Life In 19x19
http://lifein19x19.com/

Reasoning about shape
http://lifein19x19.com/viewtopic.php?f=13&t=14525
Page 1 of 1

Author:  John Fairbairn [ Sat Sep 09, 2017 2:52 am ]
Post subject:  Reasoning about shape

One of the main fascinations of amateur players of any level is with shape. There is surprisingly little good literature that deals with the reasoning behind making shape. Most discussions centre round the attributes of shapes, i.e. their static forms. But shape is also a dynamic concept (hence haengma - why we play a shape not how), and that is dealt with much less.

The following position from a 1930s pro game threw up a shape that I thought players of any level here could comment on just to show how they think about making shape.



In other words, why did White play the triangled move in answer to Black's marked slide? And from the "why" does the "how" follow? I think I know the answer but can't be sure. What I am certain of is that in the usual sort of blitz game I play nowadays just to avoid going to bed, my first thought would be to the left of the triangle - pretty symmetric patterns seem to involve less thought somehow - a good thing at my age! At least I wouldn't normally consider being forced by playing at the 3-3 point.

I think this sort of move is also very relevant in the AlphaGo age. As I have remarked several times before, much of its play was prefigured in the 1930s.

This isn't about showing how strong you are. It's about exposing thought processes, so I repeat that any level of player should feel free to comment.

Author:  EdLee [ Sat Sep 09, 2017 3:36 am ]
Post subject: 

Hi John,

A few idle thoughts (but very curious about AG's evaluation of F14 and where AG wants to play as W):

- W doesn't want to help B turn most of the top into solid cash (4th line cash seems too efficient for B), so leaning (for now), like F16, is out.

- E14, your first feeling, the V-knight, feels slow (just a feeling, no rigorous analysis to back this up).

- C17-F15 exchange seems good for B, so C17 is out.

- W wants to keep C9, H17, M17, etc. weak spots open; W is waiting.

- I'm guessing W's reply to B E15 peep starts with E16.

Author:  djhbrown [ Sat Sep 09, 2017 5:18 am ]
Post subject:  Re: Reasoning about shape

Alphago thinks about shape using a stock of 3x3 patterns. That partly explains why she likes 4th line nikken tobi, since that fits two 3x3 patterns. I can't guess whether she would like the 2-space jump here, though, as i can't read all the way to the end of the game even with the aid of a pair of 19-sided dice. Swim would like the way F14 cuts across the path between black's top and lower left and reaches out towards white's friends on the right, whilst maintaining colour connection to the D16 cluster.

Author:  gowan [ Sat Sep 09, 2017 9:29 am ]
Post subject:  Re: Reasoning about shape

My first thoughts:

The marked white stone looks toward a white invasion at H17 and also puts a little pressure on the two black stones on the left.

I recall a shape with the two space jump like thisin a game by Fujisawa Hideyuki in a Kisei match (against Otake or Rin?).

Author:  jeromie [ Sat Sep 09, 2017 9:53 am ]
Post subject:  Re: Reasoning about shape

I'll take a shot at describing my thoughts before reading the replies.

