Opening problems for AI: Problem 12

Bill Spight · Post by **Bill Spight** » Wed Jun 24, 2020 3:42 pm

Today's problem comes frome GoGoD 1997-12-11a, Rin Kaiho (W) vs. Takemiya Masaaki, who played yonrensei.

Click Here To Show Diagram Code: [go]$$Wcm24 White to play. $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . X O O O O . . | $$ | . . . O . . . . . , . . X X O X X O . | $$ | . . . . . . . . . . . . . O X X . . . | $$ | . . . . . . . . . . . . . O . . O . . | $$ | . . . . . . . . . . . . . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

This position occurs two other times on Waltheri. Each of the other pros made the same play as Rin, which Elf reckons loses 17% to par.

Bill Spight · Post by **Bill Spight** » Wed Jun 24, 2020 3:44 pm

Click Here To Show Diagram Code: [go]$$Wcm24 White moves into the center $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . X O O O O 3 . | $$ | . . . O . . . . . , a . X X O X X O . | $$ | . . . . . . . . . . . 4 . O X X . 2 . | $$ | . . . . . . . . . . . . . O . . O . . | $$ | . . . . . . . . . . . 1 . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

moves out into the center, then

plays sente in the corner. This joseki is following a familiar pattern, where one side gets the corner and plays into the center while the other side plays on both sides. Takemiya played

, but Elf prefers a.

Elf's mainline for

, which is marked

Click Here To Show Diagram Code: [go]$$Wcm28 Elf's mainline variation $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . 7 . 9 . . . X O O O O O . | $$ | . . . O . . . . . , B . X X O X X O . | $$ | . . . . . . . . . . . . . O X X . X . | $$ | . . 8 . . . . . . . 0 . . O . . O . . | $$ | . . . . . . . . . . . O . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 6 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . X . . . . 4 X 2 . . | $$ | . . . . . . . . . . . . . . . 3 1 . . | $$ | . . . . . . . . . . . . . 5 . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

invades the bottom right corner. Black takes sente and approaches the bottom left with

extends from the top left corner towards the Black group on the top side. When

approaches the top left corner, White extends again. Then

jumps into the center for a fight with the White stones there.

Elf's mainline for

Click Here To Show Diagram Code: [go]$$Wcm24 White connects the top right corner and side $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . 4 . . . . . . X O O O O . . | $$ | . . . O . . . . . , . . X X O X X O . | $$ | . . . . . . . . . . . . . W X X . 1 . | $$ | . . . . . . . . . . . 2 . W . . O . . | $$ | . . . . . . . . . . . . . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . 9 . | $$ | . . . . . . . . . . . . . . . 8 . . . | $$ | . . . O . . . . . X . . . . . X 7 . . | $$ | . . . . . 3 . . . . . . . . . 6 5 . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

connects the White stones, in exchange for allowing

in the center. It would be foolish for the two

stones to run out now. The key to this problem is the relative evaluation of the top right and the corner.

The rest of the mainline shows Elf's projected development. As always, long variations should be taken with a grain of salt. That said, this variation is based on a total of 141.2k rollouts.

"encloses" the bottom left corner. Then

approaches the top left corner. Next,

invades the bottom right corner.

Click Here To Show Diagram Code: [go]$$Bcm33 Mainline, continued $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 6 5 7 X . . . . . . X O O O O . . | $$ | . . . O 9 2 3 . . , . . X X O X X O . | $$ | . . . 0 . 4 . . . . . . . O X X . O . | $$ | . . 1 8 . . . . . . . X . O . . O . . | $$ | . . . . . . . . . . . . . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . O . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . O . . . . . X . . . . . X O . . | $$ | . . . . . O . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

makes a double approach to the top left corner.

Click Here To Show Diagram Code: [go]$$Bcm43 Mainline, III $$ --------------------------------------- $$ | . 7 . . . . . . . . . . . . . . . . . | $$ | . 6 1 . . . . . . . . . . . . . . . . | $$ | 8 3 O X X X . . . . . . X O O O O . . | $$ | . 2 4 O X O X . . , . . X X O X X O . | $$ | . . . O . O . . . . . . . O X X . O . | $$ | . . X O . . . . . . . X . O . . O . . | $$ | . . . . . . . . . . . . . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 5 , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 0 . . | $$ | . . 9 . . . . . . . . . . . . . . O . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . . . O . . . . . X . . . . . X O . . | $$ | . . . . . O . . . . . . . . . X O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

The top left joseki continues.

plays into the left side, then White captures

in the corner.

approaches the bottom left group. Then

moves further into the right side and looks towards the center.

