Discussion about R16,pass,O17

At position R16,pass,O17, @GaJ started the conversation:

I’ve read that the AI prefers this to the small knight enclosure. Why would this be? If you played 3-4, it’s for territory, so why play an enclosure where the corner can be taken unless you protect

( FYI @mark5000 )

@mark5000 said:

I’m not sure that’s true. Anyway, Go is give and take. Why play 4-4 if it can be invaded? It’s because you get influence in exchange for the territory. It’s the same here.

( FYI @mark5000 )

1 Like

@GaJ said:

Yeah, I guess that transforms the question into “What is the take here for the give? Why is O17 so much better that it is worth giving an invadable corner?”

( FYI @mark5000 )

@mark5000 said:

See comments by Takemiya. “Uninvadeable” doesn’t mean “better.” The outside-reducing options may be just as severe as invading a similar pattern.

( FYI @mark5000 )

My idiot SDK understanding of this is that the AI prefers the large knight in this case, because the shoulder hit results in an over concentrated shape with the regular knight enclosure.

At SDK/DDK play whatever you understand the follow-ups to. You’ll get way more value from that than trying to figure out what a pro / AI is doing.

Your comment has truth. I’d argue that a shoulder hit on a small knight enclosure results in a concentrated shape, but not overly so. Tencent’s AI, Fine Art, likes the small knight enclosure most. LeelaZero doesn’t, but I think it’s coming around.


If you only play what you understand, you will be slow to increase your understanding.

I like to probe the understanding of things that I don’t understand, to expand more quickly.

Often the answer is “that’s AI stuff - shrug”. Sometimes the answer is “the shoulder hit on the small knight is a thing to think about” - which is helpful.


Incidentally, invasions of the large knight enclosure are now well fleshed out.