Knotwilg's practice

[Shaddy]

I suggest getting better at fighting. Without implying that you should focus on this, I noticed when I became 3d that most 2d don't get much out of their attacks, and 3d get a lot more.

[Uberdude]

I forget, have you ever taken regular lessons from a good teacher e.g Hwang Inseong?

[Knotwilg]

I forget, have you ever taken regular lessons from a good teacher e.g Hwang Inseong?

A long time ago I studied with Guo Juan from the Netherlands.

More recently but still a long time ago I studied with Minue until he disappeared.

I'm not terribly keen on having a teacher right now but if you think it's necessary I'll consider it.

[Knotwilg]

Here's a game of today. I was happy with the collected way of playing.

I lost the game late in the middle game or early endgame, playing "honte" where I should have taken a big point and then later losing sente.

It shows errors in judgment, not so much in tactics.

[Bill Spight]

A few comments and questions.



BTW, I had to remove the [] around 2d and 1d.

[ez4u]

A few comments and questions.
...
(Regarding )
*** Why not approach the bottom left corner, like everybody else?

Who would everyone else be? In my copy of GoGoD, there are only 5 examples of the approach to the bottom left corner in this position. Three are from AlphaGo self-play games (all won by White). But this is a relatively rare position currently. In the AI era, Black's high approach in the upper left has been replaced with the low approach as the "main line". Beyond that, the hanging connection for (with or without an intervening approach at F3 to test the water) has been the main choice both before and after the AI's appeared.

Once Black connects solidly, the human choice for has historically been D10 (high extension rather than the low one in the game). This D10 is also blue for the the Katago 20-block at around 15K playouts and remains unchanged at 500K (i.e., when I leave it running over lunch ). Katago is basically indifferent between the lower left approach and tenuki to enclose the lower right.

[Bill Spight]

A few comments and questions.
...
(Regarding )
*** Why not approach the bottom left corner, like everybody else?

Who would everyone else be? In my copy of GoGoD, there are only 5 examples of the approach to the bottom left corner in this position. Three are from AlphaGo self-play games (all won by White).

Just a memory glitch. What was popular was an approach to the bottom left corner for , with the connection at . OC, with an approach to the bottom left corner in this game the play can transpose to those once popular positions.

Once Black connects solidly, the human choice for has historically been D10 (high extension rather than the low one in the game). This D10 is also blue for the the Katago 20-block at around 15K playouts and remains unchanged at 500K (i.e., when I leave it running over lunch ). Katago is basically indifferent between the lower left approach and tenuki to enclose the lower right.

Checking now, I see that AlphaGoTeach prefers the approach at C-06 slightly (by 1.2%) to the high extension to D-10, and decidedly (by 3.1%) to the low extension to C-10. Interesting that KataGo disagrees. What are the playouts and winrates for each of those moves at 500K? Thanks.

[Shaddy]

how about 43 @ r9? the extra liberty matters a lot, and i don't really feel like connecting underneath

[Uberdude]

Looking at the last game and trying to extra general principles for improvement, I would say:
- think more about maximizing local efficiency of shapes, don't just settle for a move which does the broad strategic job you want but look for more candidate moves that may do so in slightly better ways. Finesse in move order/timing as taught by book "Beyond Forcing Moves" could help.
- when fighting think about where the points for you and opponent will and will not be when the dust settles.

[Knotwilg]

Thanks all, more than I hoped for!

Bill, I see that you've already converted to more modern play. I thought many of my choices in the opening were a little old fashioned and slow but not wrong, not by Katago's account. The real mistakes were rather late in the game.

Dave, I'll investigate the high extension and may convert to it.

Shaddy, yes 43 would have been simpler at P10

Uberdude, SoDeSuNe, jlt, Shaddy, based on your general recommendations I will

- read, read, read: in my games, when doing tsumego and when studying joseki. A current goal is to read at least 10 moves deep in at least 1 variation.
- embrace fighting and apply reading there; at least for a while stop trying to hold back on fighting, as it seems to be in my nature
- take Uberdude's last recommendation to heart; rephrasing: look for the best shape that globally achieves your goal; look where the opponent may play when the dust settles
- and consider taking a teacher, but not right away

[Knotwilg]

So far I'm doing tsumego, studying joseki and playing games.
The key challenge is to read, read, read. Not going through the motions, read the solutions, check joseki ... but read.
In the games I embrace fighting and it seems to pay off.

[SoDesuNe]

I also read an insei report from Antti (I couldn't find this post anymore, so maybe it was some other report on senseis, Benjamin Teuber?, or the other parts of the internet) that he was often ahead in the opening because the other inseis had a way worse understanding of this stage then he had. But their reading was far stronger and so they often catched up.

Just if anyone was interested, I found it again: It was a testimonial for Guo Juan's go study in China (https://internetgoschool.com/ruben.vhtml).

[Knotwilg]

Since taking this new approach, I'm 6-3 on KGS. I'm doing daily tsumego and about 90% of the time I can keep the discipline of reading out the full problem, including variations that don't work even if I already found the solution. I've studied a number of corner patterns. I'm reading 5-10 moves deep and 3-5 variations. Especially in corner patterns I'm able to read fairly deeply. In tougher tsumego I have harder time picking the candidates, for example in carpenter square variations. In the games I'm embracing fights and those have turned mostly into my favor, even if they weren't always well executed but the opponents made more crucial mistakes than me.

[SoDesuNe]

[Knotwilg]

Stabilizing at 2d on KGS, lost by 4,5. The review was interesting to me.