Life In 19x19

Click Here To Show Diagram Code

[go]$$Wc White to play.
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . O X . . |
$$ | . . . O . . . . . , . . . O . O X . . |
$$ | . . . . . X . . . . . . . . O X X . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

White was Rin Kaiho, one of those textbook writers. In the commentary, Elf reckons that Rin's play loses 22½% versus Elf's choice.

lightvector wrote:Can anyone with better positional judgment understand why this result isn't mediocre for white, compared to the other line? White's wall in this one doesn't feel like it's facing anything useful, and white didn't get to move first to build in the center.

Well, I can no longer trust my positional judgement.

But out of curiosity I tried out a couple of ways of evaluating positions. The Elf position has a proximity score of +18 (for Black). The alternate KataGo position has a proximity score of +19, a 1 pt. difference. (OC, taken as a projection of the final score, this is way, way off. White is in the lead, right?) My evaluation function 0.1, which I used to think usually got within ± 5 pts. or so, evaluates the Elf position as +1.0 and evaluates the KataGo position as +1.8, a difference of 0.8 pts. That is also a static evaluation, so the value of sente is not included. This function also overestimates Black's chances, but not by as much.

However, despite their overall inaccuracy, the main difference between the compared scores lies with the bottom left corner. And both methods of evaluation evaluate that region almost the same for the two scenarios. My guess is that DrStraw's and Robert Jasiek's methods would also produce similar scores. {shrug}

Can anyone with better positional judgment understand why this result isn't mediocre for white, compared to the other line? White's wall in this one doesn't feel like it's facing anything useful, and white didn't get to move first to build in the center.

I can't put any sort of value on either position, or reliably suggest follow-up moves. But I can say I didn't feel surprised by the bot assessments, and I believe the human pro assessments would be similar. I feel the hiccup may be in too literal an interpretation of the phrase "White's wall in this one doesn't feel like it's facing anything useful". My preferred phrase for thickness in general is "it's only thickness is it functions as thickness", and there are more functions than surrounding territory or providing a cliff wall for your waves to batter a helpless mariner against. Walls, in particular, can have an important dividing effect - as here.

My feel (and, I repeat, based only on the sorts of comments I see pros make, and not on any specific insights) is that in the surrounding-move line, Black's reducing moves will give him strength not only in the centre but strength that cascades out to the top and left side. In addition, the configuration on the left edge is such that Black's corner group provides some support for his left-side group higher up.

With the dividing wall, however, Black's group on the higher left is thin (and I am struck, over and over and over again by how often pros make such assessments - I see a sturdy house, they see rotting foundations). Furthermore, Black now has no special prospects in the centre or on the upper side, not just because he lacks the strength-giving erasure moves of the other lines, but because the effects of his weakness on the left side will cascade out in to the centre and upper side.

There's another aspect where I feel, but only in my water rather than in my head, there is something wrong, and that lies in saying White comes away with sente to play the surrounding move. I can't really explain my disquiet, but to me it doesn't feel like sente. Surrounding moves (kakoi) are typically gote. They are a kind of honte. Proper sente here should mean you have the initiative. I think it's important not to mix up the endgame kind of sente ("he has to answer") with strategic sente ("I get to control what happens next, and where").

As a low dan I won't claim superior positional judgment but lightvector's question made me recall previous thoughts. I think they are in line with John's - pardon if they aren't.

1 - I remember someone mentioning Go Seigen's insight that strong groups somewhere around the 7-7 point have a lot of strategic value. Pardon again if that's not what he said. It stuck with me like this. White's strong group in the lower right is such a group. We know the proverb "use thickness to attack" so well that we may be lured into thinking there's no other value in thickness. However, thickness has intrinsic value too of being worry free.

2 - In Katago's "hard to get" diagram, White is creating another such group in the lower left, probably more so than in Elf's line. As John describes, it has a splitting effect on Black's seemingly "sturdy house" above. I don't think it's rotten per se but White's new 7-7 strength reduces Black's.

3 - A perennial difference between amateurs and professionals, which is echo'd by AI, is that amateurs will much rather fall in love with moyos like the one in Elf's line. Look at that grand area! Professionals care about thickness, territory, efficient shape ... like AIs do. So if Elf has this preferred line, it'll rather be because of its efficiency than because of the visual appeal of the moyo (obviously bots don't share our gluttonous view of the world). Katago's line doesn't have that likeable area development but it is probably just as efficient in making strength and making points.

All of this is hindsight. If I saw those 2 sequences, I'd always go for the moyo.

A perennial difference between amateurs and professionals, which is echo'd by AI, is that amateurs will much rather fall in love with moyos like the one in Elf's line.

