Tsumego theory question

(inb4 yes I’m doing some tsumego I’m stressed leave me alone)

Beginner question time!
In the situation above, I can find the right answer but with reading it out only. Is there a “rule” behind it? A way to recognize why the snapback works, for example the shape or the number of liberties? Or something else entirely?

Note: ignore the positioned black stone, my phone reads screenshot as “touch the screen” .


Training of your neural net is the real rule. Reading is just way to test it

also tsumego is cheat - you know that its possible to save this group - so S4 is wrong because it only gives you 1 eye, therefore other obvious candidate is the answer.


It’s pattern recognition. You look at the problem, is there a move that obviously work? Yes → check it with reading and done. No. Are there moves that might possible work? Yes → you check them with reading and if one of them work, done. If that fails you check all possible moves. If that fails you check impossible moves. And then you give up.

In this case there’s only one move that might work.


step that no one ever use in real game :grinning:


@S_Alexander is correct, patter recognization is the most important thing.

I actually watch pro teaching video on Tsumego. I thought he has some serious tricks. Actually no. He said it is simple enumeration. One has to go through each possible moves.

Then ofc, pattern recognization tells which position to try first.
Then he mentioned elimination. Upon all the possible moves, there are obvious not working ones. Again, that’s sort of pattern too.

My view is reading a classic tsumego book, because it lists most common patterns. I am not opposing online tsumego exercises. I just don’t know how well they are organized. I actually never seriously worked on tsumego or tesuji when I just read one book each. I thought about the problem a little, then cheated to look the solution. Then the games reinforces the memory. Lot of those are not intelligence. If you remember them, you know. If you don’t, you don’t.

The pro did mention two “tricks”:

  1. One is the opponent’s key point can be my key point. So you can read the one first.
  2. I don’t remember exactly what it is. Basically, if you play black and try to kill white, lets say, you pretend it is white’s move and white can play two consecutive moves to live. Wherever white chooses to place two stones to live are potential good candidates to read first.

In terms of pattern recognition, there’s a lot of oiotoshis like this surrounding single stones cutting under a wall, like the stone at T7. It’s almost always worth checking the moves on either side to see what they do.

This is also part of a broader pattern: look at any forcing moves that might be useful and see what they do. T6 is an atari and your opponent only has two continuations: capture or play at T8. From there, S4 (or T4) is the natural continuation since it’s another atari. Most of the forcing moves here aren’t very interesting: they either get captured immediately with no gain or lose liberties. R8 and T6 are the only ones really worth considering and one of those does nothing right now (though it might be relevant later.)


look at any forcing moves that might be useful and see what they do.

This is like the advice to begin solving a chess problem by calculating the checks.


See also Repost of a repost -- an old discussion of the correct way to solve tsumego

1 Like

I want to add, in my view, the pattern recognization is basically shape, to understand what is a good shape vs a bad shape, the key point of a shape, the weakness of a shape. Unless one has an ambition to be very good like a high D, one does not need to spend too much time in tsumego to improve reading ability. We learn tsumego to understand and memorize the shapes. Our reading ability is naturally improved in games, no need to spend extra time to improve reading.

In other words, in games, you practice to improve the number of moves you can read out and the length of time you can focus, visualize and memorize, then compare the outcomes, to make a move decision. in tsumego study, you lean how to reduce the number of moves you need to read out to solve a problem.

1 Like

Pehaps something like this?

Step 1: recognize that black’s group is cut off and surrounded without having a living shape.
Step 2: recognize that capturing white’s R6 cutting stones is the only way to save those black stones.
Step 3: recognize that white’s link from R6 to S8 has 2 cutting points and that it is on the 1st line.
Step 4: recognize that white’s R6 cutting stones have only 3 liberties and that those liberties are on the 1st line.
Step 5: remember the heuristic that throw-ins may work to reduce liberties.

If you don’t know the heuristic of step 5, you could discover it by trying (in your head) to cut at T6 after confirming that black S4 white T6 doesn’t work for black. That is actually another heurisic: white T6 refutes black S4, so try black T6 as the first move (before black S4). This heuristic comes up in some go proverbs/practices: “Your opponent’s key point is you own” and “1-2-3 reading”.


Hm, yes, I could start looking for those characteristics. It’s a “pattern” of sorts I can recognize.

For me, I agree with the above posts and simplify it further. Basically, after you play enough games, get strong enough, and recognize good shape, you clue in quickly and do a quick check by reading it out. the first second I looked at that position, the first spot that came to mind for me was t6. It isn’t even a matter of me memorizing the black L above, the number of stones, the entire position, etc., because there are situations where t6 might not work in other similar positions depending on liberty availability. Basically, you run into this kind of a situation enough and it becomes instinctive because you get used to the bad structure in the connection underneath black’s group. You know it might not work for white because you have seen the position before.

People don’t always appreciate how quickly one can read out a simple position. You can get quick at it with practice. You could probably read out that position in just under two seconds if you got really good at it, which means you really don’t loose that much time compared to just “knowing” the position outright. And it is always good to double check anyway.

I do a lot more chess puzzles these days than tsumego :grin:.

I hadn’t seen this shape before but was able to see the solution right away, and I think in a slightly different way from what others have said. Scanning over the position, I see invulnerable white stones at the top, including T7. (Ignoring the three-stone group, that’s like a typical endgame first-line hane position, in which I know white is safe.)

I can’t tell if the group of three white stones is alive or dead. But if it is allowed to connect to the invulnerable stones, it will live. So cutting at T6 is an obvious thing to try.


So… I have another question (btw, is it bad manners to post these questions here, should I post them elsewhere if at all?)

My first response was P15. App says B should play N15. I’m like “OK, I’m not good at this, let’s see”, but app says W’s response would be, to my surprise, O16! Obviously then B plays P15 and W loses 5 stones, no?

Won’t W follow N15 with P15 and lose those 3 stones, if it has to, instead of 5 and poke B’s eye in the process?


In practice the App response is bad. But it plays this way for educational purpose. You need to know how to answer any resistance.


Why in the world would it be bad manners to ask a question about a Go problem on a Go forum? We’re not just all about the memes, after all.

N15 does seem like the correct first move by Black, since then three White stones at O17 are doomed. You’re completely right, white responding at O16 is indeed a bad move, losing more stones than necessary, and wasting a later possible ko threat (and even then N15 would be a better eventual ko threat).

I think it’s just the style of the tsumego to test if one knows how to respond to O16, but I think it’s kind of pointless, since that’s a much easier follow-up to handle than the initial puzzle of finding N15.


You started this thread. As the original poster you have in my humble opinion (compared to the rest) some more freedom in how to shape your thread.


I meant I keep resurrecting this thread, when there are proper tsumego threads (that somewhat scare me, they’re too advanced for me tbh so I don’t participate…).

FWIW, following the meme thread in the beginning as an OGS newbie and trying to understand the jokes led me to some SL articles or even other threads, so I can unironically say meme thread helped me learn Go stuff.


I notice that sometimes the app has two possible responses for white in some problems and picks them maybe at random each time. So sometimes it’s actually worth replaying a solved problem as an alternative response can confirm your reading or teach you an additional thing to look for.

Having said that, I don’t like that some tsumego seem to include suboptimal responses in the “correct” sequence. This seems (as terrific said) for educational reasons or to be clear about the life and death status.