Click Here To Show Diagram Code: [go]$$Bcm53 Mainline, IV $$ --------------------------------------- $$ | . X . . . . . . . . . . . . . . . . . | $$ | . O X . . . . . . . . . . . . . . . . | $$ | O . O X X X . . . . . . X O O O O . . | $$ | . O O O X O X . . , . . X X O X X O . | $$ | . . . O . O . . . . . . . O X X . O . | $$ | . . X O . . . . . . . X . O . . O . . | $$ | . . . . . . . . . . . . . . X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . X . . . . . . . . . . . . . . O . | $$ | . . . . . . . . . . . . . . . X . . . | $$ | . 3 2 O . . . . . X . . . . . X O . . | $$ | . 7 1 4 . O . . . . . . . . . X O . . | $$ | . 5 6 . . . . . 8 . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Finally,

invades the bottom left corner.

turns, allowing Black to connect to the left side. Then White stabilizes its group with the slide,

.

While of course we cannot be sure that this variation is all correct, it does give us a picture to help us to evaluate the connection with

.

John Fairbairn · Post by **John Fairbairn** » Thu Jun 25, 2020 1:38 am

Bill,

First, a minor point. GoGoD has 11 games with this position and in 2 of them the second-line 24 is played. Both examples were in China, just after the Rin-Takemiya game, which hints at them thinking this was an important on the Japanese play. Takemiya appeared in 4 of the 11 games, so it was clearly a stylistic choice as much as a joseki choice in his case.

But the major point may have to do with 27. The Japanese commentary on this game does not mention 24 but does mention 27 and says it is the "vital point". In another Takemiya game of the same period, the commentary does not mention 24 or 27 but does mention that White was remiss in not grabbing the vital point at O13.

It seems to me that "vital point" is another term that will have to be arraigned before the hanging judge in the light of AI evidence, but for different reasons in Japan and the English-speaking world.

In English it implies the crucial point of a shape - a static thing. In that sense we would find nothing wrong with Black 27 I'd suggest. It is the crucial point of the shape.

But in Japanese the term kyuusho is the "urgent point". It's a dynamic, time-based comment. And one thing I've noticed about very many AI plays that surprise humans is that they are based on a different time frame from our usual one. I think this game is a case in point. Often bots play moves earlier than we expected. Often later. They often play (or accept) the pro move but at a different time. I'd go so far as to suggest that, after efficiency/overconcentration, timing is most important difference in AI play, and one that, unlike eff/overconc, is one that even the pros have yet to appreciate - although I did point out in another thread they are starting to get to grips with at least the "early" aspect of it, trying out early probes and contact moves before making the standard pro play. The fact they vary so much n which reconnaissance plays to try shows they haven't quite got there yet, but at least they are in the right area. In contrast, I personally see no evidence (I'm referring mainly to commentaries I read rather than my understanding of actual games) that they've got to grips with "late" timing, as exemplified by this game. Elf clearly thinks the solid watari is more urgent than prancing about in the wide-open centre.

(For those very few who are prepared to consider terminological improvements, the Nihon Ki-in definition of a kyuusho is "a point that you have to play on urgently. An important point for attack and defence." It says nothing about shape.)

Knotwilg · Post by **Knotwilg** » Thu Jun 25, 2020 4:33 am

John Fairbairn wrote:Bill,

In English it implies the crucial point of a shape - a static thing. In that sense we would find nothing wrong with Black 27 I'd suggest. It is the crucial point of the shape.

But in Japanese the term kyuusho is the "urgent point". It's a dynamic, time-based comment.

While the importance of timing is clear, the notion of static shape is probably more helpful than that of "urgency", at least for amateurs like me. When I'm looking for my next move, I will consider a.O. those that improve my shape, for which I have heuristics, while an "urgent" move is even higher up the conceptual chain, close to "the best move".

