It is currently Thu Mar 28, 2024 8:08 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 48 posts ]  Go to page Previous  1, 2, 3  Next
Author Message
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #21 Posted: Fri May 07, 2021 6:44 am 
Oza

Posts: 2356
Location: Ireland
Liked others: 662
Was liked: 442
Universal go server handle: Boidhre
Knotwilg wrote:
I equally object to Jose being portrayed as someone who knows it all and is teaching Davy to walk before he can run. That's the old paradigm of teaching Go, which I'm arguing AI should have humbled us out of. Jose is better at running than Davy but perhaps not because of his deep understanding of how the knee works.


I'm more getting at that for a beginner arguing about 3-3 or 4-4 on that board is trying to run before learning to walk. They've far bigger low hanging fruit to tackle before worrying about fraction of a point differences in options for :w2: and they're probably not equipped to understand the difference between favouring the corner or side with 3-3 or 4-4 for white yet if they're having this fuseki shown to them. It's not even that one is worse, it's that these lead to different midgames and you may prefer one kind over the other but this isn't really something beginners need to be considering yet. It is not that Jose knows all and Davy cannot grasp it, it's that it's really not relevant information to Davy yet.

My main criticism is Jose calling 3-3 slow rather than saying "that's an option, try it and see if you like it." Even pre-AI, D4 was not prescribed here especially to kyu players.


This post by Boidhre was liked by 2 people: gennan, schrody
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #22 Posted: Fri May 07, 2021 6:45 am 
Oza
User avatar

Posts: 2408
Location: Ghent, Belgium
Liked others: 359
Was liked: 1019
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
gennan wrote:
Another example where it is not very urgent to play locally.
Black a is a good local move. It gives white a secure group with a decent amount of territory, but black settles his group while making some influence.
White a is a good local move, but it may be a bit slow globally, because black can easily ignore it and tenuki (black is satisfied with preventing a shimari in sente).
So tenuki is an option for either color in many cases and the position can be statically evaluated as about even:
Click Here To Show Diagram Code
[go]$$ whose turn is it?
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . a . . . . .
$$ | . . O . . . . . .
$$ | . . . . X . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ -------------------[/go]


We know from Uberdude's great articles on AI opening gospel that this is the most urgent situation of two stones played in a corner, after the corner invasion and before the approach.

In your post you bring two things in the tempo concept: 1) is it urgent to play here with respect to the rest of the board (which is the temperature concept) and 2) whose turn is it (the sente concept)?

In the evaluation of a corner pattern both aspects play a role of course. Invading a 4-4 is not equivalent with shoulder hitting a 3-3: the difference is exactly sente.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #23 Posted: Fri May 07, 2021 6:53 am 
Oza
User avatar

Posts: 2408
Location: Ghent, Belgium
Liked others: 359
Was liked: 1019
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
Boidhre wrote:
My main criticism is Jose calling 3-3 slow rather than saying "that's an option, try it and see if you like it." Even pre-AI, D4 was not prescribed here especially to kyu players.


Agreed. So maybe that was a strawman.

Quote:
I'm more getting at that for a beginner arguing about 3-3 or 4-4 on that board is trying to run before learning to walk. They've far bigger low hanging fruit to tackle before worrying about fraction of a point differences in options for :w2: and they're probably not equipped to understand the difference between favouring the corner or side with 3-3 or 4-4 for white yet if they're having this fuseki shown to them. It's not even that one is worse, it's that these lead to different midgames and you may prefer one kind over the other but this isn't really something beginners need to be considering yet. It is not that Jose knows all and Davy cannot grasp it, it's that it's really not relevant information to Davy yet.


And this is where I disagree. I'm 2 dan and unable to carry a favorable opening into a dominant middle game. As a 2d, showing joseki and fuseki to a beginner would have me pretend I understand things, while I don't. At any point in time they can choose exposing themselves to openings and corner patterns, and when pressed, I will provide examples of such patterns. But I won't teach them. In fact I might rather talk about the inherent differece between 3rd and 4th line (3-3 vs 4-4).

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #24 Posted: Fri May 07, 2021 7:54 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
Knotwilg wrote:
Here's a conversation between Jose Kiteacher, an amateur dan teaching at a club and Davy L'Sadvocate, a beginner with a sharp critical mind.

