Two questions I'd like to ask, any input appreciated!

1. What do you think of the following tsumego-regimen?
Almost all tsumego I do are the ones I solve within 30-40 seconds. Longer than that and I tend to look at the answer, not to be stuck on one problem too long. I guess about 5% of them I spend longer on and try to read out all possible variations.
My estimate of the tsumego I do:
25% within 0-10 seconds.
25% within 11-20 seconds.
25% within 21-30 seconds.
20% within 31-40 seconds.
5% within 2 minutes.
0% longer than 2 minutes.

I think I need to be a bit more diverse with my tsumego (see question 2) and have a book with much easier problems, but also spend more time on the tough ones and not look at the answer too easily.

Currently I'm going through 1001 L&D, I'm at the 3-move problems and I think I get about 60% right at the moment. I got about 80% of the 1-move problems. Almost never longer than 40 seconds on one problem.
Now that it's getting tougher, is it wise to just start over again, or keep going until I start messing up most of the problems and then start again?

2. Do you know a good source of plenty/easy tsumego?
I've heard it's also a good training to do easy tsumego real quick. Problem is that I lack problems to do.
There's only so many times you can do GGPB 1 and 2 (the answer is 4 times) before you get sick of it. I've tried to do encyclopedia of Life & Death by Cho Chikun. I did problems 1-25 super fast. Next day I did 25-50 super fast. Then I did 50-75 and already the problems are not easy enough for me anymore to "blitz" them.
I'm about 10 kyu, but catching up on L&D so my L&D level is perhaps a bit weaker.

Any good suggestions for books/PDFs/... with good problems that I can solve within 5-10 seconds?

I am a fan of doing easy tsumego, more than most people, but if you are spending 30 seconds on each problem and getting 60-80% of them right, something is wrong. That is what happens when you solve problems by looking for the obvious shape move, reading out a line or two, saying "sure, looks good", and then checking whether you were right. That's not how to do tsumego. You are practicing for game situations where you have to read 100% accurately and need to really bear down and read it all out and be sure of your conclusions.

There is a problem, which I recently discovered myself, that it's very easy to go down the following progression of increasing tsumego difficulty:

• Obvious answers you can just blitz out (you get 100% of these)
• Problems with a clear first move try that you should verify, but you're lazy or are going for speed so you just immediately say that's your answer (you get 90% of these)
• Problems that mostly fit the shapes you know but sometimes have little tricks in them (still using intuition, you get 75% of these but sometimes shrug your shoulders and say "oh yeah, I missed that")
• Problems that really require extensive accurate reading (you get 50% of these and have trouble reconstructing some of the correct answers after you've seen them because the variations are too deep)

Now suddenly the problems are frustrating and hard, and because you have been exercising your intuition much more than your reading, you don't have the muscles built up to do well on them.

My prescription is to focus on reading from the start. Even when you know the obvious first move, read out the followup and visualize it. When you get a problem wrong, go back to the diagram and really visualize the whole sequence until you think that you could do it yourself.

I would start over and really try to get every problem right, giving yourself a 10-minute deadline after which you're allowed to guess and check the answer.

I don't think this is quite right.

Tsumego problems as presented in books tend to be graded, explicitly or implicitly, in difficulty (there's a star system, or a clue, or the clue is in the name).

Generally the level of difficulty is based on the number of techniques in the problem. There's always at least one, of course, but rarely more than three. Those with three are often amateur high-dan level.

Therefore the key to successful solving is first to learn the techniques and then and then to identify them (in terms of the surrounding position - liberties, edges, corners, etc) and then learn to recognise the structure of the problems in terms of how many techniques they have. Quite a few techniques go in pairs (e.g. hane and placement).

Seeing the structure means you have signposts even for hard problems, and treating the techniques as modular units (which can be quite long e.g. the tombstone tesuji) prunes the search tree dramatically. Reading really means reading the structure nit the moves. There's a reason why the Japanese chose the word "read" - you are expected to read words as units, not letters. Likewise with tsumego.

Of course the more interesting problems may have branches where different lines have different groups of techniques, but the principle remains intact: recognising the structure usually solves the problem.

What follows from this is that if you have done a problem, or seen the solution, and you don't recognise the technique(s), that means you haven't studied it properly. If it's an easy problem you should normally be seeing on technique (so that's a good way to start = less "noise"). If it's a high-kyu/low-dan problem expect to find two techniques, and if it's high-dan look for three.