  • The two white stones can't make much territory at this point, and making small life on the side while black plays forcing moves on the outside would be unbearable. This means white needs to jump to the outside, so the right move is going to be somewhere in the vicinity of the played move or your alternate suggestion one space to the left.
  • If I'm not trying to make territory, why are these stones important? Black would have far too much territory on the left and top if he were to surround them on a large scale, so these stones can't just be abandoned. Black shapes on the top and left also seem to have some weak points. The locations marked with a cross all look like places that could be attacked if the white group becomes strong. Alternatively, white could lean on the bottom with something like the triangled move to build strength and center influence.
    Click Here To Show Diagram Code
    [go]$$
    $$ +---------------------------------------+
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . . X . . . . . . . . . . . . . . . |
    $$ | . . . . . X . M . . . . . . . . . . . |
    $$ | . . . O . . . M . X . . . X . X . . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . O . . O . . . . . . . . . . O . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . . T . . . . . . . . . . . . . . . |
    $$ | . . X . . . . . . . . . . . . . . . . |
    $$ | . . . M . . . . . . . . . . . O . . . |
    $$ | . . M M . . . . . . . . . . . . . . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . . X . . . . . . . . . . . . . . . |
    $$ | . . . . . . . . . . . . . . . . O . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . . O . . . . . . . . . X . X . . . |
    $$ | . . . . . . O . . . . O . . . . . . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ | . . . . . . . . . . . . . . . . . . . |
    $$ +---------------------------------------+[/go]
  • Since my primary aim is to gain influence and mount an attack, I don't care if I have to abandon one of the stones in the process (e.g. after E15). A move one space to the left would perhaps be easier to directly make eye space, but if these stones live and don't do anything else they are not serving their purpose well. The move that was chosen reaches further out into the center and more directly takes aim at the vulnerabilities on the top. It also makes sure there is no outside attack that will link the top and left, so at least one group should be able to be counter attacked if black makes a move against the my stones.

Author:  Kirby [ Sat Sep 09, 2017 10:47 am ]
Post subject:  Re: Reasoning about shape

I like white's move for a few reasons:
1.) It's pretty light
2.) It aims at H17 area on the top
3.) It does well to break up the connection between black's stones on the left and top

Author:  John Fairbairn [ Sun Sep 10, 2017 1:54 am ]
Post subject:  Re: Reasoning about shape

It's very interesting to me that everyone has a completely different take on this. Shape is a basic topic where I'd expect a fair degree of uniformity. Is this lack of uniformity highlighting a weakness in western go? We haven't been taught it properly?

FWIW I consider some of the replies (or parts of some of them) are still concentrating on the static attributes of the shape (with more than hint of post hoc reasoning :)) rather than its dynamic underpinnings. I would certainly regard that as a weakness worth correcting.

Having said, that I don't know the proper answer. But I have the benefit of having seen the commentary (which only alludes to this position indirectly) and the game continuation.

The commentary says that if White 18 (the triangled move is at the 3-3 point, Black "it is obvious that Black will jump to F15." I don't think anyone here picked up on that specifically. That was, I assume, also the flaw in my own instinctive, pretty-pretty sake bottle move E14. Of course Black could still jump to F15 in the game but that would strengthen White on the outside and make the White invasion at C9 even nastier, as per the standard domino theory of fuseki.

But a possibly more important point (which I missed too) is that the triangled move is dynamic. That is: a shape is never really finished - it has to keep growing and so the way you make the shape has to allow that growth. The first growth move is at F16 (see the game below). This is standard, but I still missed it. My focus was on the two stones C14-F14 (not sure why?). If I had focused on the hazama shape D16-F14 I would have easily remembered that this is a standard way to set up an L-shape of ikken tobis. But even once White gets F16 in, the shape is still dynamic and needs stones added to it - again see the game. The rather detailed commentary didn't mention any of that, the players obviously feeling it was all too simple, and so I felt a bit :oops: Which is why I brought it up here.



Note that White 52 is nice example of growing the shape while being a domino response to a move as far away as Black 51.

Author:  EdLee [ Sun Sep 10, 2017 2:33 am ]
Post subject: 

Quote:
C17-F15 exchange seems good for B, so C17 is out.
Quote:
The commentary says that if White 18 (the triangled move is at the 3-3 point, Black "it is obvious that Black will jump to F15." I don't think anyone here picked up on that specifically.
Did I misread something ? :) :scratch:

( OTOH, if indeed :w18: at C17, I wouldn't be surprised if AG doesn't play F15; again, curious to see AG's variations. )

Author:  Uberdude [ Sun Sep 10, 2017 4:49 am ]
Post subject:  Re: Reasoning about shape

Some thoughts (before looking at John's answer -- honest!).

First of all this comment surprised me:
John Fairbairn wrote:
What I am certain of is that in the usual sort of blitz game I play nowadays just to avoid going to bed, my first thought would be to the left of the triangle - pretty symmetric patterns seem to involve less thought somehow - a good thing at my age! At least I wouldn't normally consider being forced by playing at the 3-3 point.