Jose: "The most common opening move is on the star point. Black plays here, on 4-4".
Davy: "What about playing deeper in the corner: on that ... 3-3 point"


That's also playable, as is the 3-4 point. :)

Quote:
Jose: "That's considered a little slow."


If I were to comment, which I wouldn't, I wouldn't think that a beginner would understand what I meant by slow. It's also not my preference for describing the play, but it is one way of looking at it.

Quote:
(Davy's critical mind is activated, taking no BS from anyone)
Jose: "The most common next move in that corner is White's 3-3 invasion"
(Jose shows modern joseki)


Which modern joseki? My guess is this one.

Click Here To Show Diagram Code
[go]$$ Modern joseki
$$ ----------------
$$ | . . . . . . . .
$$ | . . . . . 8 . .
$$ | . . 2 4 6 5 . .
$$ | . . 3 1 7 . . .
$$ | . . . . . . . .
$$ | . . . . . . . .[/go]


Quote:
Davy: "So why is that a good move for White?"
Jose: "Because White gets territory, and Black influence, which is a little harder to play with. And White can do that and be the first to play in another area of the board."


Well, I guessed wrong. :-| Black can certainly take sente here, and usually should. I would not show anything longer to a beginner.

My own preference would not be to talk about territory vs. influence, but about making a stable group, which I think we can say applies to both sides. :)

Quote:
Davy: "So if Black plays on 3-3 to start with, White can't go there anymore, to get this good result?"
Jose: "Err ... yes."


Moi: Good point. :) But the 3-3 is less flexible than the 4-4, more committal. Not that it's bad, in general. :)

Quote:
Davy: "So what happens if Black plays 3-3? What's White's answer?"
Jose: "Eh ... White can shoulder hit, for example" (shows traditional joseki)


i wouldn't show the shoulder hit, but say that the two space high approach is the most usual play, with the two space high response being the usual reply. Then I would point out that which side to approach from depends upon the whole board, and White will usually wait to make that decision. I might also point out that, given the approach and reply, the 3-3 stone makes solid territory, but might be considered a bit unenterprising.

I assume that this is the traditional joseki.

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


Quote:
Davy: "How can that be good? Black can ignore it and play in another area of the board, and still have that good result you just showed me?"
Jose: "Eh ..."


Guessed wrong, again. :-| Maybe Jose stopped after :w6:. In that case, my reply:

Black could play elsewhere, but :w6: is usually a mistake. If Black plays the keima, :b7:, Black not only secures territory, but aims at attacking the weak. floating White group. Better for White to play this turn for :w6:.

Click Here To Show Diagram Code
[go]$$ Variation
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 1 3 6 . . . . . .
$$ | . . . 2 4 . . . 8 , .
$$ | . 5 . . . . . . . . .
$$ | . . 7 . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


:b7: looks inefficient and slow, but if Black plays there, White will be forced into a low position with not much territory. :w8: makes a nice group.

:b5: is usually not good. Better just to turn, as a rule.

Click Here To Show Diagram Code
[go]$$ Modern joseki
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 1 3 . 7 . . . . .
$$ | . . 5 2 4 . . . . , .
$$ | . . 8 . . . . . . . .
$$ | . . . 6 . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


Each side has made a stable group. :)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #25 Posted: Fri May 07, 2021 7:58 am 
Oza

Posts: 2356
Location: Ireland
Liked others: 662
Was liked: 442
Universal go server handle: Boidhre
Knotwilg wrote:
And this is where I disagree. I'm 2 dan and unable to carry a favorable opening into a dominant middle game. As a 2d, showing joseki and fuseki to a beginner would have me pretend I understand things, while I don't. At any point in time they can choose exposing themselves to openings and corner patterns, and when pressed, I will provide examples of such patterns. But I won't teach them. In fact I might rather talk about the inherent differece between 3rd and 4th line (3-3 vs 4-4).


Yes, I'm not trying to imply that a 2 dan can. Or the teacher can ever answer the question asked. It's more about the need to gently guide people away from the the questions they don't need to be worrying overly about and towards the ones that are more relevant to them. It's not the teacher withholding knowledge but bumping the student away from things that'll waste their time overly. Like I said, I think the problem is this is all being induced by the 3-3 being called slow begging the student to go down a line that invites them to run even if the teacher is themselves only barely managing a brisk walk.


