Kageyama's Fundamentals, dia. Nets-10

Kirby · **#21**

I liked Kageyama's book. It was one of my favorite English go books when I was first starting out.

I don't think I really liked it for any technical reasons - I don't recall learning anything specific about go that was all that practically useful. But somehow, I found it somewhat inspiring in terms of the attitude or psychology behind playing go. I guess he came across as a little arrogant at times, but overall, more inspiring than arrogant.

So I don't really mind if he was fuzzy or imprecise in that text - the technical content wasn't the main value I got from that book, I think... All that being said, it's been a long time since I read that book, so maybe I'd get a different opinion if I read it now, many years later.

Ferran · **#22**

Sorry it took me so long to answer. Busy.

Thanks to everyone for the answers. I really didn't expect so many of those, much less to start a debate about the merits of Kageyama-ss' writing.

@Bill Spight: as far as I've been able to see, my handheld leela [*] wants to turn S16 & R17, at which point there's not much point for black in playing there. In the board I prepared, Leela recommends a low approach to the lower hoshi, at R6. I'm supposing it's using the group it just created as a springboard, even if it's kind of far away. If I net, then there's ot much point for white right there for now and tenuki (again in my Leela, one of meany options halfway between the two stones I had on D16 and K16), so I lose sente, don't I?

@Uberdude: Am I reading it right and you're AKA James D? My pleasure. Also my pleasure if I'm wrong, mind you, but it's good to put a... er... "face"?

Regarding your suggestion, I do like S16, and it is my preferred choice if we go down that path, but it was neither Kageyama-ss' nor Leela's. When two suggestions from so very different background agree, I tend to think I'm wrong.

J. Fairbairn wrote:

"Fuzzy-wuzzy was a bear, Fuzzy-wuzzy wasn't fussy wuzzy"

Is that... British? I think I understand, but... er... uh... You know the old error? "Type mismatch. Division by zero. Redo from start"? My brain is trying to do something similar.

And you're right about Maths and children. Last week I spent most of it trying to get my grade school kid into differential calculus. Oh, the headaches and the gnashing of teeth! But now I know I have to watch out for the youngest. I'll remain always vigilant!

Sheesh... seriously. Those teachers...

:razz:

@mhlepore: My first temptation is closing the net, either Sakata's way or the bad one. But I really like the one with aji.

Thank you all for your time; again, my apologies for the delay. Take care, stay healthy.

[*] Does the very idea of having a superhuman Go computer in the back of your pants ping your cognitive dissonance as much as mine?

Bill Spight · **#23**

Ferran, a diagram or two would help. Thanks.

Ferran · **#24**

Click Here To Show Diagram Code: [go]$$c Now... on dia 10, page 31 $$ ------------------ $$ . . . . . . . . | $$ . . . . . . . . | $$ . . . O a 6 . . | $$ . O . O X X 4 . | $$ . . O X O 5 2 . | $$ . . . X O . 3 . | $$ . . . X 1 . . . | $$ . . . . . . . . |[/go]

Okay... NOW Leela wants to escape at (a). I've done a very minimal board with no other stone than the "spare" black at the opposite hoshi. But if I put stones on the other hoshi (I mean, a completely empty board feels wrong), (a) drops in priority, sharp, and black wants to continue near the hoshi (the ones that share a side, that is, not the diagonal).

But, well, an old Leela and that, but it's giving black around 80% with the diagram above (75% with net), so maybe a better question would be how did White allow itself to get cornered like that, wouldn't it.

No worries. I mean, if you're having fun, I'm up to it, but I don't want to bother too much.

Take care.

Bill Spight · **#25**

Ferran wrote:

Click Here To Show Diagram Code: [go]$$c Now... on dia 10, page 31 $$ ------------------ $$ . . . . . . . . | $$ . . . . . . . . | $$ . . . O a 6 . . | $$ . O . O X X 4 . | $$ . . O X O 5 2 . | $$ . . . X O . 3 . | $$ . . . X 1 . . . | $$ . . . . . . . . |[/go]

Okay... NOW Leela wants to escape at (a).

I rather think that Kageyama would want to play at a, as well.

Click Here To Show Diagram Code: [go]$$Bcm7 Continuation $$ ------------------ $$ . . . . . . . . | $$ . . . . 2 . 4 . | $$ . . . O 1 O . . | $$ . O . O X X O . | $$ . . O X W X O . | $$ . . . X W 3 X . | $$ . . . X X . . . | $$ . . . . . . . . |[/go]

Now if White wants to save the corner White need to protect. OTOH, since Black's original aim was to capture the two :wc:

stones, if Black plays elsewhere and White plays at 1 we expect that Black will capture them. In that case, as a practical matter we might regard a play at 1 as a double sente. In theory, not, OC. So there seems to be something fishy about a tenuki.

Quote:

I've done a very minimal board with no other stone than the "spare" black at the opposite hoshi. But if I put stones on the other hoshi (I mean, a completely empty board feels wrong), (a) drops in priority, sharp, and black wants to continue near the hoshi (the ones that share a side, that is, not the diagonal).

