Depth of reading correlates with rank - and other insights

[John Fairbairn]

There was an interesting series of articles in the late 90s in Igo Kansai in which an amateur working on programming go gave a long account of the insights he had thus gleaned into improving in real go. As a measure of how interesting it is, it runs for almost a year, over several pages each issue, and is almost solid text. In my experience, it is utterly exceptional for an amateur to be given such space in a major magazine. In fact, it is exceedingly rare to be given even one page in one issue. Clearly, someone in charge thought the ideas were worth airing, and I for one totally agree.

Without going into detail, but also with me adding my own take to some extent, a couple of sections were of special interest to me because they touched on subjects that I myself have tried to highlight (not just here but in the "From the GoGoD Archives" books) as being concepts that have not been properly taken on board in the west (I mean qua concept - I'm not saying there are no westerners who understand them).

One is bullying or ijime. Too many western teachers just promote the idea of getting stable groups with two eyes. Our programmer had a nice approach to this (indirectly) by introducing a new "proverb" - stop fighting only when there are three eyes. The important point is that stability as represented by two eyes in the early part of the game does not necessarily mean strength. The eyes at that stage are not likely to be well formed, even though they can be made easily in response to attack. In other words, they are virtual eyes. The trouble is that every time the opponent is attacking he is probably making a point or two profit. This is bullying. Unfortunately, in go, bullying often works. But if you have three eyes you are immune from bullying, because two of them act as miai. If the opponent threatens one he is a tempo behind - you ignore him, he takes that eye away, but you still have two left. In other words, when building positions you should be trying to make three-eye shapes rather than two-eye shapes.

The other concept is kakoi or surrounding. When I first mentioned this, some years ago, it inspired a very strong Korean amateur to remark that it was this concept that got him from 5-dan to 6-dan. It's rather hard to explain without diagrams, but our programmer (I'm avoiding his name because I can't be sure of the reading: his surname could be Eda, Echida, Etsuda, Koeda, Koshita, Koshida, Otsuda or Shiota - but Koshida is the most likely; his personal name is also slightly problematic but is probably Masatsune) points out that there are situations where there is essentially a choice between reducing the opponent's territory or increasing our own, and amateurs almost invariably choose reduction (in the widest sense, not just keshi). The main reason is that this can be done in one move, and a move that the opponent has to answer, and we all know there's nothing more potent than the drug sentepezam. But surrounding (kakoi) usually takes (or appears to take) several moves. Kakoi is indeed harder, but it doesn't really have to take several moves. If you can engineer play in a certain way you can do kakoi in a single move (there are examples in the GoGoD Go Seigen books), and the benefits are considerable. A reducing move eliminates bad aji for the opponent and in some circumstances may create a burdensome reducing group of your own. It does not give you any territory. But a well-engineered kakoi move gives you territory, eliminates bad aji in your area and has no downside. It may give your opponent sente, but you can't have that all the time.

Koshida (let it be that) does not talk only about 6-dan stuff. He indicates also that teaching at beginner level could be improved immeasurably by teaching people to work like machines and count liberties more. (Bruce Wilcox made a similar pitch, and in fact his Instant Go series was a similar interesting spin-off from computer research.) But in what I've seen, few beginners are in fact taught to check the liberty situation beyond ataris, ladders and nets.

There is one long section on a topic that was discussed here very recently - which is better: a big point or an urgent move. Koshida regards the proverb as problematical, at least from a computer's point of view, though he does give mini-algorithms for identifying both big points and urgent points.

Naturally enough, given the programming background, he also talks about depth of reading. He suggests that there is a direct correlation between go strength and how deep we usually choose to read. On that basis, he gives the follwoing table (moves = ply):

3 moves = 10 kyu
5 moves = 6 kyu
7 moves = 3 kyu
10 moves = 1-dan
13 moves = 3-dan
15 moves = 5-dan

That feels about right to me, but (in other sections) Koshida brings in two other important ideas that modify this. For one he invented his own word: girichon. This is presumably a portmanteau word from girigiri and chon (like smog = smoke + fog). It will mean something like "down to the wire". He did this to stress the point that fighting is a down-to-the-wire affair. You can't just stop halfway through. That seems obvious to the point of triteness, yet I think very many of us don't fight that way. E.g. in our reading we reach a point where we spot a nice shape for us and stop there. It's not entirely stupid - it's a sort of probablistic way of playing, i.e. a position with good shape in it is more likely to favour us eventually. But it's probably not as good as digging even deeper. The ideal is to establish not just that you have an attractive position, but that you actually finish ahead of your opponent.

The other Koshida point is that, as a measure of go strength, the difference between two players can probably be determined best in how they each evaluate any two moves. The one who can discriminate better between the two is the stronger player. That may seem obvious, but there is a touch of Columbus's egg about it, and I don't think it takes a huge amount of humility to realise that too many of us don't even try to discriminate very well. We often tend to find just one plus point (e.g. that old standby good shape) and stop there. Maybe more practice in rattling off a series of evaluations checks - like a computer - could pay big dividends.

No Mickey Mice were harmed in the making of this coffee-break entertainment.

[snorri]

As usual, excellent post, John! Thanks for taking the time to summarize.

I wonder if reading and the tendency to reduce/invade rather than expand are correlated. There is (at least in Western teaching) a kind of semi-proverb that it's easier to live than to kill. (Or easier to defend than to attack.) To kill, your reading has to be pretty precise and complete, because the player trying live usually has many options even if these are self-damaging plays.

The potentially self-damaging but living plays are easier to find in blitz games that amateurs play so much. Killing sequences take time and more reading. As reading skill gets better and players get stronger, though, maybe it balances out more because reading passes some kind