How to solve tsumego

General Go Discussion
#1

I definitely miss something.

The obvious move doesn’t work, obviously.
The less-than-obvious moves don’t work either.
I must be missing a move.
Except that there are about five possible moves around, so I can’t really skip one of them.
Check the solution:

  • it was one of them! But I couldn’t figure why it works. Or…
  • no, wait! There’s a sixth move I didn’t consider!

I must have a wrong approach.
There must be some clear and rational way for checking possible solutions until you find the right one.

I’m not talking about Utaro Hashimoto or some crazy stuff like that.
I’m talking about something like this:

Situation is clear. Goal is clear. Area for solution is clear: it’s less than six intersections.
But I couldn’t even imagine that first move.
I just didn’t even take that into account.
A sequence of three moves and it’s done! :man_facepalming:
(Well, actually you need a bunch of other moves to check that the solution works)

So, is there some sort of algorithm to solve tsumego? Something that peacefully leads me around my flawed intuition and takes me to the unintuitive (for me) solution?

It can’t be just experience, bumping your head against shapes until you start to recognize them.
That can’t be the answer, can it?

Any suggestions?
I’d really appreciate that.

1 Like
:heart: PUZZLES: A tsumego a day keeps the doctor away
#2

I think this is more like a tesuji/shape problem than a tsumego problem. My teacher might call this one: “breaking the bamboo” (joint).

1 Like
#3

how do you define tsumego?

#4

A life and death problem, mostly about making/breaking eyes of the target group.

In this case the solution saves/kills stones, but it does so by exploiting a shape defect to escape, not by making/breaking eyes.

#5

I agree that the distinction between tsumego and tesuji is not very sharp.

But in my view, tesuji have a wider range of applications than making/breaking eyes. They can also be about exploiting a shape issue to make/break connections like in middle game fighting.

#6

I do like you. First I check the good candidates in most of cases it’s enough. When checking it may give me what is wrong and light up other good candidates.

Only after that I check strange moves.

Maybe do some training on tesuji problems.

1 Like
#7

Solving go problems (and playing go in general) uses intuition gained from previous experience of similar positions, combined with reading the actual position. For your example problem, any strong player will solve it on sight, but that’s not some sort of superpower, it’s basically because they’ve seen the shape before and know the solution already.

If you haven’t seen the shape before, you have to force yourself to consider every possibility carefully. There are some general principles that can help make the work easier:

Two examples of general principles applied to the example problem

Identify your goal and use that to narrow down the possibilites

image
It seems like we are supposed to save the A-stones. They obviously cannot make two eyes around D1, so we have to capture either B or C. If we try to kill B we lose by one liberty, so we conclude that the goal must be to capture C. Thus we will focus on moves in that area, like D3, D4, D5, E4, or maybe even F3, F4, F5. We start by looking at the most direct moves: D4 and D5 (ataries on the stones we want to capture). Obvious forcing moves like this are rarely the correct solution to a tesuji problem, but they are easy to read so it’s a good place to start. Even if they don’t work, the variations we find can still bring us closer to the solution, as we’ll see below.

Change the order of moves

Imagine that after a few failed attempts trying to capture C4, we finally come up with this variation:
image
Looks pretty good! Unfortunately, we soon realize that white could play 4 at 5 instead and foil our plans. Do we have to throw everything away and start over? No! This variation is close to correct, we just have to change up the order. To prevent white from getting E4, we start with E4 ourselves. There are several variations to check, but look at this one in particular:
image
Exactly the same as our “fake” success variation from before, just in a different order!

So, if you find such a “fake” success variation, try to make it into a real success variation by trying a different order of moves.

But even with these “tricks”, reading is hard work and takes time. So solving tsumego for practice is equal parts about

  1. Improving your general reading skills for novel positions
  2. Getting familiar with more shapes, so that you don’t have to read as much in the future

Both of these are necessary to become stronger. General reading skill is obviously important, since oftentimes in a real game we’re facing a position we’ve never encountered before. But “just know the solution in advance” is always the fastest and easiest way to solve a tsumego :wink: The more we can use that “cheat” of recognizing shapes in a real game, the more we can use our time and energy to read out and think about other stuff.

4 Likes
#8

In this particular case the obvious move does work though :stuck_out_tongue:

#9

I have no idea which move is the obvious move.

4 Likes
#10

In term of shape the obvious move here is:

1 To take the last move which would complete a bamboo joint

2 to clamp

3 to wedge in

Because of all these qualities, the move is pretty obvious

3 Likes
#11

I also don’t know which “obvious move” the OP @Lys is referring to (I am really curious as to the reason why), or the less-than-obvious ones (maybe D4? which could lead to a corner ko, and I feel is a more difficult to read than the clean answer).

I think understanding why you would think some moves are obvious when they are clearly not, is pretty important to eliminate them intuitively fast, and won’t fall for them in the future.

Summary

Also, If you are white and seeing black plays E4 with a similar rank to you, and the game is still salvageable, it’s probably not a good idea to bet black doesn’t know how to answer wD4 with bF4. Better cut the losses, seal in with wF4, and get sente to play somewhere else to get back to the game. Although I doubt it would be easy since white has 4 more stones locally already

1 Like
#12

Sure that insisting in a lost cause, like pushing a ladder which doesn’t work is not really a good way to follow.

But anyway that’s not really important in a tsumego as what matters is to prove it works even with the weirdest moves by white, not what white should reasonably do according to the future loss

Learning/Teaching is a bit of mystery in itself. And it varies with personalities.
There is that hard pain school on tsumego. Just read don’t check until you find it.

There is some truth in keeping trying to find by yourself but in my opinion there is some truth in the laziness too.

When you practice tsumego you re just practicing playing go. Your attitude will be the same although you may consider more strategic concerns.

Do you check all intersections? No you use your luggage of shapes and “I saw that before”. And that’s crucial in tsumego. You can first trying hard to find the shape. Then play a bit with others, but if it’s obscure, then you can completely give up and look the answer in case you meet it again later on the road.

What keep me surprised is even if you feel good on some shapes you can find harder collection of the same type, as if our eyes will never be good enough to master it.

#13
Summary

Ya, you are right, we need to think of the reaction move in tsumego as a move that if answer incorrectly can cause the most damage (like showcasing a ladder blocking the wrong way).

I subscribe to the idea that if I can not think of an answer and look at tsumego problem for more than 10 seconds. I should just skip to the next one, and try again other days, and go through a limit amount everyday, and overtime get better and faster without even reading them much.

1 Like
#14

My thought process was something like:

  1. The first plan is made: to win the local semeai.

  2. Can I win the semeai with the direct C1? No.

  3. Can I win it with the other direct move, A5? No.

  4. Can I win it with the indirect move E1, to maximum my liberties? No.

  5. D1 is not a useful shape here; this is clear without reading.

  6. Emergence of the second plan. I can’t win the semeai as it stands, so that means I need to gain liberties in the D4 area, or else capture the C4 stones and connect to the B6 group.

  7. D4 is forcing, so I’ll consider it first. Obviously White must respond D5.

  8. What if I atari again at F5, since this is also forcing? White must respond E4. White has too many liberties to be captured with F3, so this variation can be discarded.

  9. Does the D4–D5 exchange prevent White from filling at D3, thus making me an “extra liberty”? No.

  10. What if I atari at D5 instead, since this is also forcing? After D5 D4 I can force again with F5 E4, but White has still too many liberties to be captured with F3.

  11. Development of the third plan: hunt for other reasonable moves.

  12. F4 E4 can be swiftly discarded since White has connection miai of F3 and F5.

  13. E4 is the only remaining possibility. Does it work to capture the C4 stones against all resistance? Let’s check the ataris first: these are fine. Now let’s check the connections: these are also fine.

E4 must be the solution.

In short, a good metric is to first check and discard moves to which you think White has only one response – these are usually atari lines and can be investigated quickly, saving time.

A very similar technique can be applied to chess problems, the solving of which is often fastest when checks are the first moves considered.

In a true tsumeshogi (shogi mating problem), by the way, all moves by Black must be checks.

1 Like
#15

Why didn’t you consider E4?
Suppose E3 was not played. In that case you would probably play E4 immediately and connect.
But E3 is played and you don’t want to play a stone between two white stones. Long time ago you learned: do not rob your stones from liberties. But there are always exceptions to the rule. Take the the snapback: you rob yourself from these three liberties to conquer some stones. Or a throw in.
There is something in your thinking that resists you to start exploring the crazy moves; thinking No, that move is too weird, it just can’t be correct.
But if the more normal moves don’t do the trick, the weird moves become normal.
Be opportunistic. Be a bit crazy, Consider the unlikely moves just as serious as the normal ones. Think out of the box.

Well, I think @bugcat gave a good example of a line of reasoning. Testing a hypothesis, and if that one doesn’t work, test another one. Just as long until you find one that works. Hard work, no easy shortcut.

I think that is the answer :innocent:
No pain, no gain.

1 Like
#16

I was able to do by myself the first part of @le_4TC answer (identify).
Obvious moves for me were exactly D4 and D5.
And obviously they don’t work.

Less-than-obvious were F3, F4 and F5: trying to attack my target stones (C4 and C5) together with their helping friends (E3, E5).

The only intersection left is the right one: that’s why I feel so dumb when I fail to solve these problems.

I agree. And I was actually pretty fast at reading and eliminating them. So now what I am asking myself is: why not trying E4? Was it so scary to put a stone in that narrow passage? So that not being able to read two moves further?
It’s like a blind spot and it happens often to me, in problems and in games: opponent play something that I didn’t consider in advance.

It’s like a no-reading issue specific to some moves.
That’s why I was looking for some systematic way of considering/excluding moves.
But maybe the issue is different: maybe I should just face what I fear. :grin:

Yes, I think you nailed it!

Argh!

1 Like
#17

Actually there is a dirty trick that sometimes work with liberty shortage and running related tesuji if you are good at visualizing pieces in patterns. First, identify your opponent’s groups that only have 2 liberty left, or 3 liberty but two of the liberties adjacent to or can be cut by you, imagine them only has two 1/2 liberty, so in total still only count as 2 effective liberties. (and a potential ko stone will be like have an effective 0.5 + 0.5 = 1 liberties, one can be blocked, one is the ko, or if they are snapback or self-atari is also 0.5 + 0.5).

And then imagine you can place a group of 2 of your own stones and one of your opponent’ stone color “block”, like a tetris shaped, usually a long bar, or L bar, even a diagonal L, and try to fit them in the space we identify having liberty shortage. You don’t have to assign the “pattern” to the tetris like block yet, just find the space to put them like puzzle pieces first. If no answer popped out yet, increase the puzzle piece size to 4, and try to cover one of the liberty shortage stones (2 of your color, two of your opponent’s color), And then to 5, etc. Depending on your ability to keep track of the pattern, and able to come up with common patterns in these tetris like pieces.

This is my way of visualizing the “invisible stones” embedded inside the variations. And quite effective at reducing the “reading” load. It is essentially reading at bulk.

Finally, the dirty trick is, since there are only limited ways to fit these pieces, you would find common spot(s) between them. In this case, regardless L or I or V shape, E4 is the most common tile exist in all of them, which usually is the first move you should try. (second by D4 if the pattern bulk is more complicated, and can be placed on top of existing stones with match patterns, this is the way I am able to quickly confirm there could be a ko start with D4)

1 Like
#18

What’s the ko sequence starting with D4?

#19

I think I’m better at visualizing tiles than at reading sequences, so this piece of advice is super interesting!
I’m gonna try that immediately.

#20

It’s more of a possible ko shape that depends on the whole board and values of two extra steps each other can make that worth more than the losses. Remember white has 4 extra stones here, there should be 4 extra black stones already on the board.