But, well, an old Leela and that, but it's giving black around 80% with the diagram above (75% with net), so maybe a better question would be how did White allow itself to get cornered like that, wouldn't it.

A deeper look with a top bot, more than Kageyama wrote about, is a good idea. In general, I trust today's top bots over pros with regard to whole board evaluation, given enough rollouts. Even for the latest version of Leela Zero, I would want each play under consideration to get at least 10k rollouts. Also, winrate estimates and play suggestions are most accurate at around a 50-50 evaluation. To achieve that you might put one Black stone in one of the other corners and one White stone in each of the remaining two corners, or something. It doesn't have to be an actual game position. As Dave pointed out, Kageyama just made this corner position up, anyway.

Ferran · **#26**

Bill Spight wrote:

I rather think that Kageyama would want to play at a, as well.

Even so, white needs to respond. If Black nets early on, White doesn't... I think?

Quote:

[...] if Black plays elsewhere and White plays at 1 we expect that Black will capture them.

Hmmm... at 1 ? So, at the very beginning? That wasn't what I meant about tenuki. I might be miscommunicating something.

Quote:

A deeper look with a top bot, more than Kageyama wrote about, is a good idea. In general, I trust today's top bots over pros with regard to whole board evaluation, given enough rollouts.

I... sort of don't. I know bots are stronger, but...

a) They don't talk.
b) They play their own way. Specially for a DDK, I feel like following them is not even running before walking, it's rocket flying. If Yi Ch'ang-ho's games are not good for beginners, then robots...
c) Even if I could somehow absorb their way of playing... How to say this? Any roboticist can build you a robot with a better real life kill ratio than humans. Hell, cars are a better killing machine than most un-assisted humans. And I haven't gotten into tanks, or that impressive demo out there of an industrial robot doing tameshigiri [*]. All those are "better" than humans. And they do offer food for thought. But I wouldn't do kendo trying to learn from a robot.
d) ...specially until someone manages to get an AI you can ask to mimic the playing style of an historical player. Yes, you can change weights, but you need to have a catalogue of them and know each one personally. And it still doesn't cut it.

So, in general, I'd rather learn from a human. I'm enough of an introvert without need to start talking to electronics.

Quote:

To achieve that you might put one Black stone in one of the other corners and one White stone in each of the remaining two corners, or something. It doesn't have to be an actual game position. As Dave pointed out, Kageyama just made this corner position up, anyway.

Did that in my first tries, yes. Because it seemed to be obfuscating my narrative (not that I don't have that failing entirely on my own), I did that one with a simple black stone far away.

Thank you. Take care.

[*] Tatami cutting. https://www.youtube.com/watch?v=BfvVBHH2F_0

Ferran · **#27**

John Fairbairn wrote:

But bots do give us "the best move in this game position."

If I may, the best move they could play. When a robot makes an obvious mistake in ladders or ko or... it's easy to say "well, a human would never play that", but... doesn't that extend to the rest? Some moves might end up being a mistake we can't see, "proven" so some generations of robot later. Some moves might be simply too complicated for a human to keep track of. Some might be a quirk of the current network. And on an on.

I feel like we're over-trusting computers. A stronger player is not necessarily a better teacher.

Bill Spight wrote:

...but Elf has strong opinions

Do not meddle into the moyo of Elves, for they are subte and quick to anger.

Take care.

Bill Spight · **#28**

Ferran wrote:

Bill Spight wrote:

I rather think that Kageyama would want to play at a, as well.

Even so, white needs to respond. If Black nets early on, White doesn't... I think?

In general, there are few sente in the opening. But lots of close calls. As humans, we generally go by feel whether to reply locally or not. The bots have taught us that our feel is not very good.

Quote:

[...] if Black plays elsewhere and White plays at 1 we expect that Black will capture them.

Hmmm... at 1 ? So, at the very beginning? That wasn't what I meant about tenuki. I might be miscommunicating something.

I meant 1 in the current diagram, not the original.

Quote:

A deeper look with a top bot, more than Kageyama wrote about, is a good idea. In general, I trust today's top bots over pros with regard to whole board evaluation, given enough rollouts.

I... sort of don't.

Well, the bots could be stronger than humans in localized reading. But they don't do localized reading, they do global reading. And they do whole board evaluation. There are known positions where even amateurs are better than today's top bots because of their human local reading. If bots are better than humans, it is because of their global evaluation and reading.

Quote:

I know bots are stronger, but...

a) They don't talk.
b) They play their own way. Specially for a DDK, I feel like following them is not even running before walking, it's rocket flying. If Yi Ch'ang-ho's games are not good for beginners, then robots...

Who says Yi Ch'ang-ho's games are not good for beginners? Because they don't understand what he does? Beginners have the advantage of not understanding any good player's play.

There are many paths up the mountain.

Quote:

c) Even if I could somehow absorb their way of playing... How to say this? Any roboticist can build you a robot with a better real life kill ratio than humans. Hell, cars are a better killing machine than most un-assisted humans. And I haven't gotten into tanks, or that impressive demo out there of an industrial robot doing tameshigiri [*]. All those are "better" than humans. And they do offer food for thought. But I wouldn't do kendo trying to learn from a robot.

