Life In 19x19

A bit on the second skill, choosing candidate plays

In general, it seems to me that a major flaw among SDKs is never even considering the right play. You can calculate as many variations as you want, but if you don't make the right play to start with, none of them will be right.

OTOH, the main thing in choosing candidate plays is to eliminate bad plays. Otherwise, the game tree becomes too large, and you end up calculating too many irrelevant variations. But doing so risks eliminating the best move. This is a dilemma that amateurs face.

IMO, the best thing is to learn what good candidate moves look like, and doing problems is not the best way to do that. Imitation is.

John Fairbairn wrote:One reason that playing over lots of pro games is so valuable is because of the wealth of "Eh, I didn't know you could do that!" moments.

Indeed.

You can also learn what good play looks like from books that demonstrate and explain good play.

It is also possible to combine approaches.

John Fairbairn wrote: Meijin Inseki thought that the best way to teach tsumego skills was to show a technique and then just give a pile of problems with the same technique. Drills, in other words.

From what I have seen of Segoe's Tsumego Dictionary, he seems to adopt that approach. All of the problems that I saw featured the eye stealing tesuji.

Also, even though I have only seen parts of the book, may I recommend Redecker's Tsumego Strategy? (See http://www.lifein19x19.com/forum/viewto ... 57&t=13854 )

Training techniques is insufficient for the ca. half of the tactical problems for which there is no technique or none is relevant. Then one must apply general reading methods, starting with Regular Reading, which avoids much negative reinforcement.

Bill Spight wrote:Chess grandmaster Kotov advised in calculating variations to calculate each branch of the game tree only once. Obviously, that avoids dithering. That advice has generally been rejected for actual play. However, as a discipline in solving problems I think that it is excellent. You choose your candidate moves at each turn and read each branch to whatever depth seems right at the time (or to the limit of your ability), and if you find no solution, you are done, and you have not reinforced any bad moves. Now you are ready to look at the solution and failure diagrams.

It's amazing how many times I have been stuck for ten minutes on a tsumego problem, then say "OK, some move has to work, let me just try all of them in turn", and the problem is solved within the next two minutes.

When I play chess and I lose focus during my opponent's turn, or if I am totally stuck during my own turn, I sometimes quickly look at every single legal move, not even evaluating it, just acknowledging that it exists. It often turns up interesting ideas and possibilities. Of course, this is a little more time-consuming in a go game

John Fairbairn wrote: Meijin Inseki thought that the best way to teach tsumego skills was to show a technique and then just give a pile of problems with the same technique. Drills, in other words.

I picked up a nice little Japanese book (though I think it might have been translated from Korean) like this the last time I was at Kinokuniya in New York. It focuses on very simple capturing race situations and there is a chapter in which every answer features a descent to an edge, a chapter in which every answer features a diagonal descent to the edge, etc. Most people would probably consider the problems 10k+ level but I like them since at 3-4k I want to learn to be able to see the right move instantly in a game (or, even better, in a variation). I think they will be great for my flashcards. I don't have the book at hand but I'll post the ISBN later.

The Jump Level Up books are also like this. One nice thing is that when you see the same technique a dozen times in a row you start to give a name to it (e.g., "capturing 3 makes one eye").

"10 problems a day" is quite unspecific. I usually start my day by doing 10 problems, but it only takes me a few minutes. weiqiok.com has 4 daily problems and Tsumego Pro has 6. Most of them I solve at a glance, the hard ones might take me up to a minute. So, how much time are you spending on problems?

swannod wrote:I disagree about reading silly lines. To me doing problems are about finding the opponent's *best* counter. A move looks like it works but no it doesn't. In my experience when I misread a simple problem in my game it boils down to I choose a move that looks good on the surface but I didn't find the counter. Or a I choose move that works but I didn't read out what happens when the opponent plays the best counter leading to a mistake.

This is already a ton of work and you have to do this under time pressure. Reading out unlikely lines doesn't seem useful to me (unless your reading says none of the seemingly likely lines are working)

Gotraskhalana wrote:The advice "to read out every possibly sequence" is simply a lie which is easy to check by calculation. Noone can do this.

There seems to be some kind of misunderstanding. This thread is about doing tsumegos, so I advocated reading completely a problem while doing easy tsumegos. I don't see how "time pressure" and "noone can do this" applies here.

