Remembering tsumegos - good or bad?


for a few weeks now I’ve been doing easiest problems in Tsumego Pro app on android. I’ve read somewhere that, you should do tsumegos for so long that you are able to do 95%+ correctly before moving to a harder set.

There are 350 problems in the set that I’m doing and after about 1000 tries I’ve noticed that I start to remember many of the problems (switching coloros does not prevent it, unfortunately app does not apply random rotation to problems). Is this a good or bad thing? I’m afraid that, I can “play from memory” in real game, but the situation can be slightly different and I’ll fail because of that. I try to verify the solution all the times, but the more I recognize the problem, the harder it is to force myself to verify it :stuck_out_tongue:

Or is this a good sign and that’s one of the points of doing one set over and over again?

What do you think?

1 Like

I think it is a good thing but just remember to read every time you see it. Definitely helps you to understand how kill able it is and the different variations you can employ kill the group and try those variation in game. My fighting certainly improved because if doing those problems.


Remembering tsumego is common thing when the problem set is small or the program doesn’t have many similar tsumegos that differ by one liberty somewhere or one stone placed differently. It’s usually good to recognize the shape in the problem instead of remembering the solution to that specific problem. Most higher level problems combine different shapes in one tsumego so that recognizing the shapes is important to find the correct move sequence.

Most programs also don’t support or have fake tsumegos where there is no solution to the question.
I have had this feeling that 10-20% of the tsumegos should be fake. Like one liberty more in a familiar shape might change the outcome completely.
If you have dedication you could build your own tsumego set with fake problems and use it in android with Hactar Go. The app does random rotations and colour swaps to the problems.


Interesting … you mean that, for example, the right answer then would be to click a “Tenuki” button or something? Maybe we should suggest such a feature to the developer of Tsumego Pro … but I imagine spending a lot of time on such a problem then (probably too much), and it should definitely be an option that can be turned off. And maybe there should be some system, like …

  • only begin when you can reliably solve the first so-and-so many problems 95% or something, and then …
  • begin to offer fake Tsumegos that somehow look similar to Tsumego from the “reliably solvable” section.


Otherwise it might be too hard to recognise, I think.

1 Like

Yeah the interface would need one more button. Also one possibility would be to play the best endgame locally but that is quite weird since sometimes it’s better to play sente sequence that is not locally best but keeps sente.
Actually I have seen some “best move” problems that are almost like fake tsumegos.

Yeah these seem like good options to have.
I would be satisfied with just mixing the fake problems in the set and be done with it also.

Hello, kajmaj87. Let me introduce a forward written by Cho Chikun in one of his tsumego books (it is only in Japanese, but the title is Hitome no Kyuusho). He wants the readers of his Hitome series to repeat the problems 20 times, or 5 times at least. He claims that then you will be able to come up with the move in real game. He implicitly states that the goal is to be able to reproduce the moves, or "tesuji"s in real game. If you can do that, after three rounds, then you must have talent in the game and you have probably graduated from the book. But, on the other hand, you should be able to solve the 95%+ of the problems by then.

If you think you just memorized the solutions without being able to build the "tesuji"s into your intuition/default reading/instincts (or whatever you might like to call it), what I do is to move on to some other problem sets (I hop around four different books which is just right for my current strength). I still memorize some of them when I come back later, but that means I will be able to use them in real games, and it is a good thing.

1 Like

Thanks for your insight. When I wrote about remembering I meant only a subset of the problems, usually the most trivial ones, so I still cannot reach 95%.

There is nothing wrong with using our memory. If you had access to all the tsumego in the world, if you are able to remember all the solutions, then… oh well, you no longer need to understand them :wink: Every variant of tsumego becomes a tsumego on it’s own (sometimes for the other side). Unfortunately it would be next to impossible to remember all the solutions, so while memory and intuition helps us speed up the process, it’s the skill of critically verifying different variants that tsumego teach us. In the game, you won’t have the ‘try again’ option :slight_smile:

Perhaps the only thing that a tsumego app needs to change to create ‘fake’ tsumego is by letting the user decide before playing whether it’s a kill or a live problem (or both depending on who goes first, or ko / seki), which is nothing more than adding an option to hide the question and a button to choose colour. It is also more resembling of how one would approach a problem in a game.

1 Like

my two cents… for whatever it is worth, is that remembering the problem isn’t a bad thing, the more important thing is how you solve it. I know a lot of the apps you can just brute force it by trying moves til it works, and this is not good. If you do it in your head, it’s a good thing because it is strengthening your reading skills.
It is also important to remember that even though remembering the problem isn’t a bad thing, the more important thing is to identify the weak points and remember those. Getting to the point where you can look at the easy problems and automatically identify the weak point will help you much more in real games.

1 Like

I use the same app to practice tsumego. I’ve worked through the several easier problem sets methodically so I feel I can give some insight.

My goal is to always work through the problem sets methodically so that I can feel a sense of accomplishment and move on to other problem sets. I keep a paper record of what I have/haven’t solved in case my phone dies and progress resets. There’s no cloud saves.

I do “passes” through problem sets, bookmarking a problem immediately if I get it wrong. At the end of the pass, I write down the numbers of the problems I get wrong and calculate a % right (which is just a fun statistic). Then, there are two possible next steps depending on how much time I have. I prefer path #1. There are just too many problems and not enough time.

  1. Drill the problems that I got wrong until I get them right. Then, drill at least as many times as I got them wrong before removing them from bookmarks.

  2. Reset progress and conduct a second full pass through the problem set to see what % I get right this time. I might do this from time to time to see if I really can get 100%.

Lastly, here’s my opinion about memorizing answers. I know a 9d that argues that tsumego is all about patterns and memorization, but at that level it’s all pretty intuitive anyway. In my opinion, by carefully reading out each problem to make sure there are no hidden traps, you train a skill you can apply to other problems. By simply memorizing the answer, you can only solve that problem. The utility of one way seems clear over the other.

1 Like