Life In 19x19

palapiku wrote:Reading involves not only considering different variations, but also knowing when to stop, and which variations not to read at all because they're stupid (the technical term for this is "pruning"). Humans are very good at discarding stupid move sequences, and computers are very bad. Since the go board is so big, this is a big problem (unlike Chess, where considering ALL moves many turns ahead is a reality).

"Reading", "tactics", "strategy" are all human-specific terms, and saying "computers are strong at reading" is trying to fit a square peg into a round hole.

I'm fully aware of the branching problems but I don't buy this.

This appears to me like you're smuggling something else into reading which is why what computers do doesn't fit the concept. You even gave it a name: Pruning. In any case, it's more useful if we divide it up for the sake of the thought experiment:

How does one determine good pruning? Retorting 'good reading' is question begging. Tactical heuristics regarding eye shape for instance? That seems like it makes sense to me.

But what about non-local moves, over on the other side of the board. We prune on a global scale too. Wouldn't that be strategy and isn't it pretty important?

Joaz Banbeck wrote:

daniel_the_smith wrote:...
Who says reading can't be global?

The tree-trimmer.

I can't believe I'm the only person here that does "reading" in fuseki mentally placing stones all over the board??

daniel_the_smith wrote:

Joaz Banbeck wrote:

daniel_the_smith wrote:...
Who says reading can't be global?

The tree-trimmer.

I can't believe I'm the only person here that does "reading" in fuseki mentally placing stones all over the board??

I try to do it methodically. Black A1, White B1. Hm, interesting. Now, how about, Black A1, White C1. Getting better...

daniel_the_smith wrote:
I can't believe I'm the only person here that does "reading" in fuseki mentally placing stones all over the board??

That would be pretty odd alright.

Maybe it's more like they don't think strength of reading is as important because one isn't reading in extreme depth, and one will play better fuseki if one plays a consistent and intelligent strategy which is supported by reading and good choice of local play/joseki (instead of the other way around).

It's really hard to say though. Everyone seems to be using terms differently (myself included).

Pros do read the fuseki to extreme depth (and breadth), that's why they spend so much time on it.

palapiku wrote:Pros do read the fuseki to extreme depth (and breadth), that's why they spend so much time on it.

It was my understanding that this whole discussion spiraled not about whether or not it was necessary to do serious and in-depth reading at pro-level play, but whether or not it was possible to advance through much of the kyu ranks by focusing on developing strength in making strategic judgments (of the negative pruning variety and presumably also the positive selective variety). Also on whether or not reading and tactics needed the supplement of strategic strength to get to very high levels of play.

In kyu ranks it's certainly possible to get ahead of an opponent with stronger reading by having a better idea of strategy, and win that way.

But really this is just trickery - the reason you'd win would be because your opponent is playing suboptimally against you. Playing optimally would require him to start a huge fight right away and take advantage of his better fighting skills. The reason he doesn't do that (assuming he wants to win) is just because he's not aware of your tactical disadvantage or its implications. Once the opponent adjusts to you, you're toast. Of course, playing online, on a peaceful server (KGS), this may never happen. In effect, your opponents are being unreasonably nice to you.

This is why people are surprised by their rank drop when they play on a Chinese or Korean server. People there fight more.

Well, there are a lot of posts about this, and I don't feel particularly argumentative at the moment.

It doesn't really matter to me how strategy is defined. I think my main point is that, personally, I do not find it useful to study go theory at all, really. When I have seen improvement in my game, it is from reading.

Sometimes this reading is global, and sometimes local...

But if you want to study something, forget theory books and do some go problems. That's what's worked for me so far.

Studying go theory just makes me sleepy.

topazg wrote:

Kirby wrote:In both cases, I think that a fundamental reliance on reading will get you further than what you read in a book on go theory.

You'd be surprised how much your reading can be improved by books on Go theory. Reading is a sense of shape, a sense of tesuji, and visualisation skills, of which I would have said the first two of the three comprise 90% of reading ability and are very easy to improve with Go theory books, and the latter is improvable only with practice.

How often have you found a corner/edge shape where you can throw in followed by picking the key eyeshape point in what's left, or something similar? How do you know about the throw in? What about the key eyeshape point, what made you see it? Most of reading can be greatly improved by studying tesuji techniques, and shape techiques (moves that avoid being squeezed for example) as opposed to simply trying one problem after another.

Both strategy and reading/tactics require a fair amount of effort on both the theoretical side and the practical side.

For contrived illustration, try solving the tsumego below (Black to kill):

$$Bc Black to kill
$$ - - - - - -
$$ . . X O c d |
$$ . . X O b a |
$$ . . X O O e |
$$ . . X X O O |
$$ . . . X X X |
$$ . . . . . . |
$$ . . . . . . |
$$ . . . . . . |

Click Here To Show Diagram Code

[go]$$Bc Black to kill
$$ - - - - - -
$$ . . X O c d |
$$ . . X O b a |
$$ . . X O O e |
$$ . . X X O O |
$$ . . . X X X |
$$ . . . . . . |
$$ . . . . . . |
$$ . . . . . . |[/go]

How many people read out sequences starting with "b" through to "e" as the first move? If not, why not?

I will reply to this one, though. You can achieve this by simply doing go problems. You don't need to read a theory book to tell you where the key points are.