Think about these techniques hard, analysing their contexts and typical results (kill, ko, seki), give them names, work out which techniques tend to go in pairs. That "effortful" reflection will help you absorb the techniques in your subconscious.

Of course, this can't be expected to work 100% of the time, but if you find a case that doesn't fit, work out why it doesn't fit and that will make it likely that you will remember it.

OK. Everybody is different, so all I can say is what is true for me, and it may not generalize to everyone. In my case I often can deduce/guess what the techniques are (e.g., "the obvious sequence doesn't seem sufficient, so I bet I must have to take advantage of a shortage of liberties somewhere") but have trouble calculating how to make them work (because the shortage of liberties actually appears 7 moves in, say). I think a lot of players (including me) try to work too much at the "word" level when in the end things often come down to the precise calculation of whether a move works or not. For example, I often see weaker players than me try things like throw-ins and placements at inappropriate times because they have learned the techniques and hope that they will work, rather than reading them out. Of course, one could argue that if you really knew the technique you would only use it when it actually works, but, at least at my level, this generally requires doing some actual precise reading, even if it is guided by higher-level knowledge.

Based on your level I would recommend:

Jump Level Up Volumes 1-5 - ~5000 problems, $100
Essential Life & Death Volumes 1 & 2 - ~2000 problems, $40
WBaduk Pre-inter(Easy), Pre-inter(Hard) - ~1500 problems, free

I started doing these about 2 years ago. That's about 8500 problems in total - these, in combination with Graded Go Problems for Beginners, it's pretty hard to exhaust no matter how many times you go around.

That is not a correct presumption. I do start by "intuition", obvious shape moves, key points. But from there I read out several variations until I know it works or doesn't. When my answer proves to be wrong, it's because I visualised something wrong (like shortage of liberties, or a ko...) or simply missed a variation or misread one.
I hardly ever (sometimes concentration does slip) pick a move, choose a counter and say: okay that'll work... That's definitely not how I do tsumego.

Don't underestimate how long 30 seconds actually are, you can read a whole bunch of variations in that time. Or perhaps I'm overestimating 30 seconds and I'm actually spending a minute on them, thinking it's 30 seconds. Perhaps I should time myself one of these days to know for sure.




I do read the followup, several of them. Going back and visualize when I was wrong is something I do rarely, but it's definitely good advice, maybe I should do it every time I'm wrong.


However, from your comment, is it correct of me to summarize that you'd advice to do a problem the following way:
- Take a move, pick a followup and read out. Pick another followup and read out. Etc etc
- Take another starting move, pick a followup and...
- Another
- Another
-...

In that case, tsumego is not about finding the solution but rather to practise reading, visualising. That is something I can agree with. However, reading out variations that you already know don't work or do work seems a bit silly. When you start with the most obvious move and you find the variation where it works and your opponent has no good reply, do you then continue on the problem and read out other moves as well?

Cool thanks. Of those I already have Jump Level Up.
I've finished Jump Level Up 1 "fairly" easily in a few weeks.
I did Jump Level Up 2 in a few weeks, too, but I felt it couldn't hurt to redo it before doing the test. It's been lying next to my bed for a while now. I also did Level Up 5- 10 in rather quick succession so I think I'm fed up with it for now. I'll take it up again in a week or so and give it another go

I've read that Essential Life & Death aren't meant for reading exercise, rather to recognize L&D shapes, is this correct?

If I look for WBaduk Pre-inter, I only find lectures. Is there a problem set, too? If so, where?

OK, good! Then the problem is reading ability rather than reading energy.


I do know how long 30 seconds is because I time my problems! (This is also how I arrived at 10 minutes for my "give up and look at the answer" duration.) Also 30 seconds is a usual byo-yomi period. In any case, if you're only getting 60-80% on problems that you think are simple, then it's probably not long enough.


Yes, I highly recommend this. I think it is important, after every problem I get wrong, to see how I could have gotten it right. And if I still can't confidently read out the whole variation to the end even after having been shown it once, then I definitely couldn't have gotten it right.


I only read out variations that I'm not sure of. If my first try works, I stop there. However, to be sure that a move works, you have to consider all reasonable responses by your opponent. If you confidently produce a solution to a problem, after thinking that you have read it all out, and your answer is wrong, then either you misunderstood the final position or you didn't consider the best move by your opponent at some point.

Ah thanks for the clarification!
Yes, I probably have a bad idea of how much 30 seconds is. And probably also on how much 80% is
Let's just say that I did very well at the 1-move problems, almost never getting 2 wrong per page, while with the 3 move problems I'm wrong often once a page, sometimes twice.
This probably means I need to intensify my reading.