This post by Boidhre was liked by 2 people: gennan, schrody
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #26 Posted: Fri May 07, 2021 8:07 am 
Lives in sente

Posts: 757
Liked others: 114
Was liked: 916
Rank: maybe 2d
Knotwilg wrote:
Opening 2 is as if White played opening 1 as Black, then flipped stone color, except for one stone, took komi and kept sente. Flipping one stone = 2 moves. Komi and sente are one move, so the net result is equal.


Not quite equal. Obviously if you did an analogous operation in a position where somehow sente was worth literally nothing, it would be unequal and the side that gained sente at the cost of moves on the board with real value would be sad. And on the flip side, if you did it in the middle of a hot tactical situation as gennan showed, or a big midgame fight, the side that gained sente in exchange for other things of fixed value equalling that of a non-urgent big opening move, would be happy.

Your instinct (not just from book learning, but also unconsciously gained from fighting experience) should be shouting at you that the shoulder hit / 3-3 situation is a moment where play is more urgent than normal, i.e. the value of sente is higher. To the degree you trust that instinct (and ultimately, you have to trust *something*), then you can also guess that flipping the colors at this point as you did is probably not an equal trade, even if the difference is pretty small.

Knotwilg wrote:
For me it is very hard, from these diagrams, to argue that opening two has two small mistakes, which cancel each other out, rather than being two good moves. I know that :b5: in this diagram is expected to be a 3-3 in the other white corner. But why, I can't tell.


Right, you can't argue it from the diagrams alone. You need the additional information that's not part of the diagram, which is your instinct telling you (agreeing with the vast majority of other experienced players' instincts, and for good reason) that this is a time where moves are a bit more urgent than normal, and therefore swapping sides at the trade of a normal move's worth in value might not be a fair swap. And of course you can't always conclude exactly which move or moves in which diagram were mistakes from things like this, only that the final results are unfair. But knowing that gives you at least a diffuse belief over which moves might be dispreferred, relatively speaking, between the two diagrams, and you can let that be one more fuzzy data point for building your intuition, and move on.


This post by lightvector was liked by 2 people: gennan, Knotwilg
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #27 Posted: Fri May 07, 2021 9:51 am 
Dies in gote

Posts: 39
Location: Slovenia
Liked others: 36
Was liked: 9
Rank: EGF 1d
Universal go server handle: schrody
Online playing schedule: usually Sat & Sun afternoon CET
Knotwilg wrote:
Here's a conversation between Jose Kiteacher, an amateur dan teaching at a club and Davy L'Sadvocate, a beginner with a sharp critical mind.

Jose: "The most common opening move is on the star point. Black plays here, on 4-4".
Davy: "What about playing deeper in the corner: on that ... 3-3 point"
Jose: "That's considered a little slow."
(Davy's critical mind is activated, taking no BS from anyone)
Jose: "The most common next move in that corner is White's 3-3 invasion"
(Jose shows modern joseki)
Davy: "So why is that a good move for White?"
Jose: "Because White gets territory, and Black influence, which is a little harder to play with. And White can do that and be the first to play in another area of the board."
Davy: "So if Black plays on 3-3 to start with, White can't go there anymore, to get this good result?"
Jose: "Err ... yes."
Davy: "So what happens if Black plays 3-3? What's White's answer?"
Jose: "Eh ... White can shoulder hit, for example" (shows traditional joseki)
Davy: "How can that be good? Black can ignore it and play in another area of the board, and still have that good result you just showed me?"
Jose: "Eh ..."


Here's another vote for gently guiding Davy away from that precipice. We can argue about how, when and if we should teach joseki to beginners but I don't see any use in pontificating on the minute differences between various more or less common corner moves until ... well, I don't think any of us will ever be strong enough for this to become much more than a philosophical discussion.

I'd also advise Jose to add this very useful phrase to his teaching arsenal: "I don't know." Admitting that you're a bit clueless about something is often better than trying to conjure up an answer out of thin air.

Knotwilg wrote:
We can't confidently teach traditional or modern joseki to beginners. We hardly understand today's basic patterns ourselves, and since they've been evaluated as superior to old patterns, we didn't really understand those either.

