Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 3:34 pm
by PeterPeter
At the level of life-and-death tsumegos that I am currently doing (around 15 kyu level, so it says), for most of them there are only a few reasonable first moves, and then all follow-up moves stay on one or two narrow paths until success or failure. Therefore, you can almost find your first move by trial and error. My question is: will this approach work with more advanced problems, or should I get into the habit of solving them by trying to understand the key features of each position first?
(I am talking about trial and error as in: mentally choose a reasonable-looking first move by trial and error, then read it through to a conclusion, and NOT randomly click everywhere (including follow-up moves) until it says Solved. Only about 20% of the time is my instinct for a solution so strong that I click without first reading it out.)
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 3:57 pm
by emeraldemon
I think what you're calling "trial and error", most people just call reading. I don't think "understanding the key features" of a position can ever replace mentally checking each possible response from your opponent.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 3:59 pm
by jts
That sounds right to me. It won't work with more advanced problems right away, but by the time you finish your beginner problems and move to intermediate problems, you'll have much better intuitions about which moves to start with, and you'll be able to assess the follow-ups more quickly and accurately.
A few things to think about, though - when you say "all of the follow-up moves stay on one or two narrow paths", do you mean there are only one or two choices for each move, or only one or two choices that you think are reasonable? It makes sense to get into the habit, even once you're sure you have the answer, of trying at least one move that looks stupid, or to look at a sequence with the same moves, but in a different order. Sometimes the result is quite surprising - I'm sure you've encountered zwischenzug in another game...
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:01 pm
by Joaz Banbeck
The trial-and-error tree will soon get too large. You have to work on appreciating the character of the position. This will keep the size of the tree managable.
Of course, 'managable' is a relative concept. At least one pro has claimed to have read 100+ moves deep.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:04 pm
by SoDesuNe
Hm, I am a bit confused by your using of trial and error but after reading your short little disclaimer I think you mean the "right" thing. That is, mentally playing out a sequence and only if you think you will end up with a good result it gets physical.
I tend to use trial and error when I just click at the first best move and after a few moves I can fairly easy see if this works or not. Well, not so efficient for training reading ^^
But yes, this approach stays the same. At least it did for me. You will learn the key features of certain shape automatically after solving a lot of problems because the same shapes (and therefore Tesujis, respectively vital points) keep reoccurring.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:05 pm
by Tommie
PeterPeter wrote:At the level of life-and-death tsumegos that I am currently doing (around 15 kyu level, so it says), for most of them there are only a few reasonable first moves, and then all follow-up moves stay on one or two narrow paths until success or failure. Therefore, you can almost find your first move by trial and error. My question is: will this approach work with more advanced problems, or should I get into the habit of solving them by trying to understand the key features of each position first?
(I am talking about trial and error as in: mentally choose a reasonable-looking first move by trial and error, then read it through to a conclusion, and NOT randomly click everywhere (including follow-up moves) until it says Solved. Only about 20% of the time is my instinct for a solution so strong that I click without first reading it out.)
My own experience (3d EGF) is - trial & error YES - BUT in the head please ! If you just click - you will learn few from it.
You can train your first reflexes ('instinct') = narrowing down potential eye shape, hane + vitale points etc. - in such an order by doing many exercises. Often I 'see' the possible THEMES and my work is to verify whether it works that way. In a book problem that's easier, in a real situation it might not work or be another theme or a move misses etc.
When you can refrain from 'eating the Marshmallow' (google that metapher for psychology) AND you have solved a problem correctly, THEN you experience a moment of satori.
There are many good sites on senseis for a sound L&D-approach.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:14 pm
by PeterPeter
Here is an example, with black to move and live.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------[/go]
To me, on an initial review, a, b, c, d and e are the moves that look like they might be solutions, and so worth reading.
So, I read them all in turn, until I can read one to an irrefutable solution. This is a bit time-consuming, but I get there in the end. Sometimes I dismiss the correct move too early, then have to re-consider it when I cannot get any of the other options to work.
I am not sure if this is practical with bigger or more complicated problems, where there might be twice as many reasonable-looking moves.
