How to solve tsumego

Ah that’s really cool. I couldn’t find the right timing to start it (Kept getting Black killed). Didn’t see that after E4 it could work :slight_smile:

Best of luck.

One extra caution, be careful about how pieces “fell” on the first line (especially lying “on the ground”). Since the edge give each one less liberties. so limited pattern can exist (those pieces touching edge fully and touch stones on top sometimes don’t even fit any pattern on them). You need to give them a heavy discount in tallying “common tiles”.

1 Like

I heard jokish claim that top professionals solve tsumego so much they just remember solutions to all tsumego. Although Cho Chikun says otherwise.

1 Like

May I mitigate a little bit my opinion about shapes. Sometimes is really more to be exhaustively careful.

For example this problem (a relatively easy one but I didn’t get it right at first glance)

Black to play

my failure

I just considered the ko. Wrong.

1 Like

Rob van Zeijst 7d put it nicely when someone asked how long to solve a problem they posted: “40 years and 1 second”. 40 years of experience training your intuition and reading skills, 1 second to pattern match that to the problem at hand using your neutral networks and a quick verify read.

4 Likes

When you start doing puzzles, they all are about killing or making life.
“Ko is not the answer” is a recurring line. Seki isn’t even considered.
Usually the solution is just one move (vital point) or a very short sequence (falsify an eye or atari + vital point and so on).
Solution is clear and with explanation.

Then you start trying harder puzzles and everything changes: long sequences, no solution, ko and seki are valid options.
I understand that this change is matching player’s increasing skills, but I feel disoriented facing this step.

In theory it’s a quite simple path:

  1. try to kill/make life, if you can’t
  2. try to live in seki, if you can’t
  3. try to make a ko for life

I suspect there could be also puzzles where the solution is:

  1. just tenuki

It would be the logical step further. :grin:

But considering all possibilities for step #1, being sure you didn’t miss anything, trying step #2, excluding that, considering #3… and then maybe finding yourself without a convincing solution and starting again from #1… is exhausting!

1 Like

If you don’t have to play and able to live in sente, that’s a life and death problem, not tsumego.

Sensei’s library seems to disagree with you.

I suspect there could be also puzzles where the solution is:

just tenuki

I’ve read that there’s a classical problem (ie. 18th, maybe 19th-century) in which that’s the case, although I don’t know / remember which one it is or even the collection.

I agree that the solution to a tsumego can validly be to play away. That means you correctly assessed that there was no threat.

1 Like

That’s charming but I never meet that kind yet

life and death 死活 shikatsu has another word for it (I was thinking a narrow definition of the judgement of a group is alive or dead, which is usually quite simple as beginner practices, not as exciting, and usually way solid that can allow the opponents to play several moves and still alive). But you are right, most tsumego problems have the specific goal set to either play first to live, or to kill and usually the opponent has to answer as the first step. So if the first move is to do nothing, it essentially switches color, and flip the goal

If it is a problem that one side can tenuki more than one time and still able to live (to judge if one side can still live even allowing the opponent to play two moves), it will be a ko exercise. Or more precise, able to calculate the difference of losses between played one move, and two moves (answer and not answer, even not answer multiple times).

2 Likes

Well, we have plenty of them in actual games, when nobody tells you in advance “there’s a way to kill” and you must judge by yourself if you are dead, alive or you need one more stone to be sure. :wink:

Extreme tsumego: just play a game! :smile:

Yes, of course there is an algorithm to solve Go puzzles, the basic algorithm for solving two-player board game puzzles: for every possible move: [for every possible opponent next move: [for every possible next move…]] if the puzzle is solved, halt.

The hard part, in general, is that the “…” recursive repetition can require an unreasonably large number of trial moves. However, in Go puzzles, the number is usually quite small, so the general solving algorithm is practical.

Once you have the solution (or a solution), study the move sequence so you can find it quicker in real games. Each such move sequence gives you better intuition as to reasonable moves to evaluate during real games.

I feel we need to go back and exam the reason behind why tsumego existed and how they come to be.

Since a player can only play one move at a time, every move has to count for something that ended up contributing to final positions. And when stones of both come to the proximity of each other, a more refined boundary between them needs to be defined, however, those a bit away from the boundary can usually omitted be left open, and delayed as long as possible until it is absolutely necessary to play.

So the question is (the word tsumego literally means asking/questioning moves/sequences) - these haven’t played stones, can they really be omitted when the opponents added more pressures directly or indirectly to the groups? Hence, a good strategy always revolves around how to identify these “invisible vital” points (like the eye-poking spot, usually the left open root point between eye spots), the more you can quickly know where are those vital invisible spots the better, even though they are not always the first step to play. The most crucial one is usually the last step. Hence, some might say, you don’t just work your way forward, but also backward from those vital invisible spots (and some intermediate ones).

There are many tricks to help speed it up, and each player has their own ways that suit them, reading backward, bulk reading, liberty checks, isolated groups, negative space mapping (looking for the dead shape), etc, usually many in combined. Realistically, it is really not feasible to read from the start in a timely fashion.

I think I can bulk read every 1 or 2 seconds, and keep track like 4 or 5 main lines, that’s like maybe dozens of sequences at most in a minute if I am very concentrated for a deep read, where if you try all of them without techniques it would be like deep read for thousands if not hundreds of thousands which is not feasible and humanly impossible to track them.

1 Like