Follow the wise old men - and solve problems faster

[ez4u]

A typical situation is that you are reading a tsumego book on a train. You probably do not have a board and stones, but let's assume you do. You plonk down a dozen stones. Purely mechanical, but better than nothing. So far so good. You then have to remove them, but probably end up picking up too few or too many stones and have to waste time reconstructing the original position. This builds up resentment or other negative feelings which chip away at your motivation to do tsumego in the first place.

You may instead be using a smart phone. You could click through the moves on an app, but experience shows we all get click happy and the procedure ends up as a speed exercise for the thumb. Again, you also have to go through the negative and time-consuming process of unwinding the sequence.

You could watch a video and let somebody else do the work for you. That's filling in time, not using. In any case, while the demonstrator is picking up played stones and reconstructing the position, you will be switching off and thinking about what you'll have for lunch or "did I really turn the tap off?"

Or you could put in a bit of "effortful practice".

This argument seems to be: assume that all alternative methods eschew effortful practice, assume the use of arcane coordinates does involve effortful practice. Voila! We only learn via the coordinate method. The fly in the ointment is that we need to fill in the pictures with some real actors. Clearly the person on the train is someone like me. Let's call the keener with the coordinate book "Ian". I am sure you can already see the problem. Hand the coordinate book too Ol' EZ and what happens? Right! It flies through the air and lands neatly on the giant, "unread" pile. Meanwhile sit Ian down in front of the Go board and what happens? Right again! Beads of sweat form on his brow, his jaw clenches until his teeth hurt as he stuffs new insight after new insight into his brain.

[jlt]

Interesting problem. I had read the sequence of the solution but thought that it failed because of a proverb.

https://senseis.xmp.net/?CaptureThreeToGetAnEye

[Bill Spight]

Interesting problem. I had read the sequence of the solution but thought that it failed because of a proverb.

https://senseis.xmp.net/?CaptureThreeToGetAnEye

I suspect, particularly as the text makes a big deal about the eye being false, that Hashimoto constructed the problem to show how such an eye could be a false eye.

[Ferran]

I'm unconvinced I'd be able to properly read lettered coordinates. I don't count my own alphabet, much less a Goban's (without 'I', I assume). OTOH, paired with explanations (ie, "keima at hoshi 10-4") I might follow a numbered (even in kanji) board with not much of a hiccup, I believe. "Algebraic" systems a la chess have always confused me.

Take care.

[SoDesuNe]

But actually my point was that I don't see any benefit on the learning effect when solutions to problems are presented as coordinates. As someone not used to dividing up the board in letters and numbers (or numbers and numbers) it just takes more time checking moves. I also don't see the time/space-difference between writing coordinates by referencing letters and numbers and showing a board position with the correct moves labelled by numbers. In both examples moves precede one another, are sequential so both have your "time"-attribute? But a board position shown is much easier to understand.

Several points emerge from this.

1. You don't need to divide up the board into coordinates. The book editor can do that for you - see below.

2. When you are seeing a board position with all the moves on it together - the typical variation diagram - there is an implied sequence (1, 2, 4, 5...) but you are seeing it, at a single moment in space. You are also seeing the garbage left my captured stones. You are not seeing the move order in any meaningful sense. You only see it as a sequence if you play it out on a board so that each move (not just a single diagram) gets its own moment in time.

I'm still sceptical that just by delaying to show the full solution you accomplish greater learning effect.

Firstly, you - likely - read a solution already (by attempting to solve the problem) and need simple feedback on your attempt. Delaying that feedback is not beneficial according to conventional "deliberate practice"-theory. Secondly, having to backtrack to the problem diagram everytime you check a "written move" does not add anything experience-wise, does it? You already know your solution, so if you're right you will just get feedback - only later. If you're wrong, you have to go at it again anyhow - but possibly you again needed longer to come to this conclusion.

If I use board and stones and replay the solution on the board, I do add experiences. Most certainly touch and possibly also where I'm at (connecting learning material to a location is an old mnemonic). We could still argue whether this time commitment is efficient, though.

