Life In 19x19

John Fairbairn wrote:

$$W
$$ -------------------
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . O O a . . . . .
$$ | . . . X X . . . . .
$$ | . O . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . b . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .

Click Here To Show Diagram Code

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

I haven't checked all joseki dictionaries, but I don't think AI is teaching the human pros anything here. Kitani says 'a' is the joseki move. 'b' is not labelled a joseki but is commonly used when you want to emphasise Black's low position. He also says 'b' is not a move you would necessarily want to pay at once.

He seems sniffy about the various kakaris (they strengthen Black too easily) but if one is needed he seems to prefer to the two two-space third-line kakaris (i.e. high and low), and of course in that case White is at least controlling the direction of play.

John, I think this is (another!) case where you, with your broader and deeper knowledge of non-English Go sources (Kitani's dictionary is 1930s?) have a wider view than what I would characterise as the "standard western Go theory" which is mostly based on Ishi press translated Japanese books from the 70s like Ishida's joseki dictionary, the Elementary Go series, Go World and then later recycled through Sensei's Library.

Checking Ishida's joseki dictionary volume 3 (1977) it has only 2 chapters on the 3-3 points and the first is about the shoulder hit, and second the 2 space high approach, with intro;

The shoulder-hit has traditionally been regarded as the only move in answer to the 3-3 move. It is certainly a powerful move which takes advantage of the lowness of the 3-3 stones, but recently there has been a gradual shift in opinion.

After showing the crawl and slide (with comment that shoulder hitter can jump rather than extend) the very first diagram is:

Click Here To Show Diagram Code

[go]$$W
$$ -------------------
$$ . . . . . . . . . . |
$$ . . . . . X . . . . |
$$ . . . . . . . X . . |
$$ . . . 1 . . O X . . |
$$ . . . . . . O a . . |
$$ . . . . . . . . 2 . |
$$ . . . . . . b . . . |
$$ . . . . . . 3 . . . |
$$ . . . . . . . . . . |[/go]

Dia 1. (the basic pattern) White continues by jumping to 1 or blocking at 'a' or playing elsewhere. White 1 is the most common. Up to 3 is the basic pattern. Black can also play elsewhere instead of 2. If white omits 3, 'b' is a good point for Black.

Dia 2 then shows black 2 attaching to left of 1 and taking top side and letting white turn, and Dia 3 is

Click Here To Show Diagram Code

[go]$$W
$$ -------------------
$$ . . . . . . . . . . |
$$ . . . . . X . . . . |
$$ . . . . . . . X . . |
$$ . . . . 2 . O X . . |
$$ . . . . . . O 1 . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ . . . . . . . . . . |
$$ . . . . . . . 3 . . |[/go]

Dia 3 (simple). Blocking at 1 is the most straightforward move. The sequence to 3 entails a slight compromise on White's part, but not to a degree worth worrying about. This result is equal. Playing at 2 can be regarded as absolutely necessary.

So the centre jumps of Dia 1 are claimed as normal and most common, with the turn slightly inferior but not worth worrying about. This judgement is different to my pre-AI knowledge (from Matthew Macfadyen) that the shoulder hit and jumps joseki is bad for white, and AI agrees, and also agrees the turn is better, not a slightly worse compromise.

So had things changed from Kitani's time with the shoulder hit growing in favour, and then starting to wane by the 70s, or is Ishida misrepresenting a wider variety of thought prevalent in Japan in the 70s? Or was Kitani's thinking back then unusual among standard pro thinking?

Takao Shinji's joseki dictionary from 2012 also leads with the shoulder hit and jumps as the main variation, so the "gradual shift in opinion" mentioned in Ishida doesn't seem to have lasted long enough to make Takao say something different. (Or did John Power just rehash Ishida's dictionary and Takao didn't have much input of new thinking )

Andrew: I may have missed the point completely, but I get the impression you are shifting the dichotomy from 'a' magari and 'b' two-space jump to one between initial shoulder-hit and kakari. Perfectly reasonable to do so, of course, and quite possibly a more important aspect, but it puts Kitani's view in limbo, I feel.

Despite that uncertainty, what we can be certain amount is that Kitani's view was one very widely shared, simply because the research he based his dictionary on was the work done in the 1930s by a rather large team (led by Suzuki Tamejiro) for the Great Joseki Dictionary. That meant it reflected no-komi go.

Later generations had the new consideration of komi, but it must also be remembered that komi went up in stages: 4.5, 5.5 and 6.5. It seems reasonable to assume that evaluations of all sorts of plays would gradually shift along with the change in komi. A further likely cause of change in opinion would be the increase in actual-play examples. Prior to the Suzuki/Kitani work, real-life examples were very rare. It was alleged that san-san was banned in the Honinbo school. That can't be true as even Shusaku played it, but it was certainly rare. Examples of 4-4 were much more common. 3-3 was well known in old Chinese games, though, so it was no surprise when Go Seigen played it (at least to him!).

A separate point to make is that too many westerners think 'joseki' means 'equal'. It can imply something similar in Japanese, but only in the sense of 'fair division of the spoils,' allowing for possible different number of stones played by each side. The underlying meaning is something played regularly, and played regularly because both sides feel they can be satisfied with their own results (i.e. again allowing for different number of stones or even the setups elsewhere on the board). What follows from that is that there may be a shift in opinion if what is being played elsewhere is also changing.

None of that, though, seems to me to be be altering the pros' fundamental thinking about things like suji, balance, timing etc. If I had to stick my neck out, I'd say the one aspect of AI play that is creating most problems for pros (apart from depth of reading, of course) is treating moves as probes. The 3-3 shoulder hit and the 3-3 invasion against 4-4 both fall into this category, I feel, but so do many other AI moves such as shoulder hits, or other plays where they seem to dart around randomly. I have come to this conclusion partly because I have become convinced that the commonest criticism by strong pros of play by weaker pros has always been a relative lack of probes. If that's correct, the relative plethora of probes in AI play seems a natural progression.

John Fairbairn wrote:Andrew: I may have missed the point completely, but I get the impression you are shifting the dichotomy from 'a' magari and 'b' two-space jump to one between initial shoulder-hit and kakari. Perfectly reasonable to do so, of course, and quite possibly a more important aspect, but it puts Kitani's view in limbo, I feel.

I suppose I am making 2 related points:
1) shoulder hit to 3-3 is special purpose joseki for when centre is important, in large majority of real board positions approaching from the side will be better. (Corners then sides then centre, ancient theory, which AI re-emphasises).
2) Given that you did the shoulder hit (which was maybe a mistake, but not too bad yet if you keep sente), it becomes a bigger mistake if you continue with the Dia 1 floaty centre gote "standard joseki" (per what I learnt from western go books directly and and recycled through teachers like Charles Matthews, Guo Juan, Sensei's Library, teachers on KGS brought up on a diet of the same) and you should almost always turn to switch from centre to side and this is locally better than the floaty jumps. (ditto Corners then sides then centre theory)

Given that you did the shoulder hit (which was maybe a mistake, but not too bad yet if you keep sente), it becomes a bigger mistake if you continue with the Dia 1 floaty centre gote "standard joseki" (per what I learnt from western go books directly and and recycled through teachers like Charles Matthews, Guo Juan, Sensei's Library, teachers on KGS brought up on a diet of the same) and you should almost always turn to switch from centre to side and this is locally better than the floaty jumps. (ditto Corners then sides then centre theory)

OK, understood now. But that is precisely what Kitani and other pros have said in the past, no?

However, it does then raise an interesting point in relation to another thread. Kitani was notorious for tighter moves than other pros (e.g. "I don't like knight's moves"). A compact magari can be seen as an exemplar of that style. So, was Kitani also implicitly making a statement about local efficiency (in the sense of that other thread)?

It certainly was usual in the 50's and early 60's to smack down a stone of top of the 3-3 and then tenuki. I think even then it was more common to ignore the 3-3 until in the middle game or end of the opening. Approach moves were also very common but this strategy of smacking down the 4-4 and tenuki instead of making extension was very common. As far as I understand the reason this way of playing lost favor quickly is that Go Seigen got seriously burned using it in games against Sakata.

http://ps.waltheri.net/database/game/37991/
http://ps.waltheri.net/database/game/38182/

The lesson from these games was that you may still get attacked after playing the 4-4 cap but previously it was opinion that the cap couldn't be attack effectively. I have heard this same thing in lectures on these Go Seigen games, lectures on 3-3 and also been told this by Japanese pro at an EGC (which also said they didn't really know if the conclusion or the reason was correct, just that everyone stopped using this way of playing).

The idea of tenuki and defending could go back to the Honinbo school because this is how Shusai played against Go http://ps.waltheri.net/database/game/41980/, but it could also be that this was its inception.

I think the play and choices in such games used to be treated with more gravitas than was warranted and the whole Go world would really change its course depending on what was arguably a choice between various viable options.

After the 60's it is probably much more likely to find games were the cap is played with intention of playing out the joseki, not playing anything now being strongly preferred over making the cap and leaving it. The number of games preserved also increases very quickly, making it possible that this cap and leave it was only played occasionally to harken back to the famous Shusai vs Go game (and people would take note of such games). This may have lost its appeal after Sakata's success.

Btw, I don't think the order the josekis appear in the Basic Joseki Dictionary is indicative of anything.