In the urgent vs big discussion, Minue's fundamentals have helped me translating "urgent" as "improving the health of an existing group", and "big" as "developing open area". Still, sometimes a group's health needs improvement, sometimes it doesn't. How to know? Often that's again down to shape.

For "shape" we have underlying, basic concepts like number of liberties, connectivity and eyespace.

For "timing" I know less heuristics.
- we can probe an opponent for his answer, then continue in a way that benefits us more, given the answer. Knowledge of the concept, of this potential helps in general, but not in specific cases. Except for the classical 2nd line attachment to a small 3-4 enclosure, which probably explains its popularity.
- we should reduce/invade a moyo right before its completion
- and obviously, corners, sides, centre, tells us about priorities
- there's the "temperature" concept, which I (also) have found to be a fancy way of stating that playing elsewhere is or isn't an option.

John Fairbairn · Post by **John Fairbairn** » Thu Jun 25, 2020 5:45 am

For "timing" I know less heuristics.

There probably are fewer heuristics (though I can think of some more instantly: go is about the order of moves; urgent points before big points; get the tedomari). But so what?

This discussion is about trying to find out what AI can tell us. If we knew that already, i.e. had lots of heuristics, we wouldn't need to be looking.

And not only is this is not a test of what we think we know already, it's a call to re-evaluate precisely what we think we know but perhaps don't really. Sticking with influence, enclosure, vital point and so on just because you've used them for yonks can be intellectual laziness, can't it? Well, I at least have switched to reinforcement - thanks to you

On a broader tack, I have an impression from the examples you quoted that you are perhaps limiting timing either to the area around local plays, or to plays immediately before or after a local play. My impression from AI games, however, is that we have to expand that considerably (just as the Direct 3-3 teaches). In the example game of this thread, for example, AI may be telling us that not only should White not be playing in the upper-right centre NOW, but he maybe shouldn't be playing there at all for another 20, 30, 40, 50 moves. I'm cleaving to my recent observations in another thread that you should be staying well away from strength (including reinforcements!)

In short, maybe good go is about no-go areas.

As for over-reliance on shape, I thought the Koreans had already scotched that with haengma. There is a recent (post-AI) Korean book "Why is Korean go so strong" by Cheong Su-hyeon 9-dan which the Japanese themselves decided had to be translated into Japanese. Although, from memory, he doesn't talk specifically about AI, many of his vast array of topics span both haengma and AI (e.g. development, josekis for sabaki [as opposed to josekis for par of equality], the order of moves in the fuseki, the relationship between tesujis and vital [=urgent] points.) Cheong is, incidentally, one of those who have recently pointed out Dosaku's emphasis and efficiency and his ability to make moves that transcend common-sense. I think for 'common sense' we can substitute 'heuristics', innit?

Dosaku's being mentioned so often now I'm beginning to wonder whether pros are going back to studying his games?????

PS Writing the above stimulated me to good and look at the book again. The first thing that caught my eye was a chapter "Fettered by good shape". Cheong begins: "A strong point of Japanese go is the aspect of pretty shape. However, there is a downside: like two sides of a coin." And a little bit later he says: "Korean go has moved away from good shape." Nuff said?

ez4u · Post by **ez4u** » Thu Jun 25, 2020 6:18 am

"What if the bird won't sing?"

Katago (newest 40 block) seemed to agree with Elf when I set up the situation for White 24. However, there was a slight question remaining. Katago had blue as S15 but it assigned nearly as high a value to M15, the point where Takemiya finally played 27. So let's just wait a while... (remember my box kind of lumbers along under the weight of the 40 block). Well surprisingly, suddenly (around 40K playouts in this case) M15 turns blue! Katago does not agree with Elf, except that Rin played the wrong move for 24.

Now why? It turns out that M15 is sente. If Black blocks on the right, White now ignores it and plays to capture the three black stones at the top.

So Black has to respond and White gets to play S15 anyway.

Bill Spight · Post by **Bill Spight** » Thu Jun 25, 2020 6:18 am

John Fairbairn wrote:Bill,

First, a minor point. GoGoD has 11 games with this position and in 2 of them the second-line 24 is played. Both examples were in China, just after the Rin-Takemiya game, which hints at them thinking this was an important on the Japanese play. Takemiya appeared in 4 of the 11 games, so it was clearly a stylistic choice as much as a joseki choice in his case.

Thanks.

