Life In 19x19

Did you notice if Elf often wanted to play 12 at 14 instead in the joseki? I saw that once whilst reviewing. That's usually regarded as a mistake as it gives white s2 in sente which is pretty huge yose (or s9 in sente can help with an invasion on that side), but the plus is white doesn't get the p6 atari for free. The joseki usually regards that as less valuable than s2 in sente (as you can still p5 cut or n6 shape point as future moyo border / shape point), but in that case Elf disagreed.

Uberdude wrote:Did you notice if Elf often wanted to play 12 at 14 instead in the joseki?

Yes! I don't have smartgo and my little review database to hand, but I think Elf thought that this was a better idea in about 3/4 (edit: about 95%) of the pro game board positions:
(I'd be interested to see if I can identify what influences this)

Click Here To Show Diagram Code

[go]$$Wcm11 Elf misses a,b exchange
$$ . . . . . . . d . |
$$ . . . . . . . . . |
$$ . . . . . 2 X a . |
$$ . . . . b X O O . |
$$ . . . . . O X X . |
$$ . . . . . O X . . |
$$ . . . . 1 X O X . |
$$ . . . . . . O c . |
$$ . . . . . . . . . |
$$ ----------------[/go]

This reminds me of the way Ai seem to like to capture laddered stones with a big loose centre oriented move regardless of the aji. Maybe its okay to play this way if you back every move with 10k playouts. Kageyama would be furious!

To me it seems like 'c' becomes a pretty huge double sente move, but white doesn't seem to take it right away. I think white in the human variation looks thick enough that humans wouldn't like to play too near except for that shape point maybe, but since the wall doesn't quite have 2 eyes I would think that Ai might want to deny it getting the turn in sente in order to slightly threaten the wall at some point?

I actually think the really interesting stuff happens at move 6, I just haven't got my head around all the variations yet!

I'm thinking about this wrong. Elf seems to think that you can give white the sente turn at the 2-2 point and still not regard it as particularly thick.

These are the next local moves that appear in quite a few of Elf's follow up sequences:

Click Here To Show Diagram Code

[go]$$B Elf sees no thickness
$$ . . . . . . . . . |
$$ . . . . . . . . . |
$$ . . . . . X X . . |
$$ . . 4 a . X O O . |
$$ . b 3 2 . O X X . |
$$ . . . . . O X . . |
$$ . . 1 . O X O X . |
$$ . . . . . . O . . |
$$ . . . . . . . . . |
$$ ----------------[/go]

Black follows up with a or b, sometimes white seems to make the 2-2 exchange about now (I'm still surprised it's not sooner.)

As a curious (to me) tangent, I spotted that If white makes the 2-2 exchange and black has a isolated 4-4 stone in the corner where his thickness is pointing Elf likes to make this second line approach:

Click Here To Show Diagram Code

[go]$$W Elf does respect some thickness
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . O . . |
$$ | . . . X . . . . . , . . X X O O X . . |
$$ | . . . . . . . . . . . . X O X X O O . |
$$ | . . . . 1 . . . . . . . X O X . X O . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

From what I can tell, humans use this approach in pretty special circumstances, often much later in a game's development.
(This is in follow up sequences so the reading applied to each move isn't too great. Maybe take with a pinch of salt)

MikeKyle wrote: I actually think the really interesting stuff happens at move 6, I just haven't got my head around all the variations yet!

So.. from here:

Click Here To Show Diagram Code

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

Humans play a about 2/3 of the time, leading to the familiar patern. Humans occasionally try b or c, to the extent that there are a lot of paterns that have appeared frequently enough that I would call joseki.

Elf likes these rare human paterns starting with b and c.

It thinks that humans are wrong to choose a so frequently, and the more common choice here should be b with paterns like:

Click Here To Show Diagram Code

[go]$$B Elf's favourite from here
$$ . . . . b . a . . |
$$ . . . . . . . . . |
$$ . . . . . 4 X . . |
$$ . . . . . . O 2 . |
$$ . . . . . O X . . |
$$ . . . . . O X . . |
$$ . . . . . X . 1 . |
$$ . . . . . . 3 . . |
$$ . . . . . . . . . |
$$ ----------------[/go]

black a, white b is a common follow up

Click Here To Show Diagram Code

[go]$$B when white wants sente
$$ . . . . . . . . . |
$$ . . . . . . . . . |
$$ . . . . . 2 X 5 . |
$$ . . . . . 4 O 3 . |
$$ . . . . . O X . . |
$$ . . . . . O X . . |
$$ . . . . 6 X . 1 . |
$$ . . . . . . 7 . . |
$$ . . . . . . . . . |
$$ ----------------[/go]

Click Here To Show Diagram Code

[go]$$B Lader dependant
$$ . . . . . . . . . |
$$ . . . . . . 8 . . |
$$ . . . . . 0 X 6 . |
$$ . . . . . 4 O 3 . |
$$ . . . . . O X 7 . |
$$ . . . . . O X . . |
$$ . . . 9 . X O 1 . |
$$ . . . . . 5 2 . . |
$$ . . . . . . . . . |
$$ ----------------[/go]

Or c with paterns like:

Click Here To Show Diagram Code

[go]$$B black could omit the 7, 8 exchange
$$ . . . . . . . . . |
$$ . . . . . . . . . |
$$ . . . . . . X . . |
$$ . . . . . 2 O 1 . |
$$ . . . . . O X . . |
$$ . . . 6 . O X 3 . |
$$ . . 7 . 4 X O . . |
$$ . . . . 8 5 9 . . |
$$ . . . . . . . . . |
$$ ----------------[/go]

These all seem to be very rare human paterns, but Elf likes them more than the main patern.

Incidentally in the main patern, even though Elf doesn't consider white particularly thick, Elf tends to agree with white's move at every step. It seems to me that Elf thinks that approaching and being allowed to settle in sente is good for white in this case, and I'd be interested to test if Elf always feels this way about human joseki where this happens.