Kirby wrote:

topazg wrote:...

$$Bc Black to kill
$$ - - - - - -
$$ . . X O c d |
$$ . . X O b a |
$$ . . X O O e |
$$ . . X X O O |
$$ . . . X X X |
$$ . . . . . . |
$$ . . . . . . |
$$ . . . . . . |

Click Here To Show Diagram Code

[go]$$Bc Black to kill
$$ - - - - - -
$$ . . X O c d |
$$ . . X O b a |
$$ . . X O O e |
$$ . . X X O O |
$$ . . . X X X |
$$ . . . . . . |
$$ . . . . . . |
$$ . . . . . . |[/go]

How many people read out sequences starting with "b" through to "e" as the first move? If not, why not?

I will reply to this one, though. You can achieve this by simply doing go problems. You don't need to read a theory book to tell you where the key points are.

Reading does help.

I once had an amusing game against a 6D with a bulky five in the corner. It was all that remained of the eye space of a rather large group of his. I, of course, had played in the vital point.
That group of his was surrounded by an even larger group of mine which had no eye space at all. It had six external liberties. He knew, without reading, that filling a bulky five takes 8 moves. I had played one move at the vital point. So he figured that he was ahead, 7 liberties to 6.
I hadn't memorized the numbers. So I read it out. I came up with a different count. I thought that I was ahead.

The remainder of the game was bizarre. We both belived that we were the winner of a 100+ point fight, and that the other guy really ought to resign. Both of us were too polite to request that the other do it. He played wild hyper-aggressive attacks. I defended, conceding a point here and there just to guarantee stability. I assumed that he was making desperate attempts to get back in the game. He - as he told me later - figured that the game was decided in his favor and that he would test out a few ideas on attacking.

You see, in a corner, a bulky five requires only four moves to kill:

Click Here To Show Diagram Code

[go]$$Bc Black to kill
$$ - - - - - -
$$ . . X O b . |
$$ . . X O a X |
$$ . . X O O c |
$$ . . X X O O |
$$ . . . X X X |
$$ . . . . . . |
$$ . . . . . . |
$$ . . . . . . |[/go]

I knew this because I had read it out. He didn't because he had relied on memorized numbers. When I played 'a', he resigned.

Yeah... I guess I will add that I cannot really deny that "strategy" - however you define it - is "bad" for you, unless you rely on it instead of reading.

I have some go books that are not problem books. To me, I never feel like I am learning anything when I read them. I'm just trusting what some other guy says.

I prefer to just do problems... I guess that's my main stance.

...What?

Strategy is not bad for you. The scenario JB described had nothing to do with strategy, but rather failed tactics.

Kirby wrote:I will reply to this one, though. You can achieve this by simply doing go problems. You don't need to read a theory book to tell you where the key points are.

You are missing the point somewhat though. The point is, you know where that key point is, and you don't read the rest because you don't have to. It doesn't make any difference whether that came from doing Go problems, or reading a book, it's still "go theory" where the weak point in a bulky five is, however you learned it.

I remember solving a 2d problem with a 3p Korean pro, and having spent about 30 minutes on it with a 6k, a 4k, and a 1k (I was 5k at the time), we finally were pretty sure we'd cracked it. We took it to the pro, figuring that he'd be able to to it without problems. He took less than 2 seconds and picked out the right first move and the not at all obvious to me second move. The point? He barely read a thing. His shape and tesuji awareness (both based on theory, even if it is theory supported by hours of practice) were strong enough that the two necessary moves just were completely obvious to him. The only thing he said he had to read were one or two possible unorthodox resistances to the second move, and when neither worked, it was obviously the solution.

Reading is rarely being able to visualise 80+ move sequences. It's normally reading the right first move, and then the right second move, and ignoring the other stuff that doesn't work. A lot of this knowledge can be acquired by theory, and honed by practice. Much as with carpentry, plumbing, electronics, or pretty much every knowledge+practical skill based pursuit - you can trial and error your way to the knowledge 1 fail at a time (with lots of flooded houses and broken bits of wood), or you can get your theory as strong as you can by studying the material, written by others who have the experience already, and also gain the experience and knowledge of a seasoned practitioner yourself by practicing at the same time (still with flooded houses and broken wood, but hopefully a bit less of each).

Otherwise, you are just reinventing the wheel from scratch each time.

daniel_the_smith wrote:Who says reading can't be global?

Then please tell me, a or b? And please show me the sequence that tells you why.

Click Here To Show Diagram Code

[go]$$
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . X . . X . . X . O . a O . . |
$$ | . . . O . . . . . , . . . . . X b . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . O , . . . . . , . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . X . . . . . . . . . . . O . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]

Strategically, the decision is relatively easy (of course after reading out both sequences). But I have no idea how I would continue to read globally to reach a conclusion without using strategic principles...

I worry there is some confusion concerning semantics.


My 2 bits though

I think, in answer to what I think was the original question ; can you win despite normally (always?) getting a locally inferior result if your timing and direction are better?

Yes or no, depending on how much worse your reading is and how much worse their timing and direction are.


If you can't read your way out of a wet paper bag*, it's a problem. If you compulsively make weak groups next to your opponent's thickness, it's a problem.

*Or to reliably kill invasions that you know don't work in theory, for a more go-relevant situation.


If you don't get better at reading as your rank improves, your local results will get worse and worse. Similarly, if you keep choosing the wrong fights to pick, you will wonder why the game is such an uphill battle.

Of those two, though, in my experience improving reading takes longer but yields more impact.

Eh, I might as well continue this wall of text, to illustrate; I have a friend who read some book on how to play high Chinese and never lose, detailing many correct strategic decisions within that opening. However, if opposing moves were weird, he had a lot of trouble finding "refutations", as he did not strive to improve his own reading much (he is fixing that, these days). I think he plateau'd around 10k for that reason.

I'm in Aph (etc)'s camp, otherwise, though, just for the record.

Edit: By the way, if you already said one or more of the things I've written already, please consider this post a homage especially to you.