Knotwilg wrote:
And this is where I disagree. I'm 2 dan and unable to carry a favorable opening into a dominant middle game. As a 2d, showing joseki and fuseki to a beginner would have me pretend I understand things, while I don't. At any point in time they can choose exposing themselves to openings and corner patterns, and when pressed, I will provide examples of such patterns. But I won't teach them. In fact I might rather talk about the inherent differece between 3rd and 4th line (3-3 vs 4-4).

If we only ever taught things we fully understand then nothing would ever get taught. After all, the AI revolution in go has also shown that even top professional players were clueless and even worse: wrong, about many aspects of the game.


This post by schrody was liked by 2 people: Bill Spight, gennan
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #28 Posted: Fri May 07, 2021 10:36 am 
Oza
User avatar

Posts: 2408
Location: Ghent, Belgium
Liked others: 359
Was liked: 1019
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
lightvector wrote:
Knotwilg wrote:
Opening 2 is as if White played opening 1 as Black, then flipped stone color, except for one stone, took komi and kept sente. Flipping one stone = 2 moves. Komi and sente are one move, so the net result is equal.


Not quite equal. Obviously if you did an analogous operation in a position where somehow sente was worth literally nothing, it would be unequal and the side that gained sente at the cost of moves on the board with real value would be sad. And on the flip side, if you did it in the middle of a hot tactical situation as gennan showed, or a big midgame fight, the side that gained sente in exchange for other things of fixed value equalling that of a non-urgent big opening move, would be happy.

Your instinct (not just from book learning, but also unconsciously gained from fighting experience) should be shouting at you that the shoulder hit / 3-3 situation is a moment where play is more urgent than normal, i.e. the value of sente is higher. To the degree you trust that instinct (and ultimately, you have to trust *something*), then you can also guess that flipping the colors at this point as you did is probably not an equal trade, even if the difference is pretty small.

Knotwilg wrote:
For me it is very hard, from these diagrams, to argue that opening two has two small mistakes, which cancel each other out, rather than being two good moves. I know that :b5: in this diagram is expected to be a 3-3 in the other white corner. But why, I can't tell.


Right, you can't argue it from the diagrams alone. You need the additional information that's not part of the diagram, which is your instinct telling you (agreeing with the vast majority of other experienced players' instincts, and for good reason) that this is a time where moves are a bit more urgent than normal, and therefore swapping sides at the trade of a normal move's worth in value might not be a fair swap. And of course you can't always conclude exactly which move or moves in which diagram were mistakes from things like this, only that the final results are unfair. But knowing that gives you at least a diffuse belief over which moves might be dispreferred, relatively speaking, between the two diagrams, and you can let that be one more fuzzy data point for building your intuition, and move on.


This is a very profound and satisfying answer to my quibbles with the topic. Worthy of study and repetition. Thanks.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #29 Posted: Fri May 07, 2021 10:43 am 
Oza
User avatar

Posts: 2408
Location: Ghent, Belgium
Liked others: 359
Was liked: 1019
Rank: KGS 2d OGS 1d Fox 4d
KGS: Artevelde
OGS: Knotwilg
Online playing schedule: UTC 18:00 - 22:00
schrody wrote:
If we only ever taught things we fully understand then nothing would ever get taught. After all, the AI revolution in go has also shown that even top professional players were clueless and even worse: wrong, about many aspects of the game.


It depends on how you define teaching and what your objective is.

If teaching is exposure to a subject by an experienced person, things can hardly go wrong, even if the expert doesn't full grasp the subject matter themselves.

If teaching is distinguishing right from wrong, or worse, not showing things because the student is incapable, while you are incapable yourself, I find more fault with it.

You are exaggerating my point. I'm talking about the opening and joseki, which I think are not only unsuitable to teach in that second manner because the student is unready, but moreover the master is unready on many occasions. You can show things, whether they are modern or traditional, giving some background, or some arguments, but please don't teach the traditional patterns because they are "easier to understand" for a beginner. At best they are giving the teacher some comfort because they have been believing these were "true" for decades.

This doesn't apply to all aspects of Go. I can confidently teach how a capturing race unfolds, differentiating between eye vs no eye etc. I can show the vital point of a bulky five. I can explain it's valuable to know the status of an L-group. Ain't no AI ever gonna prove me wrong there.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #30 Posted: Fri May 07, 2021 10:52 am 
Lives in gote

Posts: 476
Location: Netherlands
Liked others: 270
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
Knotwilg wrote:
gennan wrote:
Click Here To Show Diagram Code
[go]$$ whose turn is it?
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . a . . . . .
$$ | . . O . . . . . .
$$ | . . . . X . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ -------------------[/go]


We know from Uberdude's great articles on AI opening gospel that this is the most urgent situation of two stones played in a corner, after the corner invasion and before the approach.

