I was in the Royal Box of the Royal Opera House in London yesterday, watching Sleeping Beauty. It's OK, I don't mind if you still call me just "Sir."

This included a visit to the Royal loo. AI hasn't quite reached there. Like many things royal in the ROH, the loo fittings are Victorian and the staff have to scour e-bay for sparts that may come up in auction, and they "fight tooth and nail."

Walking back home (I gave the carriage drivers and postilions - or at least those not struck by lightning - the afternoon off ahead of a busy weekend), I was inspired by the ballet to think a bit about a post from Robert that I had read before I left home. I was then further inspired by the following post from a different, anonymous contributor:



I thought, with some reservations, this poster (here and in the rest of his post) was on the right lines, whereas, as is very well known I disagree profoundly with his approach to tsumego (and many other things ), and I also disagree specifically with his assertion, connection with the new KBA book, that author Yoon Young Sun was wrong to say that mastery of her book would make the pupil a 5-dan at tsumego.

I suspect Robert has taken the view that he's worked hard enough (has even written a book on reading?) but has never reached 5-dan, ergo Young Sun is wrong. I haven't reached 5-dan officially yet either, and haven't worked hard, but I believe her.

This post is a coffee-break attempt to explain why. But before you get the coffee and biscuits, stand up (and hold a chair for safety, if necessary) and do a plié. That means, put your heels together and turn out your toes as far as you reasonably can, then bend your knees down (again as far as you reasonably can), then stand straight again. Your legs will make a diamond shape and (if you are like me) your mouth will make an OMG shape as you try to get back up. You can now go back to your coffee, but as it's Coronation time, do make sure that you warm your biscuits to the right temperature. King Charles has a man to do this for him, after all, and we must get in the mood. If you are not a New Man you can ask the spouse to do this (but, again, safety first!).

Are you sitting comfortably? Then I'll begin.

My musings started with the ballet, as I said. Specifically, I was pondering a comment by the great choreographer (and teacher) Georges Balanchine that a plié must be done with no pause between going down and rising, and it must be done by pressing with the toes. I going to hazard a wild guess that you did it like me - with a pause and maybe a gasp, and by pressing down with the bum.

The bum/toes press I'll ignore here as not really relevant, but it was the subtle point about not pausing that entranced me when I first read it. The point is, this is ballet - a dance to music. To achieve that you have to go down for two bars and come up for two bars. If you put a pause in, you are going to be out of synch with the music. Moreover, when you do it properly, it becomes part of your so-called muscle memory and you can do it intuitively. And you can then listen to the music with more attention and start to interpret it. You become a dancer as opposed to someone who is dancing.

Because I'm a go player, my mind flitted over to how this might offer some insights for go. How you can become a tsumego expert as opposed to someone who does tsumego problems.

I'm not going into much detail of my own thoughts. My hope is simply to stimulate you into thinking about the issue for yourself. You may wish to use no analogies or different ones, but oriental martial arts offer good examples (e.g. those who do the forms without the proper breathing) and things like music where the technique of playing an instrument cannot be allowed to interfere with interpretation are ideal. The important point is that the analogy should be something where time (or lack of it) is a major factor. I think go is somewhat confusing for we can too easily fall into the trap of thinking that we have time to think. I know we are actually given time to think, even in blitz games, but that time should not be used reading out variations by brute force. Having watched professionals do tsumego problems they have never seen before, it is quite obvious that they spend next to no time. They may hesitate for psychological reasons ("is it a trick question?") and they may even fail occasionally - the famous Yi Ch'ang-ho problem comes to mind. But by and large they are on auto-pilot with intuition or muscle memory (i.e. our own AI) doing all the heavy lifting. Even when there is an appearance of deep thought about a life & death situation of a capturing race, the pros' thoughts are more likely to about evaluating the size as opposed to moves elsewhere, or trying to find ways a line that doesn't quite work can be made to work, and so on.

