How to use an early approach to 3-4

[Gomoto]

I like to approach 3-4 early. I just wanted to keep book of how I deal with black responses.

(White approaches 3-4 at move 6.)

$$
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . W . . |
$$ | . . . O . . . . . , . . . . . , . . . |
$$ | . . . . . X . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+

Click Here To Show Diagram Code

[go]$$
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . W . . |
$$ | . . . O . . . . . , . . . . . , . . . |
$$ | . . . . . X . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

[Bill Spight]

Move 6?

Move 2, now that's early.

[sorin]

I like to approach 3-4 early. I just wanted to keep book of how I deal with black responses.
(White approaches 3-4 at move 6.)

In the SGF variations above, how did you decide when tenuki is possible and when it is not?

[Bill Spight]

I like to approach 3-4 early. I just wanted to keep book of how I deal with black responses.
(White approaches 3-4 at move 6.)

In the SGF variations above, how did you decide when tenuki is possible and when it is not?

Tenuki is always an option.

[Calvin Clark]

So you are preparing against a number of local responses to your approach, but not to black playing a double approach in the bottom left, which is the most common pro response? Your opponent may also tenuki.

[EdLee]

Hi Gomoto, Calvin,

For , what are the bots' top candidates ?

[Gomoto]

Calvin Clarke: So you are preparing against a number of local responses to your approach, but not to black playing a double approach in the bottom left, which is the most common pro response? Your opponent may also tenuki.

That is the basic idea. The approach to 3-4 is better than defending against the double approach. THe black double approach is what white hopes for:

[Gomoto]

Sorin: In the SGF variations above, how did you decide when tenuki is possible and when it is not?

I used ELF to evaluate the variations. (Not the first move only).

The sgf variations are just a simplified overview of my preferred choices. (Sometimes there are more options. But I prefer tenuki if possible.)

[Gomoto]

Bill: Move 6?

Move 2, now that's early.

I played like this in the past. But I can not support this style with the current AI opinions.

[Gomoto]

EdLee: Hi Gomoto, Calvin,

For , what are the bots' top candidates ?

[Bill Spight]

Bill: Move 6?

Move 2, now that's early.

I played like this in the past.

So did humans a few hundred years ago (with no komi). It went out of favor, but even when I was learning go Takagawa mentioned it as a possibility.

But I can not support this style with the current AI opinions.

Go Seigen in his 21st Century Go series recommends the 2 space high approach at move 2, even receiving komi.

[Bill Spight]

Hi Gomoto, Calvin,

For , what are the bots' top candidates ?

According to AlphaGoTeach ( https://alphagoteach.deepmind.com ), AlphaGo has played the keima in the bottom right corner (winrate 45.4% after 10,000,000 simulations) and the one space high pincer there (44.2%). It also mentions 2 human moves, the 2 space high pincer in the bottom right corner (44.8%) and the double approach in the bottom left corner (42.8%). I think we have to rate the double approach as questionable.

[EdLee]

Hi Bill,

I think we have to rate the double approach as questionable.

One human interpretation may be that given this position, the bot evaluates that approaching an open 3-4 is more efficient than double-approaching an open 4-4. ( Thus one of many different properties of an open 3-4 v. an open 4-4. )

[Calvin Clark]

EdLee: Hi Gomoto, Calvin,

For , what are the bots' top candidates ?

LZ 0.15, Elf V1 weights:

elfv1tenukiapproach.png (702.92 KiB) Viewed 14604 times

About the same in terms of move choice. Not sure why Gomoto's is showing black ahead for some responses.

While the double approach in this position may fall of out favor in time, I'll be the first to admit that my opponents frequently beat me despite playing 'questionable' moves and I suspect pros experience this as well.

One can't relax just because your opponent gets suckered into the 'inferior' double approach, because without the right response, any advantage is quickly lost.

For example, if white responds with the diagonal, black is back to leading:

SimpleAndSlow.png (807.54 KiB) Viewed 14602 times

The same happens if white jumps to take a 3-3 now:

AlreadyBehind.png (791.86 KiB) Viewed 14602 times

[Calvin Clark]

For those who are curious, this is the evaluation after the black double approach. Which attachment would you have played? Or would you have played something else?

TheRightAttach.png (808.72 KiB) Viewed 14601 times