One benefit I personally noticed is that in actual games, even if complete reading is not possible (nor desirable), I now read more candidate moves, not just in tsumego situations but also in middle game fights for instance. In other words, I think that forcing myself to read everything, WHEN PRACTICING EASY LIFE-AND-DEATH PROBLEMS, created the habit to instinctively look for more possibilities in actual games. This seems to be a good solution to the problem swannod describes so I don't really understand the flak here.

Reading the right sequence completely until the end is not the same thing as reading every possible sequence.
The former is necessary to understand the solution of the problem, not just guessing right for this time. The later is impossible, there are thousands of legal sequences from any local position.

I am always puzzled when people talk about the right way to do tsumego. Usually, different approaches lead to different benefits (It was the same when studying arpeggios on the piano).

There are tsumego and tsumego. I am not even talking about difficulty or the pre- or absence of solutions. There are tsumegos on paper, in apps that anwer moves, on a board, from a selection that drills a skill, from a selection that conveys many different ideas, tsumegos whose hard move is number 3, tsumegos that go deep and those that go broad, etc.

And there are different approaches: how fast, how deep, when do you give up, what do you do with the solution, what kind of help do you use

Change is spice. I try to rotate approaches. The question "what is the most efficient way to improve ones game by doing tsumego?" should only concern those who want to improve fast. I like the scenic route as long as I am enjoying it. When doing arpeggios, the scenic route might not be the quickest, but is the one that's gonna pay off in the future. I am not sure about tsumego, but I'm probably still hoping deep inside..

For many, the idea of doing tsumego probably has got nothing to do with tsumego but with getting better at visualising sequences. And I assume, me being no different, many usually will go to the next problem once they looked up the solution. So, after failure and checking the solultion, I like the approach of turning back the page, and reading it out again. Can I read it out now? Why was it difficult? Is there a common shape problem lurking? Was there a blind spot? How many refutations are sensible and do I have them covered?

dfan wrote:

John Fairbairn wrote: Meijin Inseki thought that the best way to teach tsumego skills was to show a technique and then just give a pile of problems with the same technique. Drills, in other words.

I picked up a nice little Japanese book (though I think it might have been translated from Korean) like this the last time I was at Kinokuniya in New York. It focuses on very simple capturing race situations and there is a chapter in which every answer features a descent to an edge, a chapter in which every answer features a diagonal descent to the edge, etc. Most people would probably consider the problems 10k+ level but I like them since at 3-4k I want to learn to be able to see the right move instantly in a game (or, even better, in a variation). I think they will be great for my flashcards. I don't have the book at hand but I'll post the ISBN later.

It's 978-4-488-00069-1.

bayu, there are many ways of studying tsumego but only certain general methods can solve those problems for which all more specialised methods fail. Hence every way of studying must include at least one generally applicable method and acquiring the skill to apply it.

More on the skill of evaluating the resulting positions of reading

Also on reading everything out, including "silly" lines of play

One advantage of tsumego problems is that the goal is generally given, so that the leaves of the game tree can be accurately evaluated. (That is true for a game of go, as well, but what a game tree! ) Does that mean that the skill of evaluation is not necessary? I don't think so. Having that skill means that you don't need to read variations out to the bloody end.

Here is the problem that Thofte found difficult:

Click Here To Show Diagram Code

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

No goal is given, but it is (ahem) obvious that White to play can live, so we may take it that Black is to play and kill or make ko.

What about reading everything out? Well, there are 10 open points in the corner, so there are potentially millions of variations. However, many of those include illegal moves or end at a shallow depth. So the true figure may be only a few hundred.

Are you kidding? Yes, reading out hundreds of variations could be a bracing exercise, but doing so ten times a day?! Even twice a day is too much, IMHO. Twice a year, maybe.

However, the figure may become manageable with proper evaluation of non-final positions. For example:

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------
$$ | . . . . . . . X . .
$$ | O O . 2 1 O O X . .
$$ | O X O O O X X X . .
$$ | X X X X X . . . . ,
$$ | . X . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .[/go]

Now, I consider a silly play, but reading out the variation is not particularly a problem if you stop after . My evaluation is that White is (obviously) alive. Not that this positions is devoid of interest, but we do not have to explore this branch of the game tree any further.

If you asked me to defend that assertion, I might point out that White is alive just in the very corner.

Click Here To Show Diagram Code

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

That is, with the indicated four point eye.

Really? Let's see.

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------
$$ | . 4 3 6 5 . . X . .
$$ | O O . 2 1 O O X . .
$$ | O X O O O X X X . .
$$ | X X X X X . . . . ,
$$ | . X . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .[/go]

plays within White's indicated Lebensraum and secures the one point eye in the corner. But then after cannot secure the second eye because it is self-atari.

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------
$$ | . 4 3 . 5 . . X . .
$$ | O O 6 2 1 W W X . .
$$ | O X O O O X X X . .
$$ | X X X X X . . . . ,
$$ | . X . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .[/go]

is correct, OC. Black cannot save . (She can capture the stones in gote.)

So I was right, but maybe it is not so obvious. OTOH, if you do not already know about , isn't this problem too difficult for you?

Anyway, looks pretty silly, itself. Wouldn't it be better just to take ?

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------
$$ | . a 3 7 4 6 5 X . .
$$ | O O . 2 1 O O X . .
$$ | O X O O O X X X . .
$$ | X X X X X . . . . ,
$$ | . X . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .[/go]

Yes, but. In this diagram is a disaster. OC, it should be at "a". delivers the coup de grace.

So, yes, there are lessons to be learned from playing out silly variations, but is that why we are doing this problem?

Edit: There is another lesson here.

Click Here To Show Diagram Code

[go]$$B
$$ ---------------------
$$ | . . 3 . 4 6 5 X . .
$$ | O O 7 2 1 O O X . .
$$ | O X O O O X X X . .
$$ | X X X X X . . . . ,
$$ | . X . . . . . . . .
$$ | . . . . . . . . . .
$$ | . . . . . . . . . .[/go]

is a cute way to kill.

Edited for clarity.

Thank you everyone for your replies!

So the main question was how much time I spend on a problem. Well that really depends: I have heard about that 3 minute rule (If you can't sove it by then, move on). So now about 30 minutes for 10 problems seem correct, but before that I'd spend an hour on 10 problems.

But with the 3 minute method I don't see any progress. Sure I won't be that frustrated anymore about a difficult problem, but the ones I skipped last time I still can't solve now...

From what I read most of the people seem to advice acctually looking at the solution after trying for some time. So do we have a consensus that Tatsuki acctually isn't that great of a resource?

And also some people adviced to just learn solutions for problems by heart, if I got that correctly? So no reading just knowing then?

To the original questions I'd like to add two points. First, when near or at dan level, for instance improving at a rate of 1 stone a year is not bad at all. In two months, that would be about 0.2 stones. It would be something very hard to notice, but still meaningful.

Another thing, when I used to do very intensive go study for some time, I did not feel any improvement immediately. But then maybe 4 months later, after dropping the tsumego study I felt I'm stronger than I thought. One can imagine different explanations why this could be a real effect. Maybe you need to separately learn to apply the new skills on the board. Or maybe your brain keeps processing something over months, growing new synapses or whatever, and it takes time to actually realize the reading power.

Thofte wrote: But with the 3 minute method I don't see any progress. Sure I won't be that frustrated anymore about a difficult problem, but the ones I skipped last time I still can't solve now...

Did you miss them because they were too difficult? At the right level you should miss about half of the problems. (As I indicated, I think that one I just talked about is too difficult for you right now. Too many options.)

Why are you missing them again? I assume that you remember what doesn't work, and so you try other things.

It may help to forget about the three minute rule and follow Segoe's advice to set up the problem on the board and try to solve it by playing variations on the board. Then, if you still fail, look at the solution.

I'm also a new beginner like you, but I may offer advice from my own experience. I initially did some problems, but ended up rejecting them (for now). I felt that I was not ready for it.

I put aside all books/problems, found a friend who was better than I am and just cranked out 100 games as fast as possible (aka 100 losses). I feel like this improved my reading more so than solving problems. What I did memorize were all the joseki that I could get my hands on. Went on Eidogo and ran through all the joseki and memorized all the common ones.

I recently went back to problems and everything made a lot more sense to me. Everyone may be different but just my two cents.