Do stronger players use any rules of thumb, or features of the position, to reduce my list of 5 down to 2 or 3?
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:19 pm
by PeterPeter
Tommie wrote:My own experience (3d EGF) is - trial & error YES - BUT in the head please ! If you just click - you will learn few from it.
Yes, absolutely, I tried to explain that.
Tommie wrote:There are many good sites on senseis for a sound L&D-approach.
I had a look, but I could not see much on a general approach to these sorts of problems. Do you have a link?
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:42 pm
by billywoods
PeterPeter wrote:So, I read them all in turn, until I can read one to an irrefutable solution.
Absolutely the correct way to do it.
PeterPeter wrote:Do stronger players use any rules of thumb, or features of the position, to reduce my list of 5 down to 2 or 3?
I doubt it. Here are my thoughts:
Let's try a first. a is quite an easy one to refute.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | 1 C W X X O. . . . $$ | T X C C . O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | 1 C W X X O. . . . $$ | T X C C . O . . . $$ ------------------[/go]
Black has one eye in the corner. Where's the other eye going to be? It has to be at one of the circled points. A moment's thought will show you this atari:
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | 1 . W X X O. . . . $$ | . X . 2 . O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | 1 . W X X O. . . . $$ | . X . 2 . O . . . $$ ------------------[/go]
[/quote]
and so white captures two stones and none of those circled points belong to black any more. (Pushing in from the outside at also works.) So what about b or c, to give the marked white stone fewer liberties?
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . 1 O X X O. . . . $$ | . X . . . O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . 1 O X X O. . . . $$ | . X . . . O . . . $$ ------------------[/go]
White has no hope of making a dead shape inside black's group. There are only really three moves to consider: the descent (immediately obvious it doesn't work), the hane and pushing in from outside. Again, as before, both of the latter two work:
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . 1 O X X O. . . . $$ | . X 3 2 4 O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . 1 O X X O. . . . $$ | . X 3 2 4 O . . . $$ ------------------[/go]
Even though black captures, white can just connect, and the two stones are in atari, hence a false eye.
The problem here is coming from black's shortage of liberties. Those two stones have very little freedom to move due to the marked white stone. Let's see that in action once more with c:
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . . O X X O. . . . $$ | . X 1 2 . O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . . O X X O. . . . $$ | . X 1 2 . O . . . $$ ------------------[/go]
White again can't hope to make a dead shape, so white has to capture the two stones. Check all the obvious points and you see easily that most fail. But white can play this throw-in and now black can't capture due to his shortage of liberties (it's a snapback). Here, pushing in from the outside doesn't work.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . . O X X O. . . . $$ | . X . . 1 O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | . . O X X O. . . . $$ | . X . 1 . O . . . $$ ------------------[/go]
and indeed you can check that black does indeed get one eye in the corner, one eye on the side. Notice the double table shape, by the way.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 4:46 pm
by billywoods
PeterPeter wrote:My question is: will this approach work with more advanced problems
Of course the more complex the problems get the less brute force you want to apply, but you are doing things correctly for now. Once snapbacks and throw-ins and things like that are clear in your mind, you can worry about doing things more efficiently and spotting old familiar shapes. For now, just solve lots. The shapes will become more familiar in their own time.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 5:16 pm
by Bantari
Joaz Banbeck wrote:At least one pro has claimed to have read 100+ moves deep.
I heard that too, and it has always puzzled me. What does it mean? - he read 100 moves along a singular branch? (i can do that too) - he read 100 moves along 1 branch and this is how the game went (luck?) - he read 100 moves along all the branches? was the position complex or just one-way street in yose? hard to imagine this happening in fuseki... - and so on...
You can imagine plenty scenarios which would make this feat relatively easy or very difficult. Without knowing more, it is hard to judge if one pro reading 100+ moves deep is actually something amazing or something normal....
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 6:11 pm
by Bill Spight
PeterPeter wrote:Here is an example, with black to move and live.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------[/go]
To me, on an initial review, a, b, c, d and e are the moves that look like they might be solutions, and so worth reading.