Yes, black's approach was quite urgent. It has rank +2 in The Opening Gospel.
But white responding at a is less urgent, at rank +3.5.
Black pressing at a has rank +3.

Click Here To Show Diagram Code
[go]$$ whose turn is it?
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ | . . . X . . . . .
$$ | . . O . . . . . .
$$ | . . . . . . . . .
$$ | . . . . . . . . .
$$ -------------------[/go]

But in the 3-3/4-4 situation above, continueing locally has rank -2(!), making it more urgent than all of the above opening moves.
It is even more urgent than than taking an empty corner at rank +1.
The only higher ranked move (at rank -3) is responding to a black press at a.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #31 Posted: Fri May 07, 2021 1:46 pm 
Lives in gote

Posts: 476
Location: Netherlands
Liked others: 270
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
Bill Spight wrote:
Click Here To Show Diagram Code
[go]$$ Variation
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 1 3 6 . . . . . .
$$ | . . . 2 4 . . . 8 , .
$$ | . 5 . . . . . . . . .
$$ | . . 7 . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


:b7: looks inefficient and slow, but if Black plays there, White will be forced into a low position with not much territory. :w8: makes a nice group.

:b5: is usually not good. Better just to turn, as a rule.

I think AI usually recommend :b5: as the knight move, instead of the turn (the turn can be seen as a push that helps the opponent to create more influence).
With the :b5: knight move, black's group is already alive, so black can play elsewhere after :w6:.

I think :b7: is slow. If black plays elsewhere after :w6: and white then presses at :w7:, black can push once on the 2nd line and play elsewhere again. This means that white's press at :w7: would lose sente again.
Black could even push up and cut when white does not yet have a stone around :w8:. So I think :w8: is more urgent than :b7:.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #32 Posted: Fri May 07, 2021 8:40 pm 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
gennan wrote:
Bill Spight wrote:
Click Here To Show Diagram Code
[go]$$ Variation
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 1 3 6 . . . . . .
$$ | . . . 2 4 . . . 8 , .
$$ | . 5 . . . . . . . . .
$$ | . . 7 C . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


:b7: looks inefficient and slow, but if Black plays there, White will be forced into a low position with not much territory. :w8: makes a nice group.

:b5: is usually not good. Better just to turn, as a rule.

I think AI usually recommend :b5: as the knight move,


Early in the game? Into the middle game, the keima looks better. I think that it is the human play, while the turn is the new AI joseki.

Quote:
instead of the turn (the turn can be seen as a push that helps the opponent to create more influence).
With the :b5: knight move, black's group is already alive, so black can play elsewhere after :w6:.


Well, deprived of the other keima on the second line by :w6:, I think that the usual continuation is 4th line keima for :b7: (marked), not the kosumi that I suggested. I checked with Waltheri for plays early in the game in the AI era. The 4th line keima is joseki in response to the :w6: turn.

After the 4th line keima, I think that :w8: could be elsewhere, but I wouldn't teach that to a beginner.

Take a look at the Shusai vs. Go Seigen Game of the Century. Check if your AI prefers the turn to the keima. Shusai played elsewhere after the keima, but check if the turn is not preferred at that point, as well.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #33 Posted: Sat May 08, 2021 3:58 am 
Lives in gote

Posts: 476
Location: Netherlands
Liked others: 270
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
Bill Spight wrote:
Early in the game? Into the middle game, the keima looks better. I think that it is the human play, while the turn is the new AI joseki.

Quote:
instead of the turn (the turn can be seen as a push that helps the opponent to create more influence).
With the :b5: knight move, black's group is already alive, so black can play elsewhere after :w6:.


Well, deprived of the other keima on the second line by :w6:, I think that the usual continuation is 4th line keima for :b7: (marked), not the kosumi that I suggested. I checked with Waltheri for plays early in the game in the AI era. The 4th line keima is joseki in response to the :w6: turn.

After the 4th line keima, I think that :w8: could be elsewhere, but I wouldn't teach that to a beginner.