I think that this is actually a hugely important statement. It's a valuable insight into yet another area of theory that is just not addressed yet in the west.

It's not that well addressed in the oriental countries, but there are some major contributions.

I think the best I have seen is O Rissei's 2015 book Teatsusa no Wana - about the traps inherent in thickness. Even though his book is pre-AI, it's remarkably prescient about the sort of things about thickness that AI exposed (exposed, not discovered, I'd say - given, O's book).

Two things leap out at you from O's book. One from the cover, and the other (related to the cover) from the first line of the text.

One is that for "thickness" on the cover he uses "teatsusa". I have discussed the difference between atsusa and atsumi here before, and the significance of the extra te- is something I have discussed at length in Go Wisdom. This forum has shown in the past it has little appetite for discussion of terminology, so I'm saying no about that now except to point out that the first heading on the text pages is very different from the book cover: it is "Atsumi no wana to wa" - what is meant by traps inherent in thickness. Much of the book is about this very important difference between atsumi and teatsusa.

However, the other neon light shining from the cover is a sub-heading that tells us the book is about "How to use thickness [atsumi] and moyos correctly." "Use" is the important word.

As knotwilg points out, amateurs fall in love with moyos in their own special way. We might say this is a teenage, lustful way, rather than a more mature "let's build a family life together" kind of way.

Because it's lust, the dominant terminology among amateurs is "get a moyo, build a moyo, this moyo is my territory, and so on. But O talks about using a moyo, and he shows how it is often used in the same way as thickness. And to my, admittedly inexpert, eye the examples he shows look remarkably like the sort of AI positions I, and apparently others, have most trouble with.

Here is a trivial example from the beginning of the book. It's trivial in the sense that it's just part of the scene-setting introduction. And it's about thickness rather than moyos. But I think it shows enough to indicate the sort of hidden byways down which O will take you. More to the point, I think it also resembles lightvector's position in some important respects.



O discusses this position (minus the letters) as part of explaining "what is thickness [atsumi]". It's quite rare in professional play. Pros have tried Black A, B and C here. Leela likes A, and I feel confident in guessing the reason this fuseki is rare is because pros as White don't like Black playing that. In his explanation O uses B, which in my Leela loses about half a percentage point in win rate, but I'm pretty sure O preferred to use B simply for the purposes of his follow-up explanations.

First he shows the next position: He notes that stones than can be attacked are not thick. So Black has "attacked" at the centre of symmetry. But he ends up as the rabbit being chased. He is allowing White to let "thickness function as thickness" (my phrase, not O's, but same idea).



So does that mean White's thickness is good? Leaving aside the tactical complexities that issue from A above, and so sticking B, O simply points out how Black can play round about and use reducing moves to end up with this next position:



I imagine even kyu players can see that Black dominates the upper part of board here, and White has not got much out of his "thickness". But look back at the first diagram. Would you honestly say that your assessment there would be that Black inherently can dominate the top, because of White has done below? This is one of the traps about thickness [atsumi]. (As a reminder, the book is actually about teatsusa.)

In the same way, looking at lightvectors's example, would you honestly spot that in the dividing-wall case Black inherently dominates the top half of the board?

Consider that lightvector referred to the surrounding-move line as creating a moyo. If I were bored I might quibble and say that it's now a jimoyo rather than a moyo, and if you insisted it was a moyo nevertheless, I'd ask about what O highlights: how then does White use his moyo? (The answer, of course, is that he has already used it to make a jimoyo - a territorial framework as opposed to a framework, which is by no means a bad thing, but recognising that does have implications for understanding the dynamism of the game as a whole.)

This forum has shown in the past it has little appetite for discussion of terminology

Quite the contrary I'd say. Your treatise of atsusa/atsumi is one of the things that has resonated very strongly with me. I don't remember the te- discussion as vividly though.

We might say this is a teenage, lustful way

I like the qualification of "lustful". I think that's a more accurate description of what is usually described as "greedy". That term has also been used to describe "envious" invasions. "Lustful" (building) and "envious" (invading) are useful terms to dissociate both aspects of greed.

I'd ask about what O highlights: how then does White use his moyo?

Interpreting the surrounding message: "moyo" is a particular kind of "influence", being potential territory which however can still be invaded. "using a moyo" then means: "what value will you get out of it if it gets invaded (without dying)". The "jimoyo" is a big surrounded area which may be reduced here and there but is actually close to territory already. There's no "using" it anymore because it has already found its destiny.

As for the final example: amateurs will likely overvalue the Black influence towards the top and undervalue the "overconcentrated" White group, which is however 100% alive with lots of points and as such influencing the surrounding areas, say the whole board, in a not-so-visual way.