Now, if we want to get into that way of "thinking" (i.e. not thinking) we have to try to find a way of helping out or building up our intuition. "Pattern recognition," as mentioned in the quote above, is often cited as the way to go. I'm not quite sure about that. I think it's more accurately "pattern understanding."

The problem with "pattern recognition" is that it leads us down cul-de-sacs. Usually this is in the form of dreaded "lists." I just checked one major example on SL and my suspicions were confirmed. I looked up "nakade" and sure enough found a list of nakade shapes that the student is, presumably, expected to learn. Learning such lists might seem to do no harm but I think they actually stunt development. What I believe the student should be looking at all sorts of shapes and trying to find not a pattern but what links all these patterns together, namely a specific kind of weak point. Once you can do that, you are not bound by the list and you can do Dosaku's famous 19-point makade and others at the drop of a nakade hat.

But even that is not sufficient to explain the problem with pattern recognition lists. In real tsumego, a single move is insufficient. There is rather a number of themes (or thematic movess, if you are a chess player), which normally ranges from one to three. Three usually means a dan-level problem. But this does NOT mean consulting a list of themes. It means learning them in such a way that you recognise which themes can follow which, or which have to precede, and so on. Even if you managed the aforementioned plié without the slightest grimace, I'd happily predict you couldn't possible do a grand jeté from there. You'd have to learn what needs to go in-between. What this amounts to is that tsumego is very much a dynamic process. It's to do with flows, not patterns. Flows are learnt by practice, practice, practice. And you don't practise until you get it right. You practise until you never get it wrong. And that is what mastery means.

So, I can turn now to why I think Yoon's book can provide a platform to 5-dan. Many years ago, I started collecting articles from magazines and books about the carpenter's square. I obviously had in mind a book. But as time went on, and the pile grew and grew, I noticed there wasn't much duplication. The killer for me was seeing a new series by Kitani. I already had a year-long series by him and was both started and dismayed to see that virtually every variation in the new series was different. That was when I first understood the proverb "If you know the carpenter's square you must be 1-dan." I did already know that this proverb was coined in the days when there was just one scale for go grades and 1-dan then meant 5-dan nowadays (true pros started at 4-dan).

And don't forget. That was just one corner position. So, I find it very easy to believe Young Sun's assertion that if you "master" (her word) a whole series of corner positions, you can be 5-dan. You just need to practise, practise, practise until you never get them wrong.

Of course, there are good an bad ways to practise, practise, practise. I believe that thinking about doing tsumego in the dynamic way adumbrated above is a good way. I personally would also recommend the way James Davies shows in his Life & Death book (looking at how positions are affected by extra legs, liberties or hanes) is another. In the absence of proof, I just don't believe that glaring at a position to come up with a list of candidate moves each analysed by a variation (which is just another word for an item in a list) is possible for a human, especially with time limits.

I would therefore like to sum up by saying we need to get rid of lists. However, having watched an entertaining documentary last on Franz Liszt, I have to admit we need more Liszts.

As to L+D, I have not worked hard enough (for above 5 dan level). My L+D skill when playing is 5 dan level because it is similar to the opponents' 5 dans level, but weaker than 6 dan level and stronger than 4 dan level. Your conclusion on Young relies on wrong assumptions. See also further below.



Expert knowledge and problem solving skill can differ.



Sure enough. It is noteworthy, of course, that brute force is the wrong means. Not only computer reading may be pruned but human reading ought to be pruned as well, albeit differently. In particular, it is sufficient to find one next move with the optimal solution so further alternative next moves may be pruned. (Reasonably) exhaustive reading can, however, be necessary to affirm non-existence of a good next move.



Usually. That is because they are much faster in tactical reading and endgame calculations. If necessary, they do both and take their time (if there is sufficient thinking time).



No. This is not what they have told me, although quite a few have admitted to use "intuition" a lot but they know that, if necessary, they have to analyse carefully. Only few of them have explained tactical analysis well but they sort of know how to do it properly nevertheless. For endgame, they were more explicit; none suggested "intuition" there (unless sufficient when one move is obviously larger) but everybody hinted at exploring variations and / or calculating values.