Simply just adding more time never benefits learning. That's why most of us don't write 150 words per minute despite typing multiple hours each day.

On a side note: This time factor is also - in my opinion - the reason for most of the questions to get stronger. Not because people necessarily shy away from the work but because they don't want to waste time working inefficiently. Of course running into roadblocks and experiencing what does not work for you is seldom inefficient learning-wise, only progress-wise.

Effortful practice in his case means visualising the moves by following a prompt.

But I did already visualise them in my attempt to solve the problem, didn't I? Following your own foot steps in the sand (ie checking the solution) makes for a boring journey to already known places ; )

The aim is internalising the solution so that it becomes a tool in your tool chest, and one that you actually know how to use.

No argument here.

To reinforce my claim that seeing a single picture is deceptive, or even useless, consider the following. (I'm using a discipline that most go players won't be familiar with, to ensure assumed knowledge does not get in the way of the point.)

BalletPortdeBras.jpg

The typical presentation in a go book would actually be only pictures 1 (Figure 1 in go) and 4 (Variation 1). But as a concession, if you really do believe variation diagrams show sequences clearly, I present a sequence here. What is going on?

Not sure about this comparison. Your picture shows body movements which on one hand go beyond the flat surface they're imprinted on and - more importantly - on the other hand it skips a lot of inbetween movements. A go diagram does neither. A go diagram - so to say - delivers (locally) perfect information, befitting the game.

A counter-point: The Feynman diagram. Invented primarily to offer a simple visualisation of the complex and difficult to understand interaction of subatomic particles which before had to be written as complex and difficult to understand formulas. The point being: Simple visualisation of a complex matter is an asset, not a cheat.

[John Fairbairn]

But I did already visualise them in my attempt to solve the problem, didn't I? Following your own foot steps in the sand (ie checking the solution) makes for a boring journey to already known places ; )

I won't answer every point (in various posts) - sometimes because I don't really have an answer. As usual, I'm trying to provoke a discussion, and it paid off this time because Feynman diagrams were new to me and this looks like something worth reading up. Thank you.

As to the point above, I think we are going round in circles a bit because visualisation seems to mean different things to different people here. No-one has picked up on my mention of suji, so I already suspected that was going to be a problem. For similar reasons, the lack of mention of the stones and board tells me something is amiss.

I'll try one last tack. What I am arguing for is that, to learn something from doing tactical problems, the most efficient kind of practice is effortful and rich in associations. Trying to solve a problem in your head is part of the effort, I agree, but you can add to the effort by playing the moves out on a board. This gives you the flow (suji) of the stones and creates associations. I believe that in real life, very few people do play over variations on a board. They just look at the solution diagram as confirmation they got the solution. They give themselves a pat on the back and move on to the next problem. Then they wonder why they don't seem to improve much. I believe that this standard method fails because of lack of reinforcement and associations. How you do that reinforcement is a matter of personal preference, but however you do it, I think it's got to be done. My suggestion for using coordinates or board letters definitely does not mean trying to visualise the points indicated in your head. I know that's hard and messy. That is precisely my point, so you don't have to keep telling me. What I'm assuming is that, because it's so messy and hard, you will actually get out a real board and play the solution out on the board - to absorb the flow, to make the associations.

The exact method you adopt can be totally different from what I'm proposing but I don't see that efficient learning can be achieved unless your choice of method guarantees extra such as associations and flow. Mere repetition of problems that you've solved before but can't now remember is even more boring than - and actually more similar to -tracing your own steps in the sand.

Everyone here has had to study something by means of books. The most basic way is to read the words in the book. Some people don't even get that far. They just look at the pictures, or look for a comic version of the book. Let's not be snobbish about words. Let's assume the word-readers and the picture-readers get something equal out of the book. I would argue that in either case, though, they don't get very much. They might be able to answers questions in a class test, but probably more on the basis of memory than understanding.

Reading the book's words and looking at the pictures will add someth9ng - but not very much, I suggest. (There is a worse approach, mind you, one I'm guilty of, and that is buying a book and assuming that's job done - the first step is the hardest step, and all that jazz.)