Take a look at the Shusai vs. Go Seigen Game of the Century. Check if your AI prefers the turn to the keima. Shusai played elsewhere after the keima, but check if the turn is not preferred at that point, as well.


You are right (a bit to my surprise). The slide is 0.1 points worse in this game than the turn (according to my KataGo).
Attachment:
File comment: slide or turn?
century1.png
century1.png [ 298.43 KiB | Viewed 20228 times ]


It expects the slide to be sente. Next, it prefers pincering to the 4th line keima, so in your original slide variation :w8: does seem more urgent that :b7:.
Attachment:
File comment: slide is sente, follow-up is pincer
century2.png
century2.png [ 295.09 KiB | Viewed 20228 times ]


But to my next surpise, it sometimes expects the turn to be not really sente in this game?! Black then takes :b9: later on, but subsequently ignores :w10:?!, transposing this into a variation from a 2-space high enclosure?
Attachment:
File comment: turn may be gote?
century3.png
century3.png [ 292.11 KiB | Viewed 20228 times ]

I guess that the variations from the turn are are way over my head.


Last edited by gennan on Sat May 08, 2021 5:02 am, edited 1 time in total.

This post by gennan was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #34 Posted: Sat May 08, 2021 4:09 am 
Lives in gote

Posts: 476
Location: Netherlands
Liked others: 270
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
Bill Spight wrote:
I checked with Waltheri for plays early in the game in the AI era. The 4th line keima is joseki in response to the :w6: turn.

This is what I get in Waltheri's (many of the games with this position are post-AI):
Attachment:
waltheri1.png
waltheri1.png [ 322.63 KiB | Viewed 20227 times ]

Indeed. Moving out of the corner (options 1 and 10) is played in 57.1% of cases, pincering (options 2 to 7 and 9) is played in 23.8% of cases and tenuki in 18.8% of cases.
So professionals differ with AI here, it seems.


Last edited by gennan on Sat May 08, 2021 5:20 am, edited 2 times in total.

This post by gennan was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #35 Posted: Sat May 08, 2021 4:40 am 
Lives in gote

Posts: 476
Location: Netherlands
Liked others: 270
Was liked: 147
Rank: EGF 3d
Universal go server handle: gennan
I'd like to note this related transposed "trick" variation from a position where white already has a stone on the side star point:
Click Here To Show Diagram Code
[go]$$B black's "trick" move
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . 1 . . . . . . .
$$ | . . . O . . . . . O .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


AI see this as a good result for black, although human players might argue that white choses the "correct" direction of play:
Click Here To Show Diagram Code
[go]$$B good for black
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 3 1 2 . . . . . .
$$ | . . . O 4 . . . . O .
$$ | . 5 . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


AI see that as a better result for black than this, which is black's "rightful" result:
Click Here To Show Diagram Code
[go]$$B even
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 1 2 . . . . . . .
$$ | . . 3 O . . . . . O .
$$ | . . . 4 . . . . . . .
$$ | . 5 . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


So AI usually resist by blocking on the other side with white, taking the corner territory and allowing black to make a base on the side:
Click Here To Show Diagram Code
[go]$$B good for white
$$ ----------------------
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . 2 1 3 . . 5 . . .
$$ | . . 4 O . . . . . O .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


So AI only care so much about the "correct" direction of play by human standards. AI are really greedy about secure territory.


This post by gennan was liked by: Bill Spight
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #36 Posted: Sat May 08, 2021 5:11 am 
Dies in gote

Posts: 39
Location: Slovenia
Liked others: 36
Was liked: 9
Rank: EGF 1d
Universal go server handle: schrody
Online playing schedule: usually Sat & Sun afternoon CET
Knotwilg wrote:
schrody wrote:
If we only ever taught things we fully understand then nothing would ever get taught. After all, the AI revolution in go has also shown that even top professional players were clueless and even worse: wrong, about many aspects of the game.


It depends on how you define teaching and what your objective is.

If teaching is exposure to a subject by an experienced person, things can hardly go wrong, even if the expert doesn't full grasp the subject matter themselves.

If teaching is distinguishing right from wrong, or worse, not showing things because the student is incapable, while you are incapable yourself, I find more fault with it.

