playable. However, Elf prefers the nobi to the block by 5%; so the block is a minor error. and are usual. With the enclosures in place, White does not tenuki, but plays and cuts after the Black hane. extends into the center. Then plays the S-17 attachment. At this point Black switches to the 3d line attachment in the top left corner, preparing for a large scale fight.After the hane and counter hane in the top left corner,  returns the fight to the top right. Both sides bolster their positions, and then attaches on the 4th line to the top center stone.  extends to the left, and the fighting continues. After it looks like Elf has left us in the middle of things again. However, gets 3.5k rollouts. If there is another play in this variation, there is a drop of at least 2k rollouts for it. To me that suggests that there are a number of reasonable options, and White might well tenuki.Once again the jump attachment is playable while the block is a minor error. That is not the only pattern, OC. But so far, either the jump attachment or the block has been a small mistake. |
|
| Author: | Bill Spight [ Tue May 12, 2020 5:58 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 6. Yashiro Kumiko, 6 dan (W) vs. Takemiya Masaaki, 9 dan GoGoD 2016-02-94a Elf does not rate either the jump attachment or the block as a mistake. It does prefer the keima at a to the jump attachment, and the nobi to the block, but only by 3% each. Elf does prefer another play to but only by 2½%. It's a bit of a surprise that Takemiya did not play it, because it's a Takemiya style play. How did Takemiya miss the central keima, ? (88.1k rollouts.) After ;b21: extends on the 4th line to make a center facing group. Then White plays the jump in the bottom left. Next, Black extends with as a sacrifice. accepts efficiently. Then plays to make sabaki in the bottom left corner.White replies strongly with a 2 step hane and then atari. Black continues lightly, approaching with  and leaving a ko behind.Finally we have a tsukeosae joseki that seems to be playable.
|
|
| Author: | Bill Spight [ Tue May 12, 2020 7:44 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 7. Yi Tonghun, 5 dan (W) vs. Ida Atsushi, 8 dan GoGoD 2015-08-25b In this case Elf prefers the kick at to , but only by 3%. It also prefers the nobi at 14 to , but only by 1½%. The attach and block joseki appears to be playable. is creative, but loses 6½% to par, a minor error.Here is Elf's mainline variation for .Any resemblance of White's strategy here to Black's strategy in the previous example is not coinicidental. and form a central group. is a now normal shoulder blow. is a sacrifice, and is efficient. threatens to atari at 23. Finally,  makes shape.This is an even better example of the attach and block joseki being playable. |
|
| Author: | Bill Spight [ Tue May 12, 2020 8:27 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 8. Anzai Nobuaki, 6 dan (W) vs. Ida Atsushi, 4 dan GoGoD 2012-09-17h Each play in this actual game sequence is playable. loses 3% to par, loses 2½% to par, and loses 1% to par, according to Elf. This is hardly surprising, as the only difference between this example and the previous one is that the Black stone on the right side is on the 3d line instead of the 4th line.
|
|
| Author: | Bill Spight [ Tue May 12, 2020 11:38 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 9. Takao Shinji, 9 dan (W) vs. Iyama Yuta, 9 dan GoGoD 2015-03-29c Again, this sequence from the actual game should be playable. No loss to par greater than 2½%. This was something of a surprise. I went looking for the last example, to see if there was any difference between the high Chinese and low Chinese. This occurs in the high Chinese, but I didn't go looking for it. Not that these choices are random. I chose a group of games where the attach and block occurs by move 30. There are too many factors that might affect the decision after that, I think. But from that group of games I started with games where the joseki occurred early in the game. Now I'm on games where the jump attachment occurs in the teens. That may matter. |
|
| Author: | Bill Spight [ Wed May 13, 2020 12:37 am ] |
| Post subject: | Re: Attach and block joseki |
Example 10. Zhou Junxun, 7 dan (W) vs. Cho Hanseung, 2 dan GoGoD 1998-06-16g Another playable game sequence. The jump attachment is Elf's top choice, the block loses 2½% to the nobi. BTW, the sagari, is Elf's top choice against the attach and block, and it often is.