What really allows to you start understanding the book and gleaning whatever may be useful in that book is engaging with it. There is no standard way, perhaps, but people will do things like discuss the book with other people, make notes in the margin, re-read or even memorise sections, try to write the book's ideas in their own words, get sidelights by reading similar books or criticisms, watch a film of the book, and so on and so forth.

All of these methods will add something further to your knowledge, but some methods are better than others. For example, scribbling in the margin is often close to useless, whereas scribbling the book's ideas in your own words - self-testing, in other words - tends to pay off big time. Your internalise the knowledge.

What I see in go is that most people who bother to buy books just look at the pictures. They don't really engage with the book. They even make it impossible to engage properly with the book - they read it on the train or they do five minutes a day. Or they pass the problem on to other people and ask then how to solve tsumego problems, get conflicting answers, and then give up, blaming the other people.

That's what I'm trying to overcome.

[gowan]

There are many different types of problems. Consider two cases: problems where the position might often actually occur in a real game and problems where the position is quite contrived and probably would rarely if ever occur in a real game. There are things to be learned from each of these types. When the problem could arise in a game thinking about how it might have occurred, i.e. what sequence of moves would result in the shape. This flow would show how to recognize situations that could cause the problem shape. In the case of a problem that could not occur in an actual game, studying it would challenge read-by-move reading and how to take an unusual shape and push it into something recognizable.

[Boidhre]

To briefly add on to that, for me personally, I've found real game style tesuji problems to be where the suji is really important for me interalise the pattern and where playing problems out on a board seemed very helpful. In artificial problems it's far less about this and more about shape recognition training and I didn't feel they were doing more by playing them out. I think partially because those real game problems have a natural flow to the stones.

[Bill Spight]

An unrealistic problem.

$$Wc White to play
$$ ----------------
$$ | . . . O X . .
$$ | B . . O X . .
$$ | O O O O X . .
$$ | X X X X X . .
$$ | . . . . . . .
$$ | . . . . . . .

Click Here To Show Diagram Code

[go]$$Wc White to play
$$ ----------------
$$ | . . . O X . .
$$ | B . . O X . .
$$ | O O O O X . .
$$ | X X X X X . .
$$ | . . . . . . .
$$ | . . . . . . .[/go]

I found this position through iterative deepening search on the 3x2 eye. Even as a rank beginner I never would have considered playing in the eye. For that reason I learned something about that eye, but something that I strongly doubt has any practical significance. Still, I was pleasantly surprised.

A player who has never seen the correct play by Black in that corner, can learn the same lesson from the main failure variation of the problem, which transposes to the main variation for correct play in the empty eye. Also, a player who has never heard of an approach ko can learn that they exist.

Here is another problem that arose from an iterative deepening exercise.

$$Wc Does :w1: live?
$$ ----------------
$$ | 1 . . . O X . .
$$ | . . . . O X . .
$$ | O O O O O X . .
$$ | X X X X X X . .
$$ | . . . . . . . .
$$ | . . . . . . . .

Click Here To Show Diagram Code

[go]$$Wc Does :w1: live?
$$ ----------------
$$ | 1 . . . O X . .
$$ | . . . . O X . .
$$ | O O O O O X . .
$$ | X X X X X X . .
$$ | . . . . . . . .
$$ | . . . . . . . .[/go]

? Never in a million years. Still, this position provides a nice reading problem from which even an SDK might learn a thing or two of practical value.

Edit: If Black plays first White can make independent life, something which may not be all that obvious.

It is such lessons, I think, that provide the main value of problems. OC, the composer of the problem has one or more lessons in mind. But what about the value of reading practice?

