From a dan players’ problem book, a position which Uberdude discusses in detail here.
Black to play.
As a 5k in 2021, I immediately saw A. This is the book solution, but Uberdude criticises it in the thread, supporting his position with a Leela Zero analysis.
After some time, though, I also came up with B–E, as well as the two Xes around B. I feel intuitively that D isn’t good, but after reading it there seemed to be worthwhile continuations. Leela liked B as well, he reports, but after more playouts it switched to the floaty, botty G.
I think I would’ve eventually played E, because it looks most interesting to me, appearing to carry some play against the six eastern stones.
The “incorrect solution” given by the book is F, but I feel that’s a strawman answer. Who would consider F?
To the defence of the suggested move at A: In the early days of AI, I noticed that even In-Seong Hwang 8d would usually assume that white responds with a push instead of a pincer. And top pros pushed similar shoulder hits many times in the AlphaGo vs the World 60-game series. So back in the day, the pushing “joseki” may have been actively taught to strong go students.
But nowadays, I guess that even stronger kyu players would pincer black A (or B) in this position, instead of pushing it out.
(Pre-bot) comments made by Yilun Yang in BGJ #158, 2010-11.
Here is the professional popularity of the different options according to Waltheri (admittedly including both 20th century games and ones made after this article’s publication).
Note also that Yang was only discussing local moves, so the approach statistics are not relevant to the article.
The high approach to the top right (37 games)
C (8 games)
B (8 games)
The game move (7 games)
H (5 games)
The two-space high approach to the top right (2 games)
E (1 game)
The low approach to the top right (1 game)
D (1 game)
Yang was correct that F and G are not classically good options, but who would ever play either of them in the first place?
A makes no sense to me in the situation that (4) is in its game location, since Black is unable to respond to White’s 3-3 invasion with a block towards (5). I don’t know what a player as strong as Francis Roads saw in the move (and clearly professionals never cared for it either.)
White 12 looks slow in that variation.
White has several (better) options there, but black should be able to handle them. I have played this joseki variation a lot in the past (as black and white):
Yes, this was fairly normal in the later part of the 20th century (at least in games up to my level).
It could be said that the result is slowish for black (black ends in gote and white has a fairly large corner), but black has a choice of aji around K16-L15 that threaten to capture the F16 stones (sometimes in a ladder towards the lower right), which could work nicely for moyo building. On the other hand, white has aji around C13.
Also, Waltheri has 38 instances of the A pincer with the inside block actually in a sanrensei (mainly from the '50s, '60s and '70s but with outliers extending to 2011).
Yes, Francis likes the joseki Gennan posted. I’m quite happy as white to get a big territory in sente with plenty of aji to mess with the moyo later like c13 or using the g15 stone. Iirc I missed one of those moves on the top side which are sneaky sente for h18 many years ago.
Yes, that’s also possible. The exact end result is white’s choice about what kind of aji they prefer for later. The variation I showed leaves more tricks afterwards, so I might prefer that result when playing against a weaker player. I might prefer Cho Chikun’s variation against a stronger player.
Here’s a game I played against Francis (black) with him playing this joseki with a san sen sei, from the 2011 Oxford tournament. Move 46 was a blunder when using the invasion aji.
Here’s me missing the sneaky sente in this joseki at Trigantius 2008. The best way to learn joseki traps is to lose tournament games because of them .
And a classic san ren sei with him using double Takemiya’s kosumi, from British open 2013 (I’m quite a lot stronger than him by now):
I found it interesting to analyse this position in which I collapsed in the opening as Black.
I played a sanrensei and White invaded twice at the 3-3 points. A ko developed in the top left, and I exchanged A for B as a ko threat (White’s initial threat exchange was R13–Q14). I then took the ko.
However, what I missed is that it was imperative for me to take R18 immediately myself. After R18–S17, I still have another threat at S17 and I can thus win the ko. I had fallen victim to carelessness.
After I took the ko, White immediately played R18 himself and I was galled at having missed the opportunity, since I had a lot of time to think. Annoyed at myself for the mistake, I resolved to move quickly and approach the lower left, reasoning that I’d probably have sufficient threats in an avalanche shape that I’d be able to win the ko later.
However, it was best for me to simply finish the ko and clean up the aji, building a thick – even if undercut – position. This was hurriedness.
White then took back the ko. I couldn’t bear to defend with E19 since I had lost advantage elsewhere during the ko fight, so I ignored the corner and followed up my threat with D6. I was under the influence of pridefulness.
White then resolved by capturing at (5). I exhibited some stubbornness with (6), which could have been at O3, and White continued to chomp up my top-side formation with (7).
A few minutes later, White resigned because he apparently didn’t think I actually meant to play at the time control I set. The AI says I’m behind by eight points, which isn’t great, but I had every intention of continuing into the midgame.
In the game Black played the reckless translucent move and I was soon able to capture the centre group, despite us both having Class A levels of time.
With proper play by Black, the position should continue as a running fight. Black is ahead on cash but White has the superior combat stance, so the game is even.
Takemiya played a move which even I could tell was ridiculously slack. The commentary went something like “But what on earth was Takemiya thinking with Black 43?!?!?”
– Blue Wyvern, 2003, on SL’s page on diagonal moyo; referencing a 2002 game in the Judan title match against O Rissei
It seems like the natural play to me, since otherwise Black looks to be left very thin.
I think one of the things I really enjoy about accessible strong AI is how quickly they can (potentially) validate or thrash an idea.
Now of course there’s always the ideas of bot ideas vs human ideas, where a human might play intentionally slack (lose a point or two) to simplify the variations, whereas the bot will play an incredibly sharp sequence with deep reading.
Even still, and ~only dan level support of course,
.