(I've repeatedly resisted getting into the historical argument but I still want to point out that I very much appreciate the nuances in the discussion, I acknowledge that Japanese terms have settled to convey these nuances but I prefer finding English terms that do the same (or otherwise bring nuance, using the strength of our lingua franca), rather than adopting all the Japanese terms, which I'll inevitable continue to use wrongly because I don't have a "sense" for that language which is much more remote from my native language.)

John Fairbairn wrote:So does that mean White's thickness is good? Leaving aside the tactical complexities that issue from A above, and so sticking B, O simply points out how Black can play round about and use reducing moves to end up with this next position:



I imagine even kyu players can see that Black dominates the upper part of board here, and White has not got much out of his "thickness". But look back at the first diagram. Would you honestly say that your assessment there would be that Black inherently can dominate the top, because of White has done below?

I can honestly say that that would not be my assessment, even having seen the after diagram. The reason is that, even though I am much weaker than the author, I would have my doubts about the line of play between the before and after diagram. Maybe the dominance is the result of passive play by White in that sequence. In addition, Black has made one more play than White in that sequence, which means that we have to allow for that fact in our comparison. The sequence is 15 ply long, and few of those moves are how Black and White would actually play, or we would have seen the after diagram in actual pro games. (I have similar doubts about long bot variations, as well, for different reasons.)

Consider that lightvector referred to the surrounding-move line as creating a moyo. If I were bored I might quibble and say that it's now a jimoyo rather than a moyo, and if you insisted it was a moyo nevertheless, I'd ask about what O highlights: how then does White use his moyo? (The answer, of course, is that he has already used it to make a jimoyo - a territorial framework as opposed to a framework, which is by no means a bad thing, but recognising that does have implications for understanding the dynamism of the game as a whole.)

I think that the distinction between a moyo (framework) and a jimoyo (territorial framework) is significant. In this game Rin decided to protect the corner rather than make a jimoyo. Elf disagrees, and, I suspect, Jowa and other 19th century lights would have, as well, if they were to take komi into account. I think that the bots are telling us that jimoyo are not overconcentrated. By contrast, they do not value early moyo highly.

In the problem position, I would probably make the d3 iron pillar, a move I have liked for many years, even though from studying with bots I now know they rarely like it. To my mind, it has these benefits:
- secures* the corner points
- makes eyes for a safe group, if black jumps in at 3-3 then with the large knight he can make a roomy life whilst taking all the eyes and points.
- weakens f3, aiming at h3 pincer, and even if black answers with 2 space extension to j3 then l3 is still an aim to pressure the group.
- simple

But it has these drawbacks, which bots seem to weight more highly than I tend to:
- the corner isn't actually yours: well c3 might be, but the c5/6 area isn't as black has invasions there, particularly with c9 AND e9 support.
- commits to a plan, so loses flexibility
- related to above, possible to get kikashid or overconcentrated, black may be able to bully from the outside
- f3 for c3 is a good sente exchange for black as he starts operations on the lower side if that's the direction he wants (I used to think as white I was happy black lost the chance to 3-3, but bots don't seem to like the classic 3-3 invasion of 4-4 plus big knight in most situations)

I can see on this board with a large lower side potential for white that a pincer might be better, but that still makes me uneasy as I worry that c7 can become misplaced with a pincer. However, viewing c7 as a beneficial exchange of side for centre with e9 makes it easier to treat lightly.

Anyone with better positional judgment? Come on lightvector, you know whom you got to ask.


It is not me for the slow minded

Bill Spight wrote: I think that a lot of humans would choose the one space pincer, which Elf prefers by 22½%. At first blush it appears to go against the proverb not to make territory out of thickness.

$$Wcm26 Elf's mainline
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . 0 . . . . . . a . . . . . . O . . . |
$$ | . 8 9 . . . . . . ? ? ? ? ? ? O X . . |
$$ | . . . O . . . . . ? ? ? ? O . O X . . |
$$ | . . 2 3 . X . 1 ? ? ? ? ? ? O X X . . |
$$ | . . 6 4 5 7 ? ? ? ? ? ? ? ? . . . . . |
$$ | . . . . . . ? ? ? ? ? ? ? . . . . . . |
$$ ---------------------------------------

Click Here To Show Diagram Code

[go]$$Wcm26 Elf's mainline
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . 0 . . . . . . a . . . . . . O . . . |
$$ | . 8 9 . . . . . . ? ? ? ? ? ? O X . . |
$$ | . . . O . . . . . ? ? ? ? O . O X . . |
$$ | . . 2 3 . X . 1 ? ? ? ? ? ? O X X . . |
$$ | . . 6 4 5 7 ? ? ? ? ? ? ? ? . . . . . |
$$ | . . . . . . ? ? ? ? ? ? ? . . . . . . |
$$ ---------------------------------------[/go]

I'd rather use that thickness to make this big chunky (almost) territory than this piddly little one:

Bill Spight wrote:

$$Wcm26 GoGoD 1969-00-00e, Rin Kaiho, 9 dan (W) vs. Otake Hideo, 8 dan
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . 4 . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . . a . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . O X . . |
$$ | . . . O . . . . . , . . . O . O X . . |
$$ | . . . 1 . X . . 2 . 3 ? ? ? O X X . . |
$$ | . . . . . . . . . . . ? ? ? . . . . . |
$$ | . . . . . . . . . . . ? ? . . . . . . |
$$ ---------------------------------------

Click Here To Show Diagram Code

[go]$$Wcm26 GoGoD 1969-00-00e, Rin Kaiho, 9 dan (W) vs. Otake Hideo, 8 dan
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . 4 . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . . a . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . O X . . |
$$ | . . . O . . . . . , . . . O . O X . . |
$$ | . . . 1 . X . . 2 . 3 ? ? ? O X X . . |
$$ | . . . . . . . . . . . ? ? ? . . . . . |
$$ | . . . . . . . . . . . ? ? . . . . . . |
$$ ---------------------------------------[/go]

secures the bottom left corner — well, almost. makes a base, which attacks. is joseki, but Elf regards it as losing 20% to invading the bottom left corner at a.

Sure, it also/primarily functions as an attack on the lower group, but for what profit? It only makes a piddly little overconcentrated territory. White does also have the corner profit too, though that's not solid yet.

"Don't make territory with thickness" should really be "Don't make inefficient territory with thickness", just a specialisation of the uber-proverb that pretty much covers all of Go "don't be inefficient".

Regarding lightvector/KataGo's other choice, the kakoi lower side variation does feel nicer to me too, but John's comment

John Fairbairn wrote:Walls, in particular, can have an important dividing effect - as here.
<snip>
In addition, the configuration on the left edge is such that Black's corner group provides some support for his left-side group higher up.
With the dividing wall, however, Black's group on the higher left is thin

is also what sticks out to me when judging the benefits of that line. White is playing a thick topological game, so has future aims like this textbook invasion sequence against the top left black group. Then the white wall below is doing something. White also aims at playing big endgame moves like b4 with life and death implications in the middlegame (e.g. if black comes into the lower side in later fighting white tries to get a move around g3/g2 which then make b4 sente).

Click Here To Show Diagram Code

[go]$$Bcm37
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . 3 . X . . . . . . . . . . . . . . . |
$$ | . . 2 . 8 . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . 4 . 6 . . . . . , . . . . . , . . . |
$$ | . 5 X 7 X . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . O . . . . |
$$ | . . O . . . . . . . 1 . . . . . X . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . O X . . |
$$ | . . O O O O . . . , . . . O . O X . . |
$$ | . O X O X X . O . . . . . . O X X . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

John Fairbairn wrote: There's another aspect where I feel, but only in my water rather than in my head, there is something wrong, and that lies in saying White comes away with sente to play the surrounding move. I can't really explain my disquiet, but to me it doesn't feel like sente. Surrounding moves (kakoi) are typically gote. They are a kind of honte. Proper sente here should mean you have the initiative. I think it's important not to mix up the endgame kind of sente ("he has to answer") with strategic sente ("I get to control what happens next, and where").

Well, lightvector didn't actually say sente but "and white didn't get to move first to build in the center.". But I see and agree with the feeling that the kakoi is necessary/honte to consolidate the gains made by the direction choice of the pincer and giving up the corner for the lower side jimoyo, so that sequence as a whole is gote (like this example I posted from PhoenixGo). But the point is you have sente after the local corner sequence so you can kakoi, with the dividing wall you can't kakoi as the corner sequence ends in gote so that's playing 2 moves in a row:

Click Here To Show Diagram Code

[go]$$Wcm34 White cheats
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . O . . . . . . . . X . . . . |
$$ | . . O , . . . . . , . . . . . , X . . |
$$ | . . O X . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . X . X . . . . . . . . . . . . . . |
$$ | . . . . . . . . 5 . . . . . O . . . . |
$$ | . . O . . . . . . . . . . . . . X . . |
$$ | . . . . . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . O X . . |
$$ | . . O O 3 1 . . . , . . . O . O X . . |
$$ | . O X O 2 X . O . . . . . . O X X . . |
$$ | . X X X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

P.S John, is there supposed to be a black stone at f3 in some of your later diagrams from O Rissei's book?

P.S John, is there supposed to be a black stone at f3 in some of your later diagrams from O Rissei's book?

It's early morning, so a few hours more before I wake up properly, but with that caveat: no. Black has one more stone than White.