Sakata says that reading consists of the choice of candidate plays, the calculation of variations, and the evaluation of the results of the search. Now, problems have well defined goals, so there is little to learn about the evaluation of the problem itself. However, the problem position or intermediate positions may help to evaluate similar positions in real games. The lessons certainly help in the choice of candidate plays. There are plays which accomplish or threaten to accomplish certain objectives, or subgoals of the search, and apply to real game situations. Also, one may learn heuristics for choosing candidate plays. As for the calculation of variations, research in chess (by deGroot, I think) did not find any differences in that skill between amateurs and masters in unfamiliar positions. However, I think that Kotov's experience indicates that it is possible to improve in that skill. Whatever one may think of Kotov's method, it does provide discipline in search which I think is beneficial. Not that it is the best method, but without discipline it is easy to get lost and confused. I do not think, however, that simply doing problems does much to develop such discipline.

Is doing problems the best way of learning these lessons? I kind of doubt it, especially at the low levels of skill. There I think it's better to teach the lessons first and then use problems to test the understanding and application of those lessons and to overlearn them. As for the value of effort, I do think that it is important to develop discipline, but, as Gurdieff said, why spend a lot of time and effort learning how to do something if you can take a pill that allows you to do the same thing?

I certainly disagree with the idea that the solution does not matter. After all, the solution is based upon the learnings which the composer put into the problem (or discovered in the process of composition).

[dust]

The rascal guru Gurdjieff is an unexpected reference from Bill. I haven't read any Gurdjieff myself, but thought from what I've heard he emphasised the role of constant effort and awareness.

In this connection, Gurdjieff was an influence on the late entrepreneur Nicholas Saunders who did much to revitalise the Covent Garden area of London and the whole foods marketplace in the '70s:

He (Nicholas Saunders) moved into an old warehouse in the very derelict Neal's Yard in Covent Garden and opened downstairs the first wholefood warehouse in London that sold medium bulk to the public. He was proud that their turnover per square foot exceeded Sainsbury's. The most popular items sold there led him to found a series of other shops in the Yard, ranging from the Neal's Yard Coffee House and Neal's Yard Bakery to the Neal's Yard Dairy and the Neal's Yard Apothecary. He created over 100 jobs without government aid of any kind and without any of the businesses failing. He had a belief derived from a Gurdjieff group he once belonged to, that fulfillment comes from work which is demanding, so long as it gives opportunity for variety, learning and responsibility. So rather than have a machine hoist, workers hoisted bags of grains and beans up to the first floor packing room by jumping out of the window holding the pulley rope. There was only one minor accident in 10,000 jumps.

[SoDesuNe]

I'll try one last tack. What I am arguing for is that, to learn something from doing tactical problems, the most efficient kind of practice is effortful and rich in associations. Trying to solve a problem in your head is part of the effort, I agree, but you can add to the effort by playing the moves out on a board. This gives you the flow (suji) of the stones and creates associations.

I'm with you on this. In my last post I made a short remark that replaying the moves on a physical board adds experiences (ie possible associations) but what remains is the question of efficiency (let alone practicability). More on that later in this post.

The exact method you adopt can be totally different from what I'm proposing but I don't see that efficient learning can be achieved unless your choice of method guarantees extra such as associations and flow. Mere repetition of problems that you've solved before but can't now remember is even more boring than - and actually more similar to -tracing your own steps in the sand.

Agreement with the first sentence. For the second sentence: There is some nuance. Repetition (especially spaced-repetition) is really powerful to remember stuff and for remembering facts I don't know any method more efficient (whether you enjoy this process might be personal preference?).

Now we really go full circle: The problem when learning go problems through repetition is that you have a visual (ie diagram). Since the brain does not take photographs, subconciously you will end up memorising only certain abstract elements of the visual. And here comes into play what (I guess?) Bill Spight refers to as key stones. If the way you remember the visual does not include the key stones, you can't likely claim understanding of the problem but just memorisation of this particular shape. To keep the imagine of the toolbox: You'll end up with an open-end wrench only useful for working on this problem. Been there, done that. (Anyone also reminded of the bias in maschine learning algorithms?)

So yes, (solution) diagrams might end up being a bad when you combine it with spaced-repetition. So how to proper engage with tsumegos to understand them instead of just remember them?