This myth is bad because many problems cannot be decomposed into patterns, techniques or the like but require proper tactical reading.



Some problems can be decomposed into chunks. However, see before.



This is closer to being correct but it is insufficient to get the solution right but one must also get the preceding analysis right.



The problem is: her book's series is an insufficient series for (my and my 5 dan opponents') 5 dan L+D knowledge and skill, especially at the range 3 - 5 dan. There are lots of 3 dans which L+D knowledge inferior to the scope of the book but I cannot recall any such 5 dan. 5 dans challenge each other with more difficult problems.



That is one reason why you have not reached 5 dan. (Plus your brute force exaggeration.)

Yes, lists might be the key!

Somehow you have to get the list into your head, but without writing down the list and memorising it, because that's a toxic process.

I remember being forced to learn languages at school. The teachers would literally give us sheets of paper with lists of words to memorise. It was a losing strategy for me. Surely not any sane person's idea of fun. Learning Japanese as an adult, though ... well, I didn't get much further, because life had plenty of distractions by then. But I was learning words in context, with a purpose in mind (about to visit the country, hoping to be able to function at least on the level of a 3 year old child), the mode of learning was different, and it felt a lot better. I did end up with a list of about 300 words in my head, without ever having written it down as a list.

And of course I like your music analogy :-) I think I can claim "native level fluency" in piano playing, and there is certainly a long list of scales and chords inside my head. But it comes from immersion in the language, not from systematically memorising a list of shapes.

I'm still not there with tsumego (and may not ever get "there", wherever it is). I've recently cracked open the Maeda books again, and am wondering if this sort of "artistic collection" could be a good middle ground between the chaos of games versus an encyclopedic list of shapes and variations?

(And no, we don't need to go Chopin for more Liszts. One at a time is plenty.)

Watching a pro game is like watching a beautifully choreographed ballet, while watching me play is like watching a clumsy bear stumbling around a beehive.

I may not be as smooth as a dolphin gliding through the water, but I'm still pretty adorable as a bear cub tumbling around the Go board.

Now we need a video showing John Fairbairn and Robert Jasiek side to side doing a plié, so we can judge who has the best technique.

I can assert that my subconscious has a far superior . On more than one occasion my dreams have produced songs, writings and philosophy significantly superior to what My conscious mind could produce in tens of amount of time I slept, or perhaps at all.

I was shocked when my ability to read tsumego improved in a way after watching pro games. It turns out that the way I was solving tsumego in the past was completely wrong, I thought pattern recognition was only important in whole board positions, but it turns out it's just as important in local reading. You need to have good intial prune of candidate moves, that's why you don't try every possible first move in a tsumego, that's why a plain brute force computer algorithm won't work for weiqi, shogi, chess, so not recognising the importance of intuition was daft from me, although not uncharacteristally so.





Seo Bongsoo needs to take 2 or 3 stones from Park Junghwan, so old age makes you regress at least 1 or 2 stones, so if we control for Age, John Fairbairn's ability is significantly Superior. I'd say when Vanessa Wong won the British Go Championship Matthew McFayden was already weaker. Or was it Matthew Cocke? You'd think it's Korea or China with all the duplicated names. I suspect their pilé would be at an equal objective level, and again control for age I suspect Fairbairn sensei to be the whippersnapper. I would add that I would hardly expect intuition to be a word used trying to teach someone who can't walk or ice skate how a person who's done it for years might do it! They're talking about calculating for tsumego they find hard. But the point is even tsumego they haven't seen before can be easy and not require conscious calculating because the intuition they have developed does the conscious calculating for them.