Don't put obstacles in your own path.

Quote:

d) ...specially until someone manages to get an AI you can ask to mimic the playing style of an historical player. Yes, you can change weights, but you need to have a catalogue of them and know each one personally. And it still doesn't cut it.

Imitation is a good way to learn. Why not imitate the best?

Quote:

So, in general, I'd rather learn from a human. I'm enough of an introvert without need to start talking to electronics.

I am not trying to discourage you from learning from humans. OTOH, there is nothing wrong with yelling at the TV.

And humans today are learning how to learn from the bots. You have the advantage of having less to unlearn.

sorin · **#29**

I am confused about the "sente" part: to me, whether black plays 'a' or 'b', it is gote.
It is white that has different choices in the two cases (like you posted a variation later in the thread), but that has nothing to do with the fact that black's capture is gote. Am I missing something?

Ferran wrote:

I hope this is the section to post this. It doesn't quite feel to belong to Beginners. I'll be using the 6th printing of the book, 2007.

Click Here To Show Diagram Code: [go]$$c Now... on dia 10, page 31 $$ ------------------ $$ . . . . . . . . | $$ . . . . . . . . | $$ . . . O . . . . | $$ . O . O X X . . | $$ . . O X O . c . | $$ . . . X O . d . | $$ . . . X a b . . | $$ . . . . . . . . |[/go]

Kageyama (and my handheld Leela) recommends 'a'. Then 'c' answered by 'd'. As far as can see, this gives White the NE corner in exchange for his stones, influence and sente. However, Black at 'b' settles the matter (Leela doesn't want to play anywhere close; like, really doesn't wanna) and doesn't surrender, BUT losses sente. With the game prepared so that I could evaluate the position (stones of same color on opossite hoshi, a high 2-space pincer on NW White), the difference was about 10%, IIRC, in favor of closing at 'a'.

My question is... is that difference due only to sente (and maybe the lack of defects facing South) or am I missing a lot more?

Thank you. Take care.

Uberdude · **#30**

Hopefully I can clarify. Here is my understanding of Ferran's initial post, rephrased and made more explicit to reveal the sente or not argument and why it's faulty (as the stronger players reading already know, so struggled to think like Ferran).

Click Here To Show Diagram Code: [go]$$c $$ ------------------ $$ . . . . . . . . | $$ . . . . . . . . | $$ . . . O . . . . | $$ . O . O X X . . | $$ . . O X O . c . | $$ . . . X O . d . | $$ . . . X a b . . | $$ . . . . . . . . |[/go]

Ferran2:
Kageyama says black should play a, and if white c next then black can deal with this at d (Uberdude: actually d might not be best, but it's a good tesuji to know about that readers might not so that's a valid pedagogical reason to talk about it), but I misinterpreted that as white will/should continue at c immediately. This would likely continue like so, and white takes the corner in gote, black ends with sente:

Click Here To Show Diagram Code: [go]$$c $$ ------------------ $$ . . . . . . . . | $$ . . . . 8 . 0 . | $$ . . . O 7 6 . . | $$ . O . O X X 4 . | $$ . . O X O 5 2 . | $$ . . . X O 9 3 . | $$ . . . X 1 . . . | $$ . . . . . . . . |[/go]

If black plays the net, Leela wants to tenuki, so I think the net sequence of good play ends after 1 move and is gote for black:

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

Because I considered the turn and then white c as a sequence as a whole, I consider the turn a sequence ending in black's sente, which is different to the net.

Uberdude: but of course white doesn't need to play c immediately. The turn, just like the net, can be considered as the conclusion of local play, and is gote:

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

Ferran, is that what you meant?

Uberdude · **#31**

NB: I have moved John and Bill's posts about the other Kageyama net position to a new thread https://lifein19x19.com/viewtopic.php?f=15&t=17510, to avoid cluttering this one, and will reply there.

Ferran · **#32**

Sorry by my delay replying.

@Uberdude: mostly. I'm aware (kind of; as with many things, I know but I don't quite "grok" [*]) that White doesn't have to play there. Still, there's a benefit in playing there. Might be after a tenuki or not, maybe there are more urgent plays elsewhere, but there is, as long as the position nearby allows. I can see no benefit for White in playing near the net (unless the position nearby changes a lot). If White does play there, then there's a benefit for Black to respond locally.

Thanks. Take care.

[*] Those that make sense to non-SF fans?

sorin · **#33**

Thanks Uberdude, in the meanwhile Ferran also clarified.
Indeed, if the topic of discussion is "which of two ways of capturing is locally better", going into too many details of sente/gote is a distraction and may make one confused!
I see the variation shown for white as just an argument why 'a' is a weaker way for black to capture (as in 'white has this extra option').

Up one level now: maybe black may choose the weaker 'a' just because they are aware of white's follow-up, and they want to lure white into playing that sequence, and losing sente :-)

Uberdude wrote:

Hopefully I can clarify. Here is my understanding of Ferran's initial post, rephrased and made more explicit to reveal the sente or not argument and why it's faulty (as the stronger players reading already know, so struggled to think like Ferran).

Kageyama's Fundamentals, dia. Nets-10

Who is online