I imagine that Takemiya was Black in those games and posed the question of where to play

.

John Fairbairn wrote:But the major point may have to do with 27. The Japanese commentary on this game does not mention 24 but does mention 27 and says it is the "vital point". In another Takemiya game of the same period, the commentary does not mention 24 or 27 but does mention that White was remiss in not grabbing the vital point at O13.

Click Here To Show Diagram Code: [go]$$Bcm27 Vital points $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . d . . X O O O O O . | $$ | . . . O . . . . . b a . X X O X X O . | $$ | . . . . . . . . . . . 1 . O X X . X . | $$ | . . . . . . . . . . . . . O . . O . . | $$ | . . . . . . . . . . . O z 2 X X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . X . . . . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

In this game Takemiya played

, which is a shape point, since a White stone there would make a table shape. It also (sort of) threatens z. I say sort of because who wants to play z when there are bigger and better ways to attack? Rin replied at O-13, another vital point.

As far as Elf is concerned,

is a minor matter.

in the game is Elf's third choice, with 4.1k rollouts, a being first, with 77.5k rollouts, b being second, with 4.9k rollouts, and d being fourth, with 3k rollouts. In terms of estimated winrates, all four of these moves are playable, within a range of 2%. Not a big deal.

John Fairbairn wrote:It seems to me that "vital point" is another term that will have to be arraigned before the hanging judge in the light of AI evidence, but for different reasons in Japan and the English-speaking world.

In English the term is overworked. Curiously, the SL page shows the Japanese 急所　and Chinese 要点, which strike me as different concepts, especially as I have also seen both in Japanese commentaries.

John Fairbairn wrote:In English it implies the crucial point of a shape - a static thing. In that sense we would find nothing wrong with Black 27 I'd suggest. It is the crucial point of the shape.

But in Japanese the term kyuusho is the "urgent point". It's a dynamic, time-based comment. And one thing I've noticed about very many AI plays that surprise humans is that they are based on a different time frame from our usual one. I think this game is a case in point. Often bots play moves earlier than we expected. Often later. They often play (or accept) the pro move but at a different time.

Well, go is the order of play, as someone said.

Bill Spight · Post by **Bill Spight** » Thu Jun 25, 2020 6:34 am

ez4u wrote:"What if the bird won't sing?"

Katago (newest 40 block) seemed to agree with Elf when I set up the situation for White 24. However, there was a slight question remaining. Katago had blue as S15 but it assigned nearly as high a value to M15, the point where Takemiya finally played 27. So let's just wait a while... (remember my box kind of lumbers along under the weight of the 40 block). Well surprisingly, suddenly (around 40K playouts in this case) M15 turns blue! Katago does not agree with Elf, except that Rin played the wrong move for 24.

That's wonderful, Dave!

It reinforces my belief that the top bots usually agree about human errors, even if they do not agree about best play. It also reinforces John Fairbairn's point about timing.

Uberdude · Post by **Uberdude** » Thu Jun 25, 2020 6:49 am

Ha, I thought the humans would have played obvious s15 extension and Elf was going to disagree with it! But bots really don't like letting the opponent profit in sente, and letting black get s15 is exactly that (unless you play something even more sente like KG discovers).

Harleqin · Post by **Harleqin** » Thu Jun 25, 2020 1:19 pm

Maybe you can call M15 a probe, or even an inducing move.

Probe: first see if and how black wants to save those three. If he doesn't, the right side connection would not be as important anymore.

Inducing: make black reinforce the top side, so that it becomes more pressing and useful to connect on the right side.

Uberdude · Post by **Uberdude** » Thu Jun 25, 2020 2:28 pm

I wouldn't call m15 a probe or an inducing move. A probe asks a question where there are at least 2 good answers. As ez4u's picture shows KG thinks black's choice of s15 and trading the top stones is not good. So black has to answer on the top (once black reads it really was sente). And an inducing move gives you an even better reason to play the move you wanted to play anyway, not really the case here. The value of KG's jump is to, in sente, make black's first move in the centre fight not as severe as originally. A "make my cutting stones a bit less dead in sente" move. Similar to 5 here (you need the marked side stone for it to actually be sente):

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

The tricky thing with moves like this is deciding if they are a good sente exchange (because now black can't capture you / start the fight too much advantage with one move; lower the temperature in sente) or actually just made a bigger heavier target. With KG's example in this game, we can see the m13 jump as the standard fancy shape idea of making miai of m15 and o13, but actually they aren't quite miai and m15 (or thereabouts) is more valuable, so if white can get that in sente before s15 that's a nice exchange for white.

Uberdude · Post by **Uberdude** » Fri Jun 26, 2020 11:51 am

@John, is there a Japanese term for this sort of move, which makes some stones (usually cutting ones) a bit stronger / not cleanly capturable (eg net on 5 below doesn't quite work with a famous tesuji to break out) / annoying aji for opponent in sente before you take the important gote move? It does't really fit into my understanding of "kikashi before defending" as I see kikashi as a more incidental exchange of uncertain future benefit, whereas these sorts of exchanges are made for the immediate and clear benefit of reducing how good the opponent's next move in the area will be because that's where play is expected to continue afterwards. It's an important enough concept I think it needs a catchier name!

John Fairbairn · Post by **John Fairbairn** » Fri Jun 26, 2020 4:38 pm

, is there a Japanese term for this sort of move,

I'm not 100% certain I know what you are referring to - one swallow doesn't make a summer and all that - but with that caveat it looks like you're talking about kaname ishi - linchpin stones. They are contrasted with kasui shi - dregs stone.

The contrast most often comes backed up by a pair of provers: Kaname ishi wa suterubekarazu - Don't give up linchpin stones, and Kasu ishi nigerubekarazu - Don't run away with dregs stones.

Kaname ishi are reasonably common in go commentaries, but tend not be classed as a technical word (i.e. requiring a definition) simply because the term is part of ordinary Japanese (meaning a keystone or corner stone). Kaname itself means pivot, and so you could say "pivot stones" but that runs the risk of treating only cutting stones as kaname ishi. That is certainly a common application, but usage can be much broader. It refers to stones that may look heavy and therefore disposable but which are not in fact disposable. The reasons for not treating them as disposable can be rather subtle and on a grand scale. I've noticed AI bots make lots of kaname ishi, and I've noticed that simply because many of these moves have surprised me.

In one way I really shouldn't be surprised. One of the most sublime examples is the very famous three-stone shoulder hit by Go Seigen in Game 1 of the Kamakura match. There, the stones were not disposable because they were causing too much of a nuisance to Kitani, who felt obliged to capture them - but couldn't, as Go had observed, it takes too many moves to do that. Go, of course, had also realised that they stones were not as heavy as they might have looked, because leading them to safety did not cause collateral damage (these three stones were also a sublime example of the proverb 'five alive').

But I am surprised in another way because I have only seen very few examples (maybe one, maybe two) of the level of subtlety of the GSG example - until AI came along, and now, as I say, it seems very common.

Of course, in amateur play kaname ishi are always very common. But that is only because most amateurs don't distinguish between kaname ishi and kasu ishi and try to save everything. Still, in their defence, it's not easy to distinguish. My impression is that good amateur dan players often go the other way and treat kaname ishi as kasu ishi - being a bit too clever for their own good, perhaps. Sacrificing can feed the ego so well!

In short, I agree with you - it's a good topic for study. I can't think of any books on the topic, but you'll find examples in some of the more advanced or whimsical proverbs books - and of course in commentaries. If I remember (and if I've understood you correctly), I'll make a note of examples I see.

Bill Spight · Post by **Bill Spight** » Fri Jun 26, 2020 7:28 pm

John Fairbairn wrote:
, is there a Japanese term for this sort of move,
I'm not 100% certain I know what you are referring to - one swallow doesn't make a summer and all that - but with that caveat it looks like you're talking about kaname ishi - linchpin stones. They are contrasted with kasui shi - dregs stone.

The contrast most often comes backed up by a pair of provers: Kaname ishi wa suterubekarazu - Don't give up linchpin stones, and Kasu ishi nigerubekarazu - Don't run away with dregs stones.

Kaname ishi are reasonably common in go commentaries, but tend not be classed as a technical word (i.e. requiring a definition) simply because the term is part of ordinary Japanese (meaning a keystone or corner stone). Kaname itself means pivot, and so you could say "pivot stones" but that runs the risk of treating only cutting stones as kaname ishi. That is certainly a common application, but usage can be much broader. It refers to stones that may look heavy and therefore disposable but which are not in fact disposable. The reasons for not treating them as disposable can be rather subtle and on a grand scale. I've noticed AI bots make lots of kaname ishi, and I've noticed that simply because many of these moves have surprised me.