You are exaggerating my point. I'm talking about the opening and joseki, which I think are not only unsuitable to teach in that second manner because the student is unready, but moreover the master is unready on many occasions. You can show things, whether they are modern or traditional, giving some background, or some arguments, but please don't teach the traditional patterns because they are "easier to understand" for a beginner. At best they are giving the teacher some comfort because they have been believing these were "true" for decades.

This doesn't apply to all aspects of Go. I can confidently teach how a capturing race unfolds, differentiating between eye vs no eye etc. I can show the vital point of a bulky five. I can explain it's valuable to know the status of an L-group. Ain't no AI ever gonna prove me wrong there.


There's a lot to unpack here.

First of all, I'd rather not see teachers reduced to just quoting factual knowledge from books. I agree that there is go knowledge that is (almost) certainly correct, such as our knowledge of basic l&d shapes. If the teacher is in possession of such knowledge and the student is ready to receive it then all is well. The problem here is that only a very small subset of go knowledge is of the factual, provable variety.

Based on this, there's two types of knowledge:

- knowledge that has been proven and is therefore correct
- knowledge that hasn't been proven (yet) and therefore may not be correct

On second thought, perhaps this matter isn't so black and white and it would be better to think of degrees of provability and correctness. I think that endgame theory is developed enough and close enough to being correct that we could allow for it to be taught without feeling too guilty about it. What about the opening and the middle game? We've just rewritten most of what we know about the opening, so should we really teach it at all?

Let's make another distinction:
- our current collective knowledge of the game
- complete knowledge of the game (i.e. the game is solved or close to being solved)

Perhaps this is what you meant by me exaggerating things. If we take the complete knowledge of the game as a reference point, then even the professional players would be limited to teaching just a select few pieces of provable information. On the other hand, taking our collective knowledge of the game as a framework would provide us with more freedom. Professional players can usually keep up with AI in the opening, so while the two of us may not be fully qualified to teach it, they most likely are. Still, they completely fall apart in the middle game, with their reading skills and positional judgement being no match for AI's. So, who are we to learn the middle game from?

Perhaps you'll still say I'm exaggerating and I certainly am. I sense that the major difference in our views is that you're less willing to teach "wrong" things than I am. I'd be interested to know where you draw the line for yourself and where you'd draw it for professional players.

In the meantime, here's why I'm not that bothered by teaching potentially wrong things.

I think that a perfect teacher is someone who has good (factual) knowledge and is able to understand each individual student's needs. Understandably, such teachers are difficult to find. One of the reasons is because someone who's only a few stones stronger will usually better understand how the weaker player thinks than someone who's 30 stones stronger. (Of course, teaching experience makes a lot of difference here.) We've pretty much already touched upon this but players of different strengths see the same board differently, e.g.:
A: Is my stone in atari? Can I save it?
B: My stone is in atari. I'll extend to save it.
C: My stone is in atari, but it's just one stone. I'll save these three stones instead.
...
D: My opponent has this weak group. If I use this move to attack it and chase it in that direction I'll get a wall which I'll be able to use to attack his other weak group. If I enclose it and get another wall, I'll get a huge center moyo and when the opponent invades... (the atari'd stone was ignored until endgame)

If the student wants to expedite their learning process then they'll want to progress through or skip through several different stages of understanding the game as fast as possible but even then, player C would be able to teach player A a fair bit about ataris and the value of stones.

The downside here is, of course, that a weaker player is also more likely to teach some bad habits. I've never had a regular teacher so I've picked up a fair amount of those over the years and am still trying to unlearn them with the help of AI. Unless the student wants to become a professional or high dan player, I don't find that all that tragic. If they do, then they really shouldn't be getting free lessons from a mediocre amateur.

To sum up:
- Teachers should always take the student's level and understanding of the game into account.
- Even professional players will likely teach you some wrong things.
- Amateur teachers therefore shouldn't feel that guilty about teaching some wrong things. Learning is a process and students will inevitably go through several wrong understandings of the game.
- Teachers should always be open and honest about the limits of their knowledge. I prefer saying "I would play here instead because..." rather than "Your move was wrong. You should play here instead."


This post by schrody was liked by: gennan
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #37 Posted: Sat May 08, 2021 5:49 am 
Oza

Posts: 3647
Liked others: 20
Was liked: 4626
Quote:
I think that a perfect teacher is someone who has good (factual) knowledge and is able to understand each individual student's needs.