|
|
| Author: | Bill Spight [ Wed May 13, 2020 1:07 am ] |
| Post subject: | Re: Attach and block joseki |
Example 11. Tan Xiao, 7 dan (W) vs. Ida Atsushi, 3 dan GoGoD 2012-10-18b Not much of a surprise. Another Chinese Opening position with Black strength on the top side. The bottom side is different, however. |
|
| Author: | Bill Spight [ Wed May 13, 2020 2:06 am ] |
| Post subject: | Re: Attach and block joseki |
Example 12. Luo Xihe, 5 dan (W) vs. Yu Bin, 9 dan GoGoD 1994-08-16e This sequence is not just playable, the block, , is Elf's top choice. (As are , and .) This is not a Chinese Opening position, but Black has developed on both sides. was the extension to a, but Elf has a different idea. is a sacrifice to connect underneath. White builds outside strength, and then Black plays the shoulder blow in the bottom left corner.
|
|
| Author: | Bill Spight [ Wed May 13, 2020 7:40 am ] |
| Post subject: | Re: Attach and block joseki |
Example 13. Yamada Kimio, 9 dan (W) vs. Nakano Hironari, 9 dan GoGoD 2014-01-09l The jump attachment is Elf's top choice. The block loses 5% to the nobi, a minor error.  cuts, to which is a strong extension. makes a base. Then pincers the cutting stone. plays the 4th line attachment, and  pulls back. Then plays the keima.  is a wedge on the right side. Black takes kikashi in the top right and then approaches from the top, making a territorial framework.  makes a base, threatening Black's thin corner position.At last another example where the block loses enough percentage points to the nobi to be considered a minor error. The nobi enables White to build a moyo on the top side. |
|
| Author: | Uberdude [ Wed May 13, 2020 10:05 pm ] |
| Post subject: | Re: Attach and block joseki |
John Fairbairn wrote: Quote: How bad is the attach and block joseki? It has always made me feel uneasy, because it accepts a "bad" shape. There is an interesting example in this case. He starts off by saying that the reason tsukeosae disappeared was, in a nutshell, because of the empty triangle. However, he also gives another way of looking at it (which is more attuned to AI insights, perhaps). And that is that the tsukeosae side should feel uncomfortable about giving the opponent the atari that creates the empty triangle. Uberdude wrote: one of the lessons I've picked up from bots is how much they dislike letting the opponent make gains in local efficiency of shapes in sente*, even if there is a directional benefit to it.
|
|
| Author: | Bill Spight [ Thu May 14, 2020 12:05 pm ] |
| Post subject: | Re: Attach and block joseki |
Example 14. Guo Yuzheng, 3 dan (W) vs. Chang Hao, 9 dan GoGoD 2013-05-09i Again, not too surprising, as Black only has something on the right side, while White has something on the top. As usual, Elf recommends the kick and the nobi. In the old days the kick would have been questionable, as it strengthens White without any obvious attack. makes only a 2 space high extension. plays on the frontier of two frameworks. Then plays a submarine invasion, aiming both ways. blocks off the side, and occupies the 3-3. The descent, , aims both ways. - secures the corner group with gote. Then  leans on White group in the top left with a shoulder blow.  slides to the second line, but  makes another shoulder blow on the 3d line, keeping the White groups separated. After White pushes up with ,  attacks the White group in the top right side with another shoulder blow. Note that , , and form a dogleg keima shape.After the nobi, and are usual. Then plays to stabilize the White group with strengethening Black too much. Still, Black gets the turn, , with sente and then plays the double keima with . Then plays a shallow reduction on the right side.Follow-up to the attach and block cuts in sente, as extends. In the skirmish pushes, blocks, and cuts. makes eyeshape. Then jumps into the center. turns on the 2d line, threatening the corner. grabs the  stone, securing the Black stones. Then fortifies White's top side.
|
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|