In one way I really shouldn't be surprised. One of the most sublime examples is the very famous three-stone shoulder hit by Go Seigen in Game 1 of the Kamakura match. There, the stones were not disposable because they were causing too much of a nuisance to Kitani, who felt obliged to capture them - but couldn't, as Go had observed, it takes too many moves to do that. Go, of course, had also realised that they stones were not as heavy as they might have looked, because leading them to safety did not cause collateral damage (these three stones were also a sublime example of the proverb 'five alive').

They do seem to be three sirens that led Kitani to his doom, at least, according to Elf. Not counting

, which was a joseki reply to the two space high kakari,

, all of the Black stones played near them up to

were costly, assuming komi, which the game was not.

cost 9%, but seems reasonable with no komi

cost 14½%, and always seemed dubious to me, but who am I?

cost 9½%, but also seems reasonable with no komi

cost 10%. It is at L-17, Elf likes M-16, after invading the bottom right corner.

cost 28%,

, a real blunder. It is at P-14, Elf prefers P-12, again after invading the bottom right corner.

cost 8%. Elf prefers to invade the bottom right corner, and not to come back any time soon.

These costs are by comparison with Elf's top choices, and assume no komi, but they total almost 80%. That's a lot for 6 plays.

Click Here To Show Diagram Code: [go]$$Bcm9 Go Seigen (W) vs. Kitani, Moves 9 to 18 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . 8 . 7 . 9 . . 5 X . . . . | $$ | . . . , . . . . . , . a . O 1 , X . . | $$ | . . . . . . . . . . . . . 2 3 . . . . | $$ | . . 6 . . . . . . . . . . 4 . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . 0 , . X . . . , . . . . . O . . . | $$ | . . . X . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Click Here To Show Diagram Code: [go]$$Bcm19 Moves 19 to 28 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . O . X . X . . X X . . . . | $$ | . . . , . . . . . , . . . O X , X . . | $$ | . . . . . . . . . . . . . O X . . . . | $$ | . . O . . . . . . . . . . O 3 . . . . | $$ | . . . . . . . . . . . . 0 5 . . . . . | $$ | . . . . . . . . . . . . . . b . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . , . . . . . , . . . . . , 4 . . | $$ | . . 2 . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . 6 . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . O , . X . . . , . . . . . O . . . | $$ | . . 1 X . . . . . 7 . . 9 . 8 . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Click Here To Show Diagram Code: [go]$$Bcm29 Moves 29 to 38 $$ --------------------------------------- $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . O . . O . X . X . . X X . . . . | $$ | . . . , . . . . 5 4 . . . O X , X . . | $$ | . . . . . . . . . . . . . O X . . . . | $$ | . . O . . . . . . . . . . O X . . . . | $$ | . . . . . . . . . 8 . 6 O X 3 . . . . | $$ | . . . . . . . . . . . . 1 2 . . . . . | $$ | . . . . . . . . . . . . 7 . . . . . . | $$ | . . . , . . . . . , . . . . . , O . . | $$ | . . O . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . O . . | $$ | . . . O . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . 9 O , . X . . 0 , . . . . . O . . . | $$ | . . X X . . . . . X . . X . O . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ---------------------------------------[/go]

Harleqin · Post by **Harleqin** » Sun Jun 28, 2020 5:28 am

I remember looking at that game several times, and each time not understanding the significance of those three stones. For me, they seemed like kikashi: Black answered below, so there is no loss in throwing them away. Definitely not key stones. I always thought that maybe there is a lesson here that I failed to understand.

Now with bot analysis, one might be tempted to say that I was right, but I do not think that it is so easy. There must be some merit to the moves Kitani made, it's just that other aspects dominate at that point in the game. One thing to notice is maybe that White turning where Black “crawled” would have quite an impact on the development on the right side. Maybe it is interesting to take a look at what happens in that area, and when, if the bots have their way.

Life In 19x19

Opening problems for AI: Problem 12

Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12

Re: Opening problems for AI: Problem 12