Afaik John you are about a 3 dan, but probably the sort of 3 dan who is relatively weaker in reading but stronger in shape intuition from all the pro games you transcribe. But this (one-to-the-left) move is one I would expect a 1d to know better than to play, and the further jump is a fairly common shape that's been a part of my Go vocabulary for quite some time, for example below (white shouldn't play 2 at a due to tesuji at b (though there can be counter tesuji attachment to that) and coming out with 2 is a nice shape, aiming at both c and d). In fact I think I wrote a post about this shape here recently.
Click Here To Show Diagram Code
[go]$$B
$$ +--------------------
$$ | . . . . . . . . . .
$$ | . a 1 X . . . . . .
$$ | . . O . . X . . X .
$$ | . b . O . c . . . ,
$$ | . . . . . . . . . .
$$ | . . O . . 2 . . . .
$$ | . . . . . . . . . .
$$ | . . X . . . . . . .
$$ | . . . d . . . . . .
$$ | . . . , . . . . . ,[/go]


Anyway, back to the game. First thing I would think about is what else could black have done instead of the slide? Jump to f15 is one fairly standard idea to reinforce the top side moyo. (Lower right also looks quite urgent, I guess black just ignored m3, how severe is white's follow-up there?). This threatens to make 3-3 invasion at top left more severe, so white's solid answer is to play d18, simultaneously securing corner and creating invasion/yose opportunities on the top side. c11 isn't at c12 which puts more pressure on the corner, but c11 still stops white extending down the left side. I guess black wasn't content with this (it's not clear f15 actually gained more than d18 did) so the plan was probably if white answered at 3-3 then after f15 white doesn't have such a comfortable move i.e. c18 would be solid but clearly worse than d18 without the slide.

So white coming out makes sense, and 2-space jump is a standard haengma (I particularly use the word haengma for this sort of moving out to the centre). The elephant jump shape to the 4-4 is not a worry: if black splits you just charge through at e16 and black ends up hurting f17 if he insists on separating f14. However, black can play a solid move like g16 (taking away white's option to make shape with f16) which also aims at poking at the white thinness. White aims at black's wide extensions on the top and left sides (h17 or j16 if you have both ladders, c9 or d8 if you have both ladders -- here white doesn't have ladder north east for d8, but even if you did I wonder if black giving the ponnuki is not so bad here, or the push a 2nd time isntead of connect variation).

Black could also take the 3-3 (as now if white takes it black f15 isn't really working, though I wonder how would white answer that attachment? hane g15, extend g14, extend e14 too soft?). I'm somewhat reluctant to play f16 as white because it strengthens black where I have invasion possiblities (unlike in 2-space extension joseki diagram above), but once he has the 3-3 white's group lost its base and black gained one so probably an invasion would be overplay (black could answer h17 with g15 for example, maybe white j16 a better idea) in which case f16 becomes less sad. White does still have gaps at m17/16 or r17 so isn't giving black the whole top side.

Another idea for black would be c4: my idea would be if you get the c4 c3 c5 b3 d5 e4 exchanges in sente then white c9 invasion is harder as it can't connect up to the lower left corner. However, the d7 c11 gap is still thin and it also makes white corner big and solid. Maybe c4 c3 d5 is better suji too.

What about black total tenuki, e.g. to lower right? Assuming white gets sente, what is white's follow up? c17 probably forces some defence around g16 or else invading top is powerful now that black can't live at 3-3. Or white could invade (h17, j16 etc), splitting black moyo and forcing him into 3-3. If black hasn't played 3-3 I don't want to f16.

Author:  Bill Spight [ Sun Sep 10, 2017 9:36 am ]
Post subject:  Re: Reasoning about shape

John Fairbairn wrote:
One of the main fascinations of amateur players of any level is with shape. There is surprisingly little good literature that deals with the reasoning behind making shape. Most discussions centre round the attributes of shapes, i.e. their static forms. But shape is also a dynamic concept (hence haengma - why we play a shape not how), and that is dealt with much less.


