One-way streets (ippon michi) can be the basis of practical reading. If you practice reading out long unbranching sequences you will reach a level where no conscious effort is required, the stones in your mind flow the way the stones in an SGF reader do when you use the slider. So it takes very little time to read out a 100 move unbranched sequence. Now if you have a practical problem in a game then you do know a goal, such as make life or kill, or connect out. You can follow unbranched sequences until you come to a branch point. This would take, perhaps, one second. Keep this up, as a structure of nested loops, until you determine whether you reach your goal. In a life or death problem we aren't looking at hundreds of sequences, and it doesn't take time to check out even twenty variations if you speed read all the unbranched sequences.

The late Japanese pro Nakayama Noriyuki is famous for constructing long ladder problems (e.g. http://senseis.xmp.net/?NakayamaLadderProblem). In this example the solution takes 177 moves! A long ladder might have branch points but until reaching those points it is a one-way street. That is, the solution is a concatenation of one-way sequences.

I think this is where many people go wrong. Obviously it's not untrue, and it works in chess, and it appeals to (dare I say it again) number-oriented people. But I recall a wise Irishman, a customs officer at Heathrow Airport, who answered a query from a passenger about "How do I get to the first floor?" The reply was "Go upstairs."

This was wise, not insolent, because the officer knew that the first floor in England is what some people from other countries call the ground floor.

I think we need this sort of wisdom about reading. In other words, the first step is to decide on our goal and how to recognise it. The reading tree we construct to get there is then no longer a "search tree" in the sense that such a search takes place inside a chess program, which simply chooses the line with the highest number, and even humans tend to follow the material. Go is too complex for that approach. In go we simply need a line that reaches our goal without obvious obstructions. This in itself is a fantastic pruning mechanism, and ultimately it's also what a Monte Carlo program does (win by half a point; forget everything else), although with obvious important differences.

A goal-oriented search has other advantages. One is that it can become a matter of recognition (of the goal), and often a recognised pattern can be broken down into smaller recognised patterns, all of which speed up the search.

This has already been recognised (pun intended) in Japanese studies of L&D. A crude, though remarkably effective, approach they use is to divide problems into three levels:

1. Problems where there is one theme, typically an instant tactic to solve them (e.g. placement, or nakade, on the vital point of a nakade shape). Usually easy.

2. Problems where two themes are required (e.g. first a throw-in to create a makade shape, then the nakade). Usually medium difficulty.

3. Problems where three themes are required (more is very, very rare). These are usually hard, but can be "coached" by showing the first theme, or the first two.

Studying themes like this leads to recognition of other useful information. For example, certain themes tend to be combined in a specific order, or certain themes, or combinations of them, tend to lead to kos (or sekis). As someone pointed out above, I have included a proposed long list of themes in Gateway to All Marvels (all cross-referenced to the the several hundred problems).

Further, one way that some think is useful is to compose problems on the basis of themes. This certainly reinforces knowledge and recognition of the themes.

Although crude, there are higher levels you can tune into. This was well illustrated by Maeda Nobuaki (the God of Tsumego) in a series of articles he wrote in the 1950s. I was reminded of this a couple of days ago when someone asked me a question about "rules" for problems, not recognising that the accepted conventions we use now are due to Maeda.

He was not teaching L&D directly, but as the editor of the tsumego prize quiz in Kido he was overwhelmed by a sudden surge in interest and started to find that solutions proposed by readers were not actually wrong, but jarred with him as a pro. He therefore set about devising a set of "rules" that essentially we use today. These guidelines were based on how a pro approaches a problem, and can be seen as a refinement of the themed approach.

My comments here relate to L&D but they can apply to tesuji situations just as well - the priority is still to work out what you want to achieve. First get your two bits of wood, then fix a nail, and then hit the nail with the hammer. Lashing out with a hammer, as in the old "If I had a hammer" song, and forgetting the nail, in the hope that you might accidentally join two short planks together is just being as thick as said two short planks.

Goal oriented search is one of the things I had in mind when I said that there is no single answer to the question of what reading is.

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

Visualize whirled peas.

Everything with love. Stay safe.

When I was 9k I started some lessons with someone and when we went over tsumego for the first time, the conversation went something like
>What's the best move?
For black or for white?
>Go is a game for two players. We should find the best move for both sides.

You never ask that question again when you get that answer. And it really is true. Black to play, white to play... the situation is unsettled and we should find the best move for both sides. The place to kill is not always the best place to live. Sometimes there are multiple places to play, but perhaps when one player realizes the situation is fruitless and plays elsewhere (in a real game) you leave behind a smaller or larger ko threat. Maybe you can live with three points one way or four another.

Far be it for me to take a position that might conflict with a pro, but I think reading involves simply being able to see ahead, be it one move, three moves, or maybe thirty in the case of a ladder. In some sense the pattern-training life and death style drilling allows you to jump ahead ten moves. xed_over brings up a good point about knowing the "right" move but not knowing all the failure branches, though. Pattern drilling memorization can be a little misleading in that way, but I don't think it is really a crutch, just unfinished business.

I don't think there is a problem. It's important to have feedback, psychologically, for the cases where you don't know how to proceed. There is value in self-discovery but not so much that giving all answers is going to be some kind of intrinsic roadblock. Even math books give you (traditionally) odd-numbered solutions. I could imagine a life and death book where problems are grouped thematically and a few problems have answers which introduce new tesuji or placements and the rest don't have answers and it is up to the individual to judge which to bring to bear. But I think there is a real danger of training the wrong answer if the right ones aren't provided. In the beginning you should play a lot of games and get a feel, but eventually you should probably seek stronger players' help before you reinforce the wrong ideas too much. I think that tsumego could have similar problems if used for drilling.

This seems problematic. Most of the problems I come across (e.g. Cho's or some tsumego app I downloaded recently for my phone) just give a problem and it is up to you to see what you can do. Ko? Kill/live? Seki? Go is a game for two players. It's good to know that you can live in gote and also find a sente seki. Usually living is considered bigger than seki in tsumego but is that really something we should enforce? Maybe you can tenuki and turn a live shape into a ko shape. Seems valuable to know this.

Anyway, one feature that the app I use has that is nice is that it doesn't keep presenting you with the same problems over and over again. If you miss them they reappear in the working set, but once you solve them in the set it is pushed back to an hour, and then if you succeed again it is gone for the day from the working set. So the working set is eventually filtered to those problems which present you the most challenge. I think this is really a great feature and should be the core of any drilling software. I think an option for a timer should also be given, something like standard byo-yomi periods and quantity which the user can select; tenuki is always an option, which resets the timer and moves to a different problem. Sort of like the time attack mode at goproblems, only without the ridiculous whole board endgame problems or crazy ladder problems that appear out of nowhere and ruin you (and the fact that the timer doesn't reset when you make a move). I think a third feature is the ability to tag problems so that themes can be explored specifically or searched for in case you want to find a problem for others.

Unbranched sequences of obvious moves play an important role, and techniques can sometimes be used to construct them. However, there are problems defying an overly simplistic approach. Quite a few intermediate to difficult problems require methodical reading of lots of variations. Despite still applicable simplifications, not everything resolves to unbranched sequences and reference to prior knowledge of shapes and techniques.

Three ideas that have been mentioned, spaced repetition, thorough paths of refutation and categorizing are a must, but before re-inventing the wheel, we should recognize that these ideas have already been implemented in one form or another. Spaced repetition is available, though poorly implemented on the android app go tesuji, and I also recall hearing that Guo Juan's training program involves spaced repetition of tsumego. Boidhre mentioned GoChild as a site where wrong paths are followed, and I would also mention WBaduk's android app which does the same, and also offers hints at all junctions(several possible moves, all with continuations). As to categorizing, some problem collections, for example 200 Tesuji Problems by Shirae Haruhiko are sorted to compare similar situations.

I would like to see a coaching program do something different. Although a bit utopian, I would like to see it work in a way similar to a real-life high-level sports coach. By this, I mean that the coach should verbally train you to be thinking about the right thing at each stage of solving a problem. What are the questions you should be asking yourself? Are you asking them?

I had this idea, because I recently discovered that while trying to read a go problem, I had developed a peculiar and useless habit of counting moves in my head, meaning that as I try to read a sequence, I would say to myself "one, two, three, etc.," in an attempt to help visualize the stones. I think this came about from reading solutions where the moves are so numbered. Upon realizing that this mechanical chanting, similar to chanting "black, white, black, white etc.," was not helping me remember which stones I had mentally placed, I started saying the liberty count of the strings affected by the stones I placed and I'm finding that this helps my visualization and understanding of the moves. The problem is bad habits are hard to break. I'd like a coach to say: "liberties?" after each move.

This may be a bad example, because not everybody has the same lousy habits, and maybe my idea of counting the libs after each move is not what good players do, but I am fairly sure that good problem solving habits exist, usually combined with recognizing what type of situation one is facing. A coach should train you to recognize where a hane might be a likely first move or whether you should first be considering a placement. Ideally, the coach should remind you what you should be saying to yourself, be it "Where is the vital point of this shape? Does it work to play there first?" or whatever is actually appropriate from the perspective of someone who solves problems well and efficiently.

_________________
Patience, grasshopper.

Out of curiosity, when you try and visualise as you read do you "see" the stones on the board? I'm just asking since this is something I'm very conscious I can't do at all - when I'm reading I'm literally just trying to remember where the black and white moves are (rather than visually imagining them there) which mostly works but as soon as the variations start to branch I lose track pretty quickly.

One idea to improve this aspect of reading that I've recently wanted to try is playing blind go. I thought a useful utility would allow me to play against an AI with just an empty board, except that when a move is played it flashes on the screen briefly before disappearing. It would have a "reveal" button that would, when pressed, reveal the whole board briefly.

I try to do exactly this and found it pretty frustrating that few people talk about WHAT they do when "reading". Less "reading" more "painting", I used to lose track rapidly, whenever I tried to "talk" stones into place (you are here, next stone is there). Good practice for this kind of reading are imo long, non-standard ladders, see Gowan's post. Don't listen to detractors, it may well be "not reading" for professionals and high-dans, but you would be surprised how many players well into dan levels "feel out" moves based on knowledge and experience instead of going through these little one lane sequences.

I talk about seeing, but I don't try to visualize. Everybody is different. Not visualizing does not mean that I don't see things. I once took a test where one of the tasks was to use painted blocks of wood to recreate a pattern. I set up the last and most difficult pattern in 12 seconds, and the tester remarked that the average time for that pattern was 2 minutes. So there is some sense I have of an "image", but it is not a visual image.

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

Visualize whirled peas.

Everything with love. Stay safe.