The current computer-assisted research on the 3-3 invasion. Shown by my (and many people’s) favourite WeiqiTV pro Meng Tailing 6p in a commentary of a game against fresh Korean 1p, in the 2nd round of the Samsung Cup preliminaries.
I basically screenshotted and transcribed loosely what he said in the first few minutes of this video. In the following, first-person pronouns refer to Meng Tailing.
(Game) White extended 8 à la Master. Against the slide 10, I think the best for Black is tenuki. I prepared the invasion 11.
Though I’m not sure whether it’s better to omit 9-10, it’s certainly fiercer…
(Var.) …because now White does not have the option to block from this side.
White extending towards 7 is not meaningful and after splitting at 10 I think I’m good.
(Game) So White blocked the other side. When White gets a wall to extend at m17, then the 9-10 exchange would be a minus.
In the game White chose the keima 14.
(Var.) If he plays nobi at 1, jumping out at 2 is a good move.
(Var.) If he plays the hane 5, I would push up at 6. Nobi at 7 would be too slow, but White does not seem to have a good move otherwise.
(Var.) Therefore in AlphaGo vs. AlphaGo games White usually turns at 5 to cut at 7 and nose at 9.
(Var.) And this shape emerges. White gets to approach at 13.
I think this is the best option for White. If I were White, I would play like this.
Black defends r17 and White can jump m15. Maybe it’s not good for Black and Black should extend on the upper side instead of 12.
(Var.) But where to extend is not easy to decide.
First White is going to get a smooth atari at 13 which is annoying.
Then if Black goes two-space low, White can still extend.
(Var.) If Black goes high, White can invade immediately. Not easy to counter.
(Var.) …Could this be the move? This is really tight, not sure if it’s good.
In any case, White should settle the corner shape like this and see what Black does.
(Game) The keima 14 in the game is probably not the optimal joseki.
Attaching below is a usual response.
(Var.) I’m certainly not satisfied with bumping like this and letting White extend majestically.
(Var.) White dares not extending directly in the game because he was probably afraid of the hane-cut 2.
(Finally someone says what I have been suspecting for so long~)
(Var.) Once I get to shoulder-hit 10 in this fight, White is in trouble. The three stones are too heavy.
(Var.) So White is going to respond somehow locally. But if White retreats 1, I can hane up at 2. I don’t need to put the stone where my mouse is so Black gains a little.
(Var.) If White extends here, I’ll bump, crawl, and tenuki. It feels like White’s wall is restricted no matter where precisely Black extends.
(This is different from the AlphGo shape that already blocked the left side. Though there is no cutting point here, Black couldn’t have cut anyway.)
(Var.) And I’ll try to keima somehow when White invades.
So by restricting this wall, Black’s fuseki strategy is successful.
(Game) Thus, White played hane at 16, recently a popular move among humans. Black of course bumps 17.
(Var.) If White blocks etc. and plays hane at 5… should I take one stone or extend the upper side?
(Var.) Let’s say I take and let him approach 11. Locally White is not that great. So for now White does not feel like playing in this way.
(Game) As White chose to hane 16, he was certainly going to nobi 18.
However, in retrospect, this choice was questionable.
(Game) I pushed and cut and hane-connected. If the ladder was good for White, Black couldn’t play in this way. Black would have no continuation.
(Var.) For example if Black pushes 1, White taking at 2 feels like game kaput.
(Var.) Fortunately, the ladder was mine and Black has this wonderful keima 27. Now it’s become difficult for White.
We can couclude that White cannot choose this variation without a favourable ladder.
(Var.) If White takes one stone, Black squeezes and blocks and White is toast.
(Var.) There is nothing for White to do in the centre as White has no ladder…
(Var.) And Black wins the local semeai by one liberty.
(Maybe it’s not even related to the ladder because Black can hane in sente where my cursor is.)
(Var.) If White connects, Black can push once to keima down at 5. The four white stones are in trouble.
(Game) White can only atari 28 and connect 30 and allow the hane.
(Var.) Connecting 30 is necessary, or else White is dead.
(Game) So finally the keima 27 prevented White from taking at 35 and at the same time gave Black the hane 31.
White’s shape is horrible, he can only connect 32. Black simply connects 35 after throwing in above 34 (for later damezumari.)
(Game) Black got 10 points in White’s corner. So White has fewer points and the outside fight is still even.
Therefore, we can say that this result is slightly better for Black.
[The rest is omitted.]