My experience differs. A lot of strong amateurs, in both the East and the West, don't give a fig about shape. Not that weak amateurs do, either. ;) Yes, IMX most discussions by Western amateurs about shape -- I don't know about Eastern amateurs -- focus on static forms. I guess I was lucky in that when I first learned about shape it was about shape as a dynamic concept, about efficiency, not about form per se. From 2 kyu to 3 dan I was a shape player. Hardly any of my opponents were.

Quote:
It's very interesting to me that everyone has a completely different take on this. Shape is a basic topic where I'd expect a fair degree of uniformity. Is this lack of uniformity highlighting a weakness in western go? We haven't been taught it properly?


Shape may be a basic topic, but I don't think that it is an easy one. The lack of uniformity does not surprise me. Yes, I think that it highlights a weakness in Western go, and in Eastern go, as well.

Click Here To Show Diagram Code
[go]$$Wc One space jump
$$ +----------------------
$$ | . . . . . . . . . . .
$$ | . . . X . . . . . . .
$$ | . . . . . X . . X . .
$$ | . . . O . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . O . 1 . . . . . .
$$ | . . . . . . . . . . .
$$ | . . X . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . X . . . . . . . .
$$ | . . . . . . . . . . .
$$ | . . . . . . . . . . .[/go]


When I was learning go this jump was recommended, the implication being that it was good shape.

Click Here To Show Diagram Code
[go]$$Wc Too solid
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . O . . . . . X . . . X . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . 1 . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . B . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . X . X . . . |
$$ | . . . . . . O . . . . O . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]


But in this position I do not find :w1: appealing. One reason that nobody has mentioned is that :bc: is two spaces away instead of one. It does not threaten White very much, nor does :w1: threaten it very much.

From a shape standpoint I find two plays worth considering.

Click Here To Show Diagram Code
[go]$$Wc More efficient
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . O . a . . . X . . . X . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O . . 1 . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . B . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . X . X . . . |
$$ | . . . . . . O . . . . O . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]


The two space jump in the game looks more efficient, especially given the distance to :bc:. It also, as has been pointed out, aims at "a".

Click Here To Show Diagram Code
[go]$$Wc The opponent's play
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . X . . . . . . . . . . . . . |
$$ | . . . W . . . . . X . . . X . X . . . |
$$ | . . . . . 1 . . . . . . . . . . . . . |
$$ | . . W . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . X . X . . . |
$$ | . . . . . . O . . . . O . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]


Another shape play worth considering, IMO, is the keima, :w1:. Maybe the opponent's play is my play. Besides, the keima triangle that :w1: forms with the :wc: stones is a solid shape. Maybe too solid, which is why White played the two space jump in the game.

Author:  jeromie [ Sun Sep 10, 2017 1:31 pm ]
Post subject:  Re: Reasoning about shape

I actually found the responses remarkably uniform, especially given the strength differences involved. Almost everyone mentioned not wanting to be sealed in, and Ed, Kirby, gowan, and I all mentioned an invasion at H17. (Being a bit weaker, my eye on invasion points was a bit fuzzier.) Several people also mentioned black's possible attack at E15. Similarly, I think all of these answers were looking at shape in a dynamic matter. At least, they all understand that the proper relationship between my stones depends on the position of my opponent's nearby stones.

Of course, I suppose it could be debated whether these answers really focused on shape, per se, and none of us described the continuation or shapes that would develop in the game. (Uberdude did when he jumped in - a factor of being stronger, or just a different way of thinking about the position?)

Author:  Knotwilg [ Sun Sep 10, 2017 1:45 pm ]
Post subject:  Re: Reasoning about shape

First a word on terms: I understand we are trying to translate words like suji or haengma and when finding a noun, sometimes lose the dynamic meaning of the original word. Nevertheless, one cannot blame people to adopt a static concept when learning about "shape". If we want people to adopt a dynamic concept, we should use continuous present, like "making shape" or otherwise nouns that hold a dynamic concept, like "flow". I think I remember "haengma" means "the way of the moving horse" but we don't usually take such a lyrical approach in game theory - maybe unfortunately so. In short, while you can criticize a static interpretation of "shape", we could use dynamic words to convey concepts that should have a dynamic aspect.