So far I've just assumed that it's my first time going through 1001 L&D and I'd just do a "quick scan" of the book first, not spend too much time on each problem.

Problem with tsumego is that everyone has a different opinion on them. Even pros can't seem to agree. I've heard:
- Never look at the answer
- Always look at the answer
- Look at the answer after x seconds/minutes
- Do easy tsumego
- Do tsumego that really challenge you
- ...

So basically, it comes down to preference, I think. Whether I'm doing tsumego 100% effeciently, I seriously doubt it. But I'm definitely increasing my reading ability, getting stronger and recognizing shapes more easily. So let me leave it at that and say I'm doing all right.

Thanks for the tips, though. I'll take them to heart.

IIRC, you have started to enjoy tsumego, so it's hard to say that that is a bad regimen. But I notice the large gap between 40 sec. and 2 min. You might put a cap at 1 min. and take more time on other problems. It seems like you are missing problems that you spend 30 sec. on, for instance.

One thing you can do is, if you don't have a solution within, say 20 sec., move on to the next problem, and come back to the unsolved problem later within the same session. You unconscious will keep working on that problem while you are doing others.

Could well be.



That sounds like a good level for you.



IMO, you should aim at 100% for the 1 move problems. (50% - 60% is fine for the 3 movers.) 40 sec. max is fine at the one move level.



Review is good, at least on the problems you missed the first time. You should do them over until you get them perfect, and then review some more (overlearning). As you I'm sure know you should space out the reviews.


I wouldn't worry about it. As I said, when I was playing bridge I was very quick at figuring things out, but I never ever trained for speed.



There is some value, I think, to using short time limits as a way of making problems more difficult. By super fast I take it to mean in 1 - 2 sec. At whatever speed, you missed some of the problems in the 50 - 75 group. Fine. You have some problems to review. How many did you miss? If you missed 15 you should probably take more time on the next set.

Why was I so quick at bridge? As I said, it wasn't through trying to solve problems quickly. I'm guessing, but I think it because I knew what I was doing. Thinking along those lines, I would say, study the problems you missed, and study them thoroughly. Don't just look at the main line in the book and say, Oh I see, yeah, yeah. What are the key features of the position? What did you miss and why? What stones can you move or remove and the problem essentially remains the same? What stones have to be exactly where they are? Can you reconstruct the problem, or essentially the same problem, on the board, or in your mind? Can you refute every move, not just the ones in the main line? One drawback to review is recognizing the problem and solving it, but later encountering a similar position over the board and getting it wrong because a key detail is different. If you know the problem thoroughly you won't do that, or hardly ever do that. If you spend 10 min. solving a problem set, there is nothing wrong with spending another 10 min. studying the ones you missed.

Edit: I don't think that there is anything wrong with looking at the answers. However, since you are feeling the lack of problems, maybe if you can't solve a problem in one minute, save it and try it again later.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

It seems to me that there is something that dfan, John Fairbairn, and I all agree upon. Study the problems that you miss.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

IMO, what it comes down to is that people mostly tell you what they think worked for them. But psychologists and cognitive scientists have done, and continue to do, research on learning, if not much specifically on tsumego per se. A lot of the results of that research is available online. The research on overlearning goes back over 60 years, research on the 50% rule of thumb goes back over 55 years. Recently I read something about the idea of varying learning tasks and spending only 15 minutes or so on one type of task. Research on eye movements when solving tsumego life and death problems indicates that good solvers look at the location of possible eyes instead of possible moves. Don't dismiss people's experience, but check out the research, as well.

Edit: A lot of the research has been discussed here on lifein19x19. You don't have to go far afield to find it.

Edit2: I was also reading recently about the value of varying the learning task slightly. OC, varying a tsumego problem can destroy it. Black to play and kill? Sorry, now it's alive. But with thousands of tsumego available online, drawing from different sources should yield many similar problems.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

1.

Play, review, play, review, play review, ...
throw in some tsumego
play, review, play, review, play review, ...

2.

Magic Baduk Go App
(L&D and Tesuji, verrry nice)

@John Fairbarn: I am quite interested in your tsumego solving technique.
If I understood well, one shall first know all the techniques and recognize wether they solve the tsumego or not right?

Do you have any reference explaining it?

Many thanks for your help!

