Learning about when the corner is safe

Quite often, I find myself agonising about whether my 4-4 based corner position is safe from invasion.

It can (appear to) make the difference between “I am ahead so I should play conservatively” or “I am behind if that gets taken so I need to find a safe way to secure it even in gote”.

How can I learn about “reading the corner”? Are there any specific puzzles or books or just plain “rules of thumb” about this?


If you play 4-4, you have to assume it’s invadable. Even with one enclosure, there are still invasions so you need to take that into consideration when assessing the current board state. If you get to the point where the rest of the board is settled and a 10-15 point gote move is the biggest on the board… then you can consider taking away the invasion with a third stone in the corner. Nice double enclosures around the 4-4 point are 6-3 / 3-5 (note the double knights enclosure is considered bad) If outside influence has been rendered unimportant given the board state, stronger corner enclosures include the slow moves 3-3 or 2-4.

1 Like

Seeing is believing. Start with an empty board with a hoshi stone and self play. You will see it’s impossible to prevent life in the corner. Now go back and add a single support stone at various places like 3 3 and try again. The secondary benefit is you can learn invasion sequences for when it’s your turn to invade. Then add more stones until you find a setup you know you can defend.


Just a note to clarify the question: I do know that 4-4 is invadable, and the related 3-3 josekis.

What I find hard to judge is after an “enclosure” (say, the small knight) then what?

The advice to self-play is of course good, but what I’m afraid of is that I find the corner very tricky - all sorts of unexpected throw-ins etc work - so re-inventing this wheel myself is risky.

Thus if there’s some thought-through body of work about this area, I’d love to read it.

1 Like

Josekipedia, shows all the moves are responses for the most part.

http://ps.waltheri.net/database/ Could help as well.

1 Like

The rule of thumb is that if there is a large nights enclosure, you can live. If there is a small nights enclosure, it is ko, and if there is a diagonal move at the 5-3 point, or then you can kill. Obviously, there are special cases that can change any of these.




Quick reference (N.B.: I’m an sdk, so there might be mistakes):
type of enclosure -> invasion aji
1 small knight -> ko
2 small knights -> ko
1 large knight -> live
2 large knights -> live
1 one point jump -> live
1 one point jump + diagonal -> dead, but small corner or live and enclose
1 small knight + diagonal -> dead
1 small knight + diagonal, but diagonal is result of a kick -> dead, but with a nasty potential cut
1 small knight + 2nd line one point jump -> dead
1 small knight + iron pillar -> dead

Situation might change if one side is very thick near by.


Even if you learned, you still need to kill the invasion of your opponent tries it. Knowing it is not possible doesn’t make his group dead yet. So, learning life & death is a other approach to solve your problem.


I second life and death as the key idea. You asked for a rule of thumb. A rule of thumb I find useful is to play the loosest most outward-facing move that still kills your opponent. Some examples may help.

Exhibit 1 from the Master series games:

In this game, AlphaGo enclosed the corner with a second small knight. Without more, the double small knight ends in ko. But AG had extra stones nearby at R7 and K3 and K4 that caused the small knight to kill, too. AG played the looser defense because it worked. Reading is key.

Exhibit 2 from the Master series games:

Here, AG defended a small knight by just bending up at the distant D12. Again, nearby stones in the J17 and C12 areas made a difference, and Black couldn’t live inside. AG did not play C15. AG played a looser move because it still worked. Again, reading is key.

My suggestion is to learn common killing sequences for the small knight enclosure. Once you learn some killing sequences, you can better avoid overly conservative defenses.

EDIT for working links:


This is exactly the sort of list I was looking for.

Now I can go through each one of these and try to learn/figure out how to achieve the results described! :slight_smile:

Note that I appreciate that nearby stones can change everything.

It would be awesome to have a series of Tsumego around this topic!


I appreciate the advice, I just wish I could see the examples! It seems like either your links are to the wrong games or the coordinates you mentioned are inaccurate. I don’t see anyone playing at K3, K4. It would also be helpful if you said “white” and “black” instead of referring to both players as AG.

Mh, no, I can’t link the comments to either game, either.

Change “event/7” to “event/11” in the URLs. Looks like the links changed over time.

P.S.: This reminds me of an article I read in the Yale Journal of Law and Technology lamenting the number of dead links in our Supreme Court opinions. That was a good read, too.

I think I have read a dan player’s opinion that a hoshi and two small knights is not invadeable.

Edit: Scrolling down I see Mark say that there is a ko: OK, I trust him. An example of how we kyus wander around in the dark…


These answers are not mutually exclusive. See http://josekipedia.com/#path:pdttqfttncqc

In the link, 3 leads to ko and is generally the right answer. 1 and 4 kill if and only if White has no friendly stones nearby and no aji to exploit.