Myself I have crossed that bridge in the realm of articulation but still, like John said, I spotted the elephant's eye first and thought about what would happen if Black played there (not that I thought he would, but just to understand the shape).

I didn't think of developing the shape into the trapeze next.

It's also interesting to note that White's pressing move does not look very appealing for Black to play himself. This is another aspect of shape: it's not about your own stones only.

But I wonder why Black takes the corner and allows White to force him to take territory, while strengthening his "shape" and invade next.

Why doesn't Black forestall the invasion instead? How is White going to use his move to harm the top?

Otherwise stated: has White "made miai"?

Author:  djhbrown [ Fri Sep 22, 2017 5:47 pm ]
Post subject:  Re: Reasoning about shape

you can tie yourself in knots over semantics and/or reading and/or circular reasoning, or you can use your commonsense, or if you don't have any of that, you can use Swim's commonsense:

q: when is a shape not a shape?
a: when it's not colour-connected* (then it's 2 shapes)

q: when is a move a "make shape"?
a: when it makes a shape (cf prev answer)

qed

* https://github.com/pnprog/gomap

Author:  John Fairbairn [ Sat Sep 23, 2017 2:37 am ]
Post subject:  Re: Reasoning about shape

Quote:
you can tie yourself in knots over semantics and/or reading and/or circular reasoning, or you can use your commonsense, or if you don't have any of that, you can use Swim's commonsense:

q: when is a shape not a shape?
a: when it's not colour-connected* (then it's 2 shapes)

q: when is a move a "make shape"?
a: when it makes a shape (cf prev answer)

qed


Swim can happily flounder in its own paddling pool but human common sense must not ignore the oceans of the real world.

Making shape is ultimately a translation from Japanese (which in turn takes it from Chinese). Katachi is one of those very many words which have connotations according to context. You can say 形が悪い (the shape is bad) but without qualification 形 is usually taken to mean いい形 (good shape). Not just in go - you can see the same implication in the military context of the Art of War.

So four stones in a dense square shape and not being a nakade shape is a qed Swmm shape but is usually not considered a good shape by humans and so we would never say "make shape" about it (though we might say make a shape).

I'm too lazy to think about an ideal example from ordinary English, but maybe a passable one is "handsome". "Handsome man" is usually a term of praise. A "handsome woman" can be a term of praise or a veiled way of saying "not beautiful". Human common sense tells us context is all.

Author:  djhbrown [ Sat Sep 23, 2017 4:34 am ]
Post subject:  Re: Reasoning about shape

John Fairbairn wrote:
I'm too lazy to think... context is all.
yes, shape is not a unitary attribute - as you say, there is good shape and bad shape.

so now, the only question that remains, is: what makes a shape good?

if you temporarily suspend disbelief and go along with Swim's definition of a shape as a colour-connected cluster, it follows immediately that the goodness of a shape is its lightness, ie the ratio of the number of colour-controlled empty points within its colour map perimeter to the number of stones making it, as defined in:

Seeing is Believing (Or How To Make Sabaki) SIGART Newsletter 80, 91-94. Reprinted in M.Bramer (Ed) Computer Game Playing: Theory and Practice, 177-184. Horwood, 1983.

That was 34 years ago; things have moved on since then - today's Swimmer doesn't bother about measuring dangosity ratios, because its methods - which reason about higher-level concepts such as paths and groups, (which create a context) - only suggest moves that produce light shapes.

yes, context is what it's all about.
good grief, i agree with The Bishop ! who would have thunk it? :study:
Attachment:
Screenshot from 2017-09-23 21-27-42.png
Screenshot from 2017-09-23 21-27-42.png [ 175.13 KiB | Viewed 9531 times ]

Page 1 of 1 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/