One of the most interesting themes of the book "Make It Stick" on the latest developments in how we learn is that even intelligent people (the book is by academics and so they mainly reference college professors and students) have major blindspots. One is that they confuse learning with memorising. Another is that they refuse to listen to experts because they are convinced they know what works for them (it may work but be very inefficient). Yet another is that they often believe they are a particular type of learner: auditory as opposed to visual, say. This is not backed by science. If it's accepted as a preference, it may appear to help with motivation for study, but it's actually another example of the above case of "may work but is very inefficient". Other misconceptions include the belief that an hour's hard study must be better than 15 minutes, or that writing down what the lecturer says is a good way to learn.

All of this is relevant to tsumego study. I happen to believe that many pros have come to the right way to study, even if by trial and error rather than formal research. I say this because I see time and again advice from them that is flatly contradicted by amateurs (e.g. they say it's OK to look at the answers even if you don't come up with a solution - Make It Stick agrees). They often say they don't spend too long doing tsumego (i.e. they do the equivalent of 15-minute bursts instead of an hour or two) and that they are not very good at reading - but they have absorbed the techniques that make much reading redundant.

So, assuming you are not one of the Unbelievers pinpointed by MIS and are prepared to examine a different approach, here is what I infer, on the basis of wide reading of what pros say but with little actual competence in tsumego myself, is part of what they study so as to avoid reading.

Most strongish amateurs know a good example of that already: the proverb "The L shape is dead." Which means that if you see an L-shape in a game you don't need to bother working it out. Of course, if the opponent plays it out you need to be careful if you want to kill it, so there is some post-facto reading involved, but you know the result already and that makes it quite easy.

It's a little trickier if there are hanes around, and what happens if there are legs attached? Dan players probably know these variations, but do they know instantly whether a liberty affects the status of the group. They can easily work out the answer, but if they have to work it out, can they reasonably say they "know" the L-shape.

Let's say they don't need to work that out. Can we then say they know the L shape? I would have once said yes, but I was brought up sharp by a question from a pro or an insei - I forget who - who asked: how many liberties does an L-shape have if you need to kill and remove it? And if you ever need to do it in a game, I can tell you it's rather difficult to get right. How much better (much quicker and safer) if you just know the answer in advance. I don't know how many pros would know the answer in advance, but at least one does, and at the very least this example is an indication of the sort of preparatory study a pro or insei expects to do. (Someone may wish to pose this problem on a separate thread.)

So, with that brief glimpse into the value of knowing a technique in advance, and what it really means, let us look at the tombstone tesuji. It's not a great name and in fact there are about six or seven names in the oriental literature, but "tombstone" (from the Japanese and so referring to a Japanese grave) seems to have longest pedigree in the west. The problem below is an example. Black to play. Look at the answer when you are ready.



If you do several problems of this type, you will soon recognise some of the characteristics that make this an identifiable technique. At the highest level it is usually part of a capturing race, so we know liberties must be important. At a lower level it involves adding a stone to make a two-stone line down to the edge (burying the body - imagery to help you remember) and a throw-in (tossing the dirt into the grave). You could call these techniques if you want but I prefer to see them as sub-techniques. Not because they are single moves but because they don't actually solve the problem. Their function is to make the tombstone technique "flow" and work.

Now to stop there and go onto the next problem would be a mistake. There are other things we can learn even from this base example. Thinking this little bit further is what MIS means by "effortful" practice which leads to learning as opposed to memorising. Learning means you can use it intuitively and flexibly as opposed to just recalling it.

I'll suggest a couple of things this base example tells us. One is that the gravedigger can apply this technique with just three liberties of his own. That's valuable time-saving information if you know it. The second thing is that, though long, this technique is pretty much a one-way street. That's important to know because it tells us something about how reliable this technique is and how much we can trust it.

Assuming you have sucked all the juice out of that problem, is there anything else "effortful" you could do to reinforce your learning? Well one thing is to consider what happens if it's in the corner. Do you gain an extra liberty? Try Example 2. Black to play again.



So, a little twist there, and you can refine this base example even further by doing some more examples (nb by spaced repetition rather than being a Stakhanovite). But knowing just what you have seen up to now lets you just look at the position in Problem 3 (B to play) and "know" instinctively that there is a promising play - and what it is. This is just like the L shape knowledge, though of course you may need to practise a few more problems of this type to embed it that intuitively in your neural network.



The next problem (B to play) is similarly a look-and-see solution of the basic theme, the difference here being you have to count extra liberties before judging whether you can win the race. It's a long solution but it is definitely a one-track look-and-see, at least if you have done a few of this type.