Personally, I would label that level as "adequate" rather than "perfect". For me, the perfect teacher has mainly to be able to inspire the pupil to work hard. The right facts don't matter too much - if the student works hard he will discover the corrections for himself, and learn oven more from that very process.

But all this talk of go teachers baffles me anyway. I never had a go teacher and never even thought of having one. Most things I have learned have been without a teacher, or with a teacher just hazily in the background. Rather like a go pupil having a pro "teacher" who never played him and maybe rarely spoke to him. I think that was normal for my generation. And I'm not complaining, because I had teachers (and bosses) who inspired me to work hard. They were in control of HOW I worked rather than WHAT I learned. Far from complaining, I'm actually very grateful.

I find the modern obsession with having a teacher either snobbish lifestyle-ism (rather like having a personal trainer), or the offspring of a craze with cramming schools and higher education with teachers judged solely by exam results. When I compare my schooldays with the schooldays of my grandchildren, I think I had the better deal.

There are thankfully plenty of people who are using the old way successfully, of course. In go, I'm pretty sure Sumire wasn't specifically taught much - there hasn't even been the time for that. But what she has picked up is the ability to work hard on her own. Exactly the same for Go Seigen. And look where that got them.


This post by John Fairbairn was liked by 2 people: gennan, gowan
Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #38 Posted: Sat May 08, 2021 6:09 am 
Dies in gote

Posts: 39
Location: Slovenia
Liked others: 36
Was liked: 9
Rank: EGF 1d
Universal go server handle: schrody
Online playing schedule: usually Sat & Sun afternoon CET
John Fairbairn wrote:
But all this talk of go teachers baffles me anyway. I never had a go teacher and never even thought of having one. Most things I have learned have been without a teacher, or with a teacher just hazily in the background. Rather like a go pupil having a pro "teacher" who never played him and maybe rarely spoke to him. I think that was normal for my generation. And I'm not complaining, because I had teachers (and bosses) who inspired me to work hard. They were in control of HOW I worked rather than WHAT I learned. Far from complaining, I'm actually very grateful.


There were subjects and skills where I had a teacher and those where I did not. Where I had a teacher, I was more likely to stick with it, I learned more, faster and enjoyed the overall learning process more. Where I didn't have a teacher, the opposite is true but those experiences taught me how to work harder, be more independent, how to persevere and how to motivate myself when the going got tough.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #39 Posted: Sat May 08, 2021 8:27 am 
Lives in sente
User avatar

Posts: 714
Liked others: 109
Was liked: 138
Rank: Shokyu
Universal go server handle: CDavis7M
This is all interesting discussion but why not teach both old and new joseki as both are viable and both won't be remembered anyway. The student will have to play, see what happens, and learn the proper sequence from experience. But it's nice to have some guidance as to the first move and possible responses even if the joseki isn't remembered.

Back to the hypothetical discussion, starting 3-3 avoids the need to learn the AI direct 3-3 invasion and joseki. It avoids a lot of complication and any disadvantages are vastly outweighed by other plays a beginner might make. I think it's the best opening for beginners. Maybe some bias as I've opened with 3-3 a lot.

Top
 Profile  
 
Offline
 Post subject: Re: Teaching beginners oldschool or new AI joseki?
Post #40 Posted: Sat May 08, 2021 9:53 am 
Honinbo

Posts: 10905
Liked others: 3651
Was liked: 3374
John Fairbairn wrote:
Quote:
I think that a perfect teacher is someone who has good (factual) knowledge and is able to understand each individual student's needs.


Personally, I would label that level as "adequate" rather than "perfect". For me, the perfect teacher has mainly to be able to inspire the pupil to work hard.


I agree that a major task, if not the main task, of a teacher is inspiration. :) John Conway had a great knack for infecting children with the fun of mathematics.

John Fairbairn wrote:
The right facts don't matter too much - if the student works hard he will discover the corrections for himself, and learn oven more from that very process.

But all this talk of go teachers baffles me anyway. I never had a go teacher and never even thought of having one. Most things I have learned have been without a teacher, or with a teacher just hazily in the background. Rather like a go pupil having a pro "teacher" who never played him and maybe rarely spoke to him. I think that was normal for my generation.


Not everybody has your talent, John. :)

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 48 posts ]  Go to page Previous  1, 2, 3  Next

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group