All of these methods will add something further to your knowledge, but some methods are better than others. For example, scribbling in the margin is often close to useless, whereas scribbling the book's ideas in your own words - self-testing, in other words - tends to pay off big time. Your internalise the knowledge.

...and I believe the same is true for tsumegos. It's not revolutionary either as far as I know: Start making your own problems! "The asians" did that all along when I correctly remember my reading of various go exchange students' experiences.

You have various ways to turn that into practice: In a book you can pose problems the usual way but instead of just given the answer on the next page you can ask the reader whether the problem would work with the same solution when you take away this or that stone. For more engagement, don't show a diagram with less (key) stones but instead ask the reader which stones they consider key for their solution to work. Now you can present the answer to both the initial problem and the key stones on the next page.

If you structure the problem book thematically you can end each chapter with a couple of empty diagrams encouraging the reader to make their own problems based on what they have seen and learnt about key stones (and ask them to upload them to your site for interaction with other players... not even discussing the business value behind this). Like this you still keep the entry level for engagement really low (even better: make an app), since no board or stones are needed (and no place to put them) and you can at least spark engagement, ie by given the reader the choice to immerse themself. I'd say this choice is better than designing a cumbersome solution and (more or less) forcing the reader to use "third party devices".

[Bill Spight]

The rascal guru Gurdjieff is an unexpected reference from Bill.

Oh, I was heavily into Gurdjieff 50 years ago. I even picked up my copy of All and Everything: Beelzebub's Tales to his Grandson by walking into the publisher's offices when I couldn't find it in any of the metaphysical bookstores I knew of in Manhattan.

One of my favorite Gurdjieff anecdotes:

Some of Gurdjieff's students in Paris complained that since studying with him they had become boring, according to their friends. Gurdjieff replied, "You haven't become boring, you have become honest."

[Bill Spight]

Repetition (especially spaced-repetition) is really powerful to remember stuff and for remembering facts I don't know any method more efficient (whether you enjoy this process might be personal preference?).

Now we really go full circle: The problem when learning go problems through repetition is that you have a visual (ie diagram). Since the brain does not take photographs, subconciously you will end up memorising only certain abstract elements of the visual.

Memory is in general reconstructive.

And here comes into play what (I guess?) Bill Spight refers to as key stones. If the way you remember the visual does not include the key stones, you can't likely claim understanding of the problem but just memorisation of this particular shape.

And not even that.

So yes, (solution) diagrams might end up being a bad when you combine it with spaced-repetition. So how to proper engage with tsumegos to understand them instead of just remember them?

These days there is a plethora of problems at almost every level of difficulty. With many problems at the same level, they will contain repetitions of ideas couched in slightly different shapes. For instance, I remember a problem where the key play was a throw-in that allowed the opponent to make an eye, but removed a liberty, and that allowed the Golden Cock tesuji to work. The throw-in to take away a liberty in conjunction with the Golden Cock occurs in Hashimoto's problem above. In the main line, however, White does not capture the throw-in stone but ataris the stones forming the Golden Cock's leg. That allows Black to use the throw-in stone to take away White's eye. Hashimoto's problem is more difficult than the one I remember, but the same theme is repeated.

[CDavis7M]

I have been keeping a log of positions/problems that I should return to. I was using GOWrite to reproduce the problems and sequences on the computer but for many reasons I found this to not work for me. I ended up just inputting the positions as Unicode text using ┼○●●┼┨. This format has the limitation that I can only setup the baseline stones and input numbers/letters that are not circled. This turned out to be a good limitation because it obscures the solution a bit, requiring more mental effort to play it out. This reminded me of "Go Wisdom" format which is very similar. I searched for "go wisdom" and ended up here.

Very interesting discussion. I'm happy I found it because I have also been looking for a good way to consolidate "correct" and "mistake" sequences. Listing the moves in multiple sequences takes less space than showing multiple board positions. Thanks to this post and to Kobayashi Tetsujiro for providing a simple and easy solution. Writing the solution becomes more tedious but I suppose it helps with learning as well.