All these problems involve a single technique and so would normally be classed as Easy. "Easy" covers a range, of course, and you may wish to use a different term (e.g. Simplex), but the essential point is that a single identifiable, learnable, reliable and flexible - as well as common and useful - technique is all that you need to solve even long problems essentially without any reading (i.e. next to no branching).

I have identified a little over 50 of these techniques and they were sufficient to encompass all the Chinese classics of tsumego.

Obviously some problems involve more than one of these techniques (Duplex and Triplex, but that is the usual limit), and I'll deal with that in a later post. But for the time being here is some more practice of the basic (simplex) theme. All Black to play.







The next one is still the basic theme but with a slightly different finish. Different possible finishes apply to quite a few techniques, which is one reason to keep practising each type of problem. There may be some reading involved but by superimposing the template you know, and counting liberties instead of trying out moves, you can still eliminate almost all of it.

Hi John,

Thank you very much for this great post.
After looking at your different examples, I could really feel something was integrated.

Are there tsumego books teaching with this approach?

Thanks again,

Martin

Part 2 here.

First, though, on the principle of spaced repetition, here are a few more practice problems of the basic tombstone tesuji pattern (simplex type - one technique). Black to play in all cases.









When it comes to duplex problems, where two techniques are involved in the main line (variations lines may of course involve some other techniques but we are talking of the flow of the problem), the techniques can come in either order (AB or BA) and could be any of the 50-odd easily identifiable techniques. But in practice one order tends to occur more often, and the paired technique tends to fall within a small sub-set. Both facts are useful to know as they aid the trust you have in applying templates rather than going: if he goes there, I go here, which tends to end in an expletive.

Clearly it is better if you have studied the other techniques in their simplex form first, but we'll gloss over that here.

Our first example involves a well known second technique.





















Duplex problems are obviously harder then the "easy" simplex ones. Generally duplex problems cover a range of high kyu to low dan. Triplex questions also have a range, low dan to high dan, but can appear much simpler than that if you have absorbed the basic template of the simplex type and the common AB/BA associations of the duplex type.

I give just one example here - at the lower end of the range - but I hope this and the problems above are enough to show why this approach works so well.





To answer the question whether this approach to tsumego exists elsewhere, I have seen small portions of it, which is one reason why it could exist in bigger form. But the truth is, I don't think they bother with that because if you are a serious student you are expected to follow the proverb "Hear one, understand ten" (一を聞いて十を知る). THt is, go away and do the work yourself. I think I have shown that it is easy enough to do on your own. It's just a matter of collecting the examples and the counting up to three techniques. The collecting process itself is of course a form of spaced repetition, which implies it is necessary as well as useful.

If you want to try another technique for yourself (i.e. collecting and analysing it yourself), I'd recommend Rooster Standing On One Leg as examples are easy to find and it has very wide application in real play. One way to start may be to look at my Gateway To All Marvels (SmartGo e-book - I think) as it lists 50+ techniques with pointers to many example problems for each. The only thing it doesn't do is show the simplex/duplex/triplex analysis, and as these are classical problems it's a little on the hard side (i.e. most are duplex or triplex). But you have to do something a little more effortful than just read a book if you really want to learn.

Thank you so much for all of this, John.

Small nit.


And another.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Bravo!

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

White to play (Thanks John!)

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------------------------
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . O O O O X . . X X O O . . . . |
$$ | . . O O X X X O . X X O X X O . . . . |
$$ | . X X X O . . . . O O O . . . O . . . |
$$ | . . . . X X O O . . . . . X . . . . . |
$$ | . . . . . O . . . . . . X . . O . . . |
$$ | . . . . X X O . . . . . X O . . . . . |
$$ | . . . . . O X . . . . . . O . . . . . |
$$ | . . . . . . X . . . . . . . . . . . . |
$$ | . . X X . . . . . , . . . . . , O . . |
$$ | . . . . . . . . . . X . . . . . . . . |
$$ | . . . . . . . . . . . . O . . . . . . |
$$ | . . . . . . . . . . . X O . . . . . . |
$$ | . . . X . . . . . . . X . . O O . . . |
$$ | . . O X . . . . . . . . O O X O . . . |
$$ | . X X O X X X X X , . X X O X O . . . |
$$ | . X O O O O X O . . X . . X X O . . . |
$$ | B O . . . . O O . . . . . X O . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]

Solution: