Attach extend joseki

Knotwilg · #1

Well, one of the great redemptions in the AI revolution is the attach-extend-joseki.

https://senseis.xmp.net/?44PointLowApproachAttach

Joseki

Click Here To Show Diagram Code: [go]$$W Joseki $$ -------------------- $$ | . . . . . . . . . . . $$ | . . . . . . . . . . . $$ | . . . 6 5 1 . . . 7 . $$ | . . . X . 2 3 . . . . $$ | . . . . . 4 . . . . . $$ | . . . 8 . . . . . . . $$ | . . . . . . . . . . . $$ | . . . . . . . . . . . $$ | . . . . . . . . . . .[/go]

This pattern has repeatedly been exposed as a "beginner joseki" and even today teachers (like dwyrin) warn against playing it as Black.
I believe the reasons for dismissing this pattern are: Black loses sente, in his own corner, Black is overconcentrated, and White gets an easy base.

Being a docile amateur, I hardly ever played it, except when building a framework which makes a leaning move like :b2:

obviously good. But I can now safely admit that it never felt bad for Black. And as I understand it, the bots say it's good for Black. And they go on saying that White better stretched at :w5:

directly.

All of this is briefly explained in this video: https://www.youtube.com/watch?v=PZ1DfWnki7M, continuing with many new variations and modern insights into this corner pattern.

Uberdude had already pointed out some new insights on some of the SL variations, but SL largely keeps telling the traditional story.

Uberdude · #2

Yeah attach and extend is back, and I always was dubious of the rationale for it being bad because white 5-6 is a pretty terrible exchange giving a 4th line solid corner when you do the classic tewari vs "approach, one point jump, 2 space extend, then attach". So it's very interesting that the standard bot reply after 4 extend is not 5 but 2nd line kosumi. Also I don't think you'll see bots recommending 8 at the end there (or 7). Black doesn't need to defend the cut because white didn't defend hers. If 7 is closer then black may spend a move, but 1 line lower (or even further) is often more efficient.

On a similar note I always felt comfortable in the 4-4 double low approach joseki when you attach out and they hane and extend: you got a tenuki and you still get a live group with the corner points and yes they get groups both sides but both have problems. Bots often like a at b now though, that's not one I particularly understand yet but I do use it sometimes. The bot move is usually 8 at 9 instead (unless black has ladder on the 6 stone).

Click Here To Show Diagram Code: [go]$$B $$ +---------------- $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . 9 8 2 . . $$ | . . . X . 5 6 . $$ | . b a . . 7 . . $$ | . . 4 . . . . . $$ | . . . 0 . . . . $$ | . . . . . . . .[/go]

Gomoto · #3

Cho U and AI agree on the attachment in this game.

Click Here To Show Diagram Code: [go]$$B $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . O X . . . . . . . . . . . . . . | $$ | . . O X . . . . . , . . . . . X . . . | $$ | . . . . X . . . . . . . . . . . . . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . X O . . . . . , . . . B . X . . . | $$ | . . X O . X X . O . . . . O . . . . . | $$ | . . X X O O O . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Shibano Toramaru, 8d - Cho U, 9d, W+R (Komi 6.5)
44th Meijin, Final, Game 3, 2019-09-17,18

Gomoto · #4

reaping the fruit :study:

one bad and one good attachment

https://lifein19x19.com/viewtopic.php?p=251630#p251630

Gomoto · #5

forget about :b8:

in the original post ;-)

Gomoto · #6

Today I encountered an interesting variation:

Click Here To Show Diagram Code: [go]$$W $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . O . . . . . | $$ | . . O , . O . . . , . . O X . X . . . | $$ | . . X X . . . . . . . . 1 X . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Gomoto · #7

Click Here To Show Diagram Code: [go]$$W $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . 3 O 2 . . . . | $$ | . . O , . O . . . , . . O X . X . . . | $$ | . . X X . . . . . . . . 1 X . . . . . | $$ | . . . . . . . . . . . . 4 a . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

4 at a is also possible

Gomoto · #8

Where is the good move for black? (Probably better to not follow your first instinct here.)

Click Here To Show Diagram Code: [go]$$W $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . 3 O 2 . . . . | $$ | . . O , . O . . . , . . O X . X . . . | $$ | . . X X . . . . . . . . 1 X . 8 . . . | $$ | . . . . . . . . . . . 6 4 5 7 9 . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Gomoto · #9

Gomoto · **#10**

And some variations:

ez4u · **#11**

A couple of alternatives if we run katago for 40K visits:
After move 15

After move 18

Gomoto · **#12**

added ez4u move suggestions:

sorin · **#13**

Knotwilg wrote:

Well, one of the great redemptions in the AI revolution is the attach-extend-joseki.

I blame the early go books available in English (or maybe it was just my misunderstanding of them, but then it looks like many other Western go players are in the same boat) on the misconception that attach-and-extend ("tsuke-nobi") is not suitable for "high level players".

Obviously, pros don't suffer as much from such misconceptions: I think pros' rate of playing this in pre-AlphaGo era is higher than that of Western amateur dan players :-)

And then, the rate goes up in post-AlphaGo era.

Uberdude · **#14**

sorin wrote:

Knotwilg wrote:

Well, one of the great redemptions in the AI revolution is the attach-extend-joseki.

I blame the early go books available in English (or maybe it was just my misunderstanding of them, but then it looks like many other Western go players are in the same boat) on the misconception that attach-and-extend ("tsuke-nobi") is not suitable for "high level players".

Obviously, pros don't suffer as much from such misconceptions: I think pros' rate of playing this in pre-AlphaGo era is higher than that of Western amateur dan players :-)

And then, the rate goes up in post-AlphaGo era.

Here's a thread from before AI where I repeat a lesson from Guo Juan 5p, cautioning people not to play attach extend, at least until they are 5 dan in which case you are strong enough to identify the rare cases it is good. https://lifein19x19.com/viewtopic.php?f ... an#p168059. Post-AG that advice is wrong in that it's not so rare to be a good move. What changed, new sequences, new evaluations? The principle of beware of attaching to your opponent's weak stones is still valid, but I suppose there's a change in emphasis that attaching to settle shapes early is ok if you can cause inefficiency/overconcentration and attach extend can naturally consolidate the corner territory/eyes which we now realize was playable to take directly. Also was the approach really weak? Maybe we are attaching to a strong stone which is ok! Actually I think that aspect of when it's good persists from pre- to post-AI. (Personally I also quite liked attach and crosscut as an overconcentration technique when it's already strong and boring direction, but I've not seen that from AI, but then again maybe the extend instead of hane is to prevent that, I'll check to see if LZ agrees with times I've played it). At least Guo's advice which I repeated is qualified as "until you are strong" rather than "this is always bad" but that bar of needing to be an advanced player to not get told off for playing it has come crashing down. Now I'd say pretty much anyone (e.g 15 kyus or 5 kyus) can use it because it'll never be that terrible of a move, though the wrong follow ups might be.

John Fairbairn · **#15**

It's over 50 years ago since I started reading English go books, and it wasn't too long after that that I stopped reading the beginners' books, even if I still bought them. So my memory can't be trusted. But I don't ever remember anything that said tsukenobi was bad.

What I do remember, and have just partially checked, is that tsukenobi itself was not question. It was the knee-jerk ikken-tobi connection that was questioned.

Takagawa, for example (in "Vital Points of Go"), says absolutely nothing about tsukenobi being bad but says of the sequence up to that connection: "It seems there are many people who consider the sequence from White 1 to Black 8 [not numbered here] to be just the typical tsukenobi joseki and who have not the slightest suspicion of it." He then goes on to say, "But the new feeling concerning this is that the reinforcing play Black 8 O-4 shows signs of some duplication of effort."

He further goes on: "Needless to say, the correct play for Black is at the strategic point [K3] on the lower side, which he should occupy without hesitation."

Later on he goes on in similar vein:

"The reinforcing play Black 8 [triangled] following White 7 [square] is often encountered, but it betrays Black's undigested knowledge of joseki." He then points out that Black ought to have simply jumped to D6 without the tsukenobi, so that he could attack White from around K3. This attacking aspect is also what I remember from the many other references to tsukenobi in Go Monthly and so on. The advice was that Black [the context was nearly always handicap play] should eschew solid defence and use his handicap stones to learn to attack.

Takagawa does also show examples where even the ikken-tobi connection is acceptable, which is in cases where he has an attack available anyway. And there's no excuse for overlooking all of this. It dominates the first part of his Chapter 2.

A case of misrepresenting the books, to add to the many cases of L19 misrepresenting the pros?

However, maybe even this book (or its translator) misrepresented the pros. I have never seen anything that justifies the reference to a "new feeling". The great joseki books of the time did show the ikken-tobi connection, but only in the case where White had made a very tight tiger's mouth connection himself first. Yet even then it's not called joseki (e.g. Kitani, working fron the Great Joseki Dictionary of the 1930s) simply says "Black is solid". Otherwise he tells us exactly what Takagawa said, about opting for K3.

Allen · **#16**

is responding to the attachment by extending OK?

Click Here To Show Diagram Code: [go]$$W $$ +---------------- $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . 3 1 . . $$ | . . . X . 2 . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . .[/go]

Bill Spight · **#17**

When I was learning go there was a certain disdain among amateur dans, which I picked up, for the attach and extend joseki. This was not because of the later one space jump, which most people had learned not to do, but because a lot of kyu players always played it, without reflection. The most popular dan response to the keima approach was the simple one space jump, which, truth to say, was also usually played without reflection.

As for the "new feeling" that Takagawa talked about, perhaps it had to do with the fact that, according to magazine articles that I read, one space jump attachments were becoming less popular in joseki, because the strength gained from them was not as good as the strength the opponent gained from the hane in response. Most of these cases, as I recall, were underneath jump attachments on the third line, which were being replaced by the slide to the second line.

Bill Spight · **#18**

Allen wrote:

is responding to the attachment by extending OK?

Click Here To Show Diagram Code: [go]$$W $$ +---------------- $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . 3 1 . . $$ | . . . X . 2 . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . .[/go]

Yes, when Black already has an extension on the side, maybe otherwise.

Uberdude · **#19**

Allen wrote:

is responding to the attachment by extending OK?

Click Here To Show Diagram Code: [go]$$W $$ +---------------- $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . 3 1 . . $$ | . . . X . 2 . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . . $$ | . . . . . . . .[/go]

Yes, it's definetly on the list of candidate moves to consider, indeed bots do fairly often think it is BETTER than the hane. Here's a video from Yeonwoo about it.

https://youtu.be/PZ1DfWnki7M

xela · **#20**

That's the same video that started this whole thread, right? (Scroll to the top.)

Attach extend joseki

Who is online