I came across something today that seemed to relate, in one way or another, to several threads here. One is the thread on moyos, another is a thread that mentions some equipment failure (too few stones?), and of course lots of threads that mention thickness and AI. It is also connected with me re-reading my favourite play, Le Bourgeois Gentilhomme by Molière, and with the fact that, today, I was doing a dance called the Highland Rambler, also nicknamed the Highland Wobbler.

The context is that one of the great go journalists, Akiyama Kenji (born 1946), was talking about go stones being too thick. Honinbo III Doetsu (1636 ~ 1727; roughly contemporary with Molière, as it happens) supposedly set the first standard for go-board size - together with Itagaki Yusen, it should be said. The cynics among us may choose to believe Yusen did the work and Doetsu took the credit. Either way, they set the thickness at 3.75 sun. Sun is 1.2 inches, which my ruler tells me is about 11 cm (whatever they are).

As time passed, the thickness crept up. In middle Edo, the standard thickness reached 4.8 sun. In the Meiji era, 5 sun became the norm. At present 6 sun is a lower limit. To some extent, this may be connected with increasing average height of the players, but probably it also had a good deal to do with showing off wealth. At any rate, as boards got thicker, so stones got thicker, just as, no doubt, the waists of the Japanese equivalents of M. Jourdain got thicker. But this, apparently, did not please go professionals.

Akiyama recalled Game 1 of the 50th Tengen title-match in 1980. This story was not reported in Go World so will be new to most of you. The game was played in a VERY posh hotel called Kagaya in Wakura Spa in Ishikawa Prefecture, in a room once occupied by Emperor Hirohito. What was considered a fitting go board and (apparently) thick was therefore procured the day before, by a very rich patron, from a local craftsman. Everyone agreed it was a masterpiece.

Kato Masao (the holder) was Black and played his first move at the upper-right star point. Challenger Yamabe Toshiro countered with White 2 at the diagonally opposite star point. At least he thought he did. The stone moved a fraction. There had been an earthquake the day before on the Noto Peninsula, and he seems to have assumed there were still minor earth tremors. So he tried again, but again the stone wobbled away. And again. And again. Eventually, the earthquake was discounted and everyone turned their scrutiny to the board and discovered that the lacquer was too proud at certain points. Because the purchase had been so recent, no-one had checked the board beforehand. Akiyama didn't say what they did about that, but from the context it seems they switched to some thinner stones.

There is something else to say about that game, but before that, I will mention another tale by Akiyama from a Meijin title-match ten years later, held in Miyazaki Prefecture, part of which was called Hyuga in Edo times. That area is famous for its kaya boards and "Hyuga White" hamaguri clam stones. Because it's a big local industry, competition among manufacturers is intense, but a new board (kaya, of course) was settled on first. Then came the questions of stones. The board-maker wanted the very best stones, the thick ones, to be sued with his magnificent board. But prudence too over and so the go officials got him to bring along three sets for the plsyers to choose from. Sure enough, the players chose the thin ones.

Now back to the Tengen game. Step though first to move 11 and then guess where Yamabe played White 12.



Yes, an early-early 3-3. Take that, AlphaGo, which takes the credit for innovation just like Doetsu (probably) overshadowing Yusen. Remarkably, katago seems to say this is the best move. More remarkably, when the AlphaGo 3-3 hit the headlines, one conclusion drawn was that pros had assumed, wrongly it was said, that they had to finish off with the hanetsugi (White 20 and 22 here) but was a mistake. Well, Yamabe played taht here and my KaTrain says that was correct!

In the post-mortem Yamabe (who got a good game initially but lost) asked whether White 12 was too early? Kato thought that perhaps it was, but did not criticise it. Yamabe did, however, reveal that he played the 3-3 mainly because he didn't know what to do in the lower right.

Now the next interesting feature of this game. Step through to move 53 and you will see the mother of all moyos - big enough to two sumo wrestlers, said Yamabe. How do you play against it. Bear in mind that moyo maestro Takemiya says the ideal way to play WITH a moyo is to invite the opponent in let him live small. But, actually, both players agreed that Black 53 was a mistake. Black should have played at J11.

To go from one extreme to the other, and thus to show how rich go is, take a look at another game from the Kagaya venue (which seems nowadays to have lost favour with go organisers - too up-market?). This is a Gosei title-match in 1986, with Cho Chikun as Black and Otake Hideo as White. It can be seen as relevant to the moyo discussion, but first step through to White 28 and what do you see?

But how thick were the "thin" stones? Or the "thick" ones?

"Kato Masao (the holder) was Black and played his first move at the upper-right star point. Challenger Yamabe Toshiro countered with White 2 at the diagonally opposite star point."

And now, according to the thread title, I expected, of course, 3-3 under 4-4, which, according to Katago, is the best move.

Cho Chikun appears to repeat this third line theme occasionally. I posted about his game with Xie Yimin.

The theme was only broken in this diagram and looks like white is inviting himself to black's moyo. This time Cho Chikun lost but I recall that the score evaluation wasn't that big of a deal yet when he resigned. I checked just now, initially it is B+5 but then it drops to B+4.5. The final position does look like black has won when it is a long game between top pros but maybe it wasn't as easy as it looked.

Click Here To Show Diagram Code

[go]$$B
$$ +--------------------------------------+
$$ | . . . . O X . . . . . . . . . . . . .|
$$ | . . O O O . X . . . . . . . . . O . .|
$$ | . . O X O O X . O O . O . . O . O X .|
$$ | . O . X X X . . X , . . . . . X X . .|
$$ | . . . X . . . . X . . . . . . . . . .|
$$ | . . X . . . . . . . . . . . . X . . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . , . O . . . , . . . . . , . . .|
$$ | . . . X . . . . . . . . . . . . X . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . X . . . . . . . . . . . . . X . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . , . . . . . , . . . . . , . . .|
$$ | . . O . O . . . . . . . . . O . O . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ | . . . . . . . . . . . . . . . . . . .|
$$ +--------------------------------------+[/go]

I recall that in "My views on Go" he writes about Kitani Minoru something to the effect that no one can mimic Kitani's style, he has tried and failed himself, and that Kitani played much on the third line. Maybe this is the motivation behind the 3rd line theme

Game record:

I recall some post game/post match interview with Kato where he talked about the stress of playing Yamabe - because the interview featured some unusually strong/candid views. IIRC Kato said something like Yamabe would play moves 'out of spite' that had an unclear outcome for both players and, despite the one sided result of the match, the stress on him (Kato) was almost unbearable.

There may almost have been a psychological factor at play as well in that I think Kato had been in a relative slump (for him) the previous year, and was extra keen to demonstrate his return to form.

My recollection (perhaps faulty) from the 1970s-80s go books I learned from (Elementary Go Series etc.) is that I was told that the full 3-3 invasion (ending with the hane and connect) is reasonable once Black has played on both nearby sides (K16 and Q10). The one example I could quickly find is from "The Power of the Star-Point" by Takagawa. He gives a sanrensei example where Black has not played K16, and says the 3-3 invasion is premature, then gives an example where Black has, and says that at this point the 3-3 invasion is playable because K16 would prefer to be one point to the left.

Anyway, the subtleties are beyond me, but when I saw the 3-3 invasion played out here it didn't strike me as being at odds with either 1980-era Japanese or modern-era AI theory. (Of course here one stone is on R9 instead of Q10; I'm not qualified to assess how much of a difference that makes.)

I had the exact reaction when reading John's post - I could clearly remember "from the old days" that extending on both sides from 4x4 is inviting a 3x3 invasion, as a common-sense thing. Except that I couldn't remember at all the source of this, I would have guessed some Go World game commentary.

Kato and Yamabe comment on their game in Go World 24, March - April 1981 [translated from Ito Club)

Instead of the 3-3 invasion at 12, Kato suggests capping move 5 at O9.

Incidental additional detail includes that Yamabe won the right to challenge when his opponent, Ushinohama Satsuo of the Kansai Ki-in, went to sleep in his room in the evening break and let his time run out.

There was a game involving Ishida Yoshio, maybe in a Honinbo title match, in which Ishida played a 3-3 invasion very early. I saw the game record in Go World (Ishi Press) magazine. If I recall correctly the commentator opined that Ishida played it as a yosu miru

In Kawabata's novel Meijin or in English The Master of Go, there are some remarks on the thickness of the white stones used in the game. In my copy of Seidensticker's translation it is towards the end of chapter 9, page 36. It is on page 29 in my Japanese edition. One of the observers commented on how thick the white stones are, and the thickness is given in the Japanese edition as 3 bu and 6 to 8 rin. In modern Western units, the stones were about one centimeter thick.

Like others above have mentioned, a 3-3 invasion when there are two side extensions from the star point, has been conventional wisdom in the amateur world around me, the reason being that one of these two extensions will be made inefficient. The novelty of AlphaGo and later AI was to play the 3-3 invasion even when there are no such extensions. The novelty included omitting the hane & connect and therefore not allowing easy eye shape to the outside wall resulting.

So I would not call it an "early 3-3": yes, it comes as early as move 12 but that's not the point. The earliness is about playing it before the sides.

In spite of the conceptual understanding, I still feel more comfortable with the "old" 3-3 invasion than the new "early" one. When the opponent makes 2 side extensions, 3-3 is almost natural to me. An early 3-3 still feels awkward and forced.

That may be conventional wisdom in some amateur world, though even there it sounds like wilfully ignoring the converse: one extension will be made efficient.

That precise point applied in the professional teaching of those days. In Vital Points of Go, Takagawa gives the following position (page 130, 1964 edition) minus the letters, and asks "How should White play when Black makes a wide extension to R11?"

Click Here To Show Diagram Code

[go]$$W
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . X . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . b . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . X a . . |
$$ | . . . . . . . . . X . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ----------------------------------------|[/go]

The answer given is that in this case it is good play to attack with White 'a'. Takagawa goes on to say that "It would be wrong to play White 1 at the three-three point, since Black would inevitably answer with Black 2 at 'a', thus raising the Black extension at R11 to its maximum value." The recommended move is White 1 at 'b'.

Even in the British amateur world, this philosophy can be seen in Matthew Macfadyen's concept of virtual territory. When you make a moyo (i.e. virtual territory), you should only expect to turn half of it into real territory. In other words, the value of a moyo can be expressed as actual territory = virtual territory / 2. But is it good play to go along with the moyo holder's expectations? You would normally want to ruffle his feathers at least a bit.

I acknowledge that AI disagrees with Takagawa, but at the time of the game shown, Takagawa and his fellow pros apparently did not normally accept that. And even if we do know better now, I'm not at all sure any amateur really understands why. Or even most pros?

Agreed John, when one extension is wider than the other, the invader should try spoiling the wider side and the attachment shown would come to my amateur mind as well.

This and some other similar positions are interesting. Approaching was the usual move but if you check game databases then the 3-3 has much better win rate.

Click Here To Show Diagram Code

[go]$$B
$$ ------------------------+
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . X . . . . . . a . .|
$$ . . , . . . . . X . . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . b . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . , . . . . . , . . .|
$$ . . . . . . . . . X . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|[/go]

a is 29 / 75 or 39%
b is 9 / 20 or 45%

Click Here To Show Diagram Code

[go]$$B
$$ ------------------------+
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . a . .|
$$ . . X . . . . . X . . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . b . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|
$$ . . , . . . . . , . . .|
$$ . . . . . . . . . X . .|
$$ . . . . . . . . . . . .|
$$ . . . . . . . . . . . .|[/go]

a is 61 / 142 or 43%
b is 36 / 72 or 50%

This second position is much more common, it occurs in the Chinese opening. There are various whole board positions mixed into the statistics above.

I suspect that b was preferred by players who thought that they could fight better than their opponents. In later years there are mostly moves at b despite the lopsided win rate. For example Gu Li and Cho U played b.

The argument that the 3-3 makes one stone inefficient and the other efficient doesn't sit right with me. The way I see it is that black is already a bit inefficient, say 70%, and when the 3-3 comes one stones becomes normal and the other becomes worse. Maybe that comes out less than 70% total? If white approaches then both of black's stones could become efficient, say above 80%. I think at least one of them does become more efficient and the other could stay about the same, let's say one is above 80% and the other stays at 70%, that is at least 75% efficient.

Saying a stone is this and this efficient is a bit of a game of numbers. However, if some stone is so misplaced as to inefficient, how can we expect to make up for that by simply playing normal moves? To put it differently. If every normal move is the same 80% efficiency but we find a 70% move it is rather optimistic to think that we have more than 90% efficient move somewhere else to get back to our baseline.

It becomes more difficult to get an advantage at higher levels and that probably has huge effect of what is played in pro games. Maybe Gu Li and Cho U were optimists by nature but probably they were trying to take the game into situations were they believed they had chances to outplay their opponents. This is probably necessary in order to have a high win rate at a level were everyone plays very well. If players at that level simply played the easiest 80% move all the time their opponents would know how to match that and the stronger player wouldn't get the same advantage as when he plays more speculative.

For the rest of us it isn't that simple. If the opponent can't match our baseline efficiency then we would appear to have an easy way to win, at least we could wait for them to gamble. On the other hand if we can't match the opponent's baseline, then what to do? Should we gamble that we can sometimes outplay them in other ways or accept that we may lose many games while we learn to play more efficiently?