So, I read them all in turn, until I can read one to an irrefutable solution. This is a bit time-consuming, but I get there in the end. Sometimes I dismiss the correct move too early, then have to re-consider it when I cannot get any of the other options to work.
I am not sure if this is practical with bigger or more complicated problems, where there might be twice as many reasonable-looking moves.
Do stronger players use any rules of thumb, or features of the position, to reduce my list of 5 down to 2 or 3?
In the future, you will solve similar positions at a glance.
Now, since you have solved this problem, let me ask you a few questions.
What are the key features of this position? When you were solving the position, which move looked best at first? Why? What was wrong with it? As you worked on this problem, did you pick up clues about where to make your first move? After Black's initial correct play, what is White's best play? (From an endgame point of view, OC. ) After Black's initial correct play, are there any White moves that are obviously useless? Why? Having solved the problem, can you go through the main variations quickly? Are you confident that you have a reply for every White response?
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 9:16 pm
by Shaddy
PeterPeter wrote:Here is an example, with black to move and live.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------[/go]
To me, on an initial review, a, b, c, d and e are the moves that look like they might be solutions, and so worth reading.
So, I read them all in turn, until I can read one to an irrefutable solution. This is a bit time-consuming, but I get there in the end. Sometimes I dismiss the correct move too early, then have to re-consider it when I cannot get any of the other options to work.
I am not sure if this is practical with bigger or more complicated problems, where there might be twice as many reasonable-looking moves.
Do stronger players use any rules of thumb, or features of the position, to reduce my list of 5 down to 2 or 3?
I do not understand how, but when I looked at this problem and your a-e, I immediately dismissed all but d without even really thinking about it. Some part of me recognized it as the only reasonable shape, and there was no conscious reading. It might be that I've internalized that problem, but I think there's also some kind of unconscious shape rules-of-thumb happening here.
edit. I'll add that this isn't something I've practiced, it's just the result of having done thousands of life-and-death problems.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 10:07 pm
by Joaz Banbeck
PeterPeter wrote:Here is an example, with black to move and live.
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . O . . . . . . . $$ | . . . . . . . . . $$ | . O O O . . . . . $$ | X X X O O O . . . $$ | a b O X X O. . . . $$ | . X c d e O . . . $$ ------------------[/go]
... any rules of thumb, or features of the position, to reduce my list...
At first glance, I see this:
$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . C . . . . . . . $$ | . . . . . . . . . $$ | . C C . . . . . . $$ | C C C O O C . . . $$ | . . O X X O. . . . $$ | . C . d . O . . . $$ ------------------
[go]$$B $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . C . . . . . . . $$ | . . . . . . . . . $$ | . C C . . . . . . $$ | C C C O O C . . . $$ | . . O X X O. . . . $$ | . C . d . O . . . $$ ------------------[/go]
I think 'belly tesuji is what he wants'. Therefore, 'd' goes to the top of my list of candidates, on the theory that a good point for my opponent is a good one for me. After about 5 seconds of reading, I have assured myself that it works. Next, I ask myself if I can get more territory by living with a different move. I analyze 'e' for about 1/2 second and conclude that it does not work. I never consider 'a' or 'b' or 'c'.
The important point here is that I have an internal library of shapes, and when I look at a position, they jump out at me, as in the diagram above.
Re: Solving life-and-death tsumegos by trial and error
Posted: Wed Mar 06, 2013 10:20 pm
by Kirby
Bantari wrote:
Joaz Banbeck wrote:At least one pro has claimed to have read 100+ moves deep.
I heard that too, and it has always puzzled me. What does it mean? - he read 100 moves along a singular branch? (i can do that too) - he read 100 moves along 1 branch and this is how the game went (luck?) - he read 100 moves along all the branches? was the position complex or just one-way street in yose? hard to imagine this happening in fuseki... - and so on...
You can imagine plenty scenarios which would make this feat relatively easy or very difficult. Without knowing more, it is hard to judge if one pro reading 100+ moves deep is actually something amazing or something normal....
also works.) So what about b or c, to give the marked white stone fewer liberties?