This works only for easy or some relatively easy tsumegos. The difficult tsumegos defy the possibility of subconscious thinking revealing the combination of finding and understanding the solution. In other words, subconscious thinking might find correct first moves accidentally in a few of many difficult tsumegos but does not find correct first moves in all of many difficult tsumegos. For difficult tsumegos, conscious tactical reading is mandatory.

There are mistaken views in both quotes, I submit.

First, you are not supposed to start by trying to find or prune candidate first moves. You are looking for candidate final positions. Once you know how the tsumego is likely to end, then you only have to select moves that get you to your final destination. So, to take what is often considered a difficult category of problems, under-the-stones, if you can see that you can make a zigzag shape on the edge, you are already halfway there. Or you can see that a rooster on one leg is plausible. You know, through previous study, that such solutions can often be rather long. But because you know the final destination, you know you'll get there in the end.

Of course, you need to check the solution, in which case it is not a matter of "conscious tactical reading is mandatory" but, rather, "some conscious tactical reading is mandatory." There's a huge difference. "Some" is very often tantamount to "trivial" as in the typical 'rooster on one leg' problem.

Even where a devisor attempts to create a difficult problem where the usual patterns + sequences seem to work but don't, that in itself creates a pattern. You should therefore start not by reading every move, but by looking for the usual tricks that invalidate the usual sequences (e.g. different order of moves). If you buy a book of such problems by one author, you can also learn to recognise that devisor's style by the end of the book. Furthermore, if you think about it, the very fact that the devisor is tricking you by making the usual sequences not work implies that he believes standard sequences that do work must exist.

It's best to think of tsumego as a journey. It's like finding your way in the street. You want to go to the nearest supermarket. You know your destination but don't know ho to get there. You ask someone. He says, "Go to the traffic lights, turn right, go past the library, and it's either the second or the third street on the left." He doesn't tell you wait tillt he lights change to the right colour, to be careful not to trip over the curb, to stay on the pavement, to avoid stepping in the dog poo. He doesn't tell you how big the library is, whether it's clearly marked or what sort of library it is. He expects you to do "some" tactical reading - you have to ckeck which final street to turn into it. And it's possible there's more than one supermarket down that street - your interlocutor hadn't told you which was the best one. He hasn't even told you how far it is. But still you set off in total confidence. And 9 times out of ten find it your destination with ease, even if was a bit further than you hoped. And even if the shop was closed and so your journey was pointless.

Most amateurs would be over the moon if they could do 9 out of ten tsumego problems reliably. It involves work, but the work is done at home, at one's leisure, by continually feeding final positions into your subconscious brain. This brain is an expert at finding patterns that you are not aware of. That is how humans work. It's essentially how AI works, come to that.

And by doing even more such work you can move from 9 out of 10 to 99 out 100. But, of course, the law of diminishing returns kicks in and most of us are too bored or too busy to do that extra work. But for exactly the same reasons, few of us would even consider the insane work required i, say a problem with 11 empty points, to try calculate 11x10x9x8x7x6x5x4x3x2x1 moves. Even with some pruning.

But, you know, even with that combinatorial approach you can make big inroads by using final positions. I have to say this with diffidence, not being a numbers guy, but if you use the "meet in the middle" approach to tree searching, and search simultaneously from the start and from the end and see what you get when you meet in the middle, I believe your computer gets to your destination much quicker thsan by brute force from the beginning.

But for humans even that is out of the question. Best advice for us is: start at the end.

How do you atart at the end and do - only - some reading for it in the following cases, provided knowledge of shapes or techniques does not or hardly apply?

- The correct move is a tenuki.

- The correct solution, even if it does not depend on ko threats etc. in the environment, is neither unconditional life of all stones nor unconditional death of all stones.

Be practical and apply what was said above: "Most amateurs would be over the moon if they could do 9 out of ten tsumego problems reliably."

A game of go is not about trying to explain the 1/137 fine structure constant.

As long as you give advice to 5 kyu and below, every advice is more or less helpful. Stronger than 5 kyu, my mentioned aspects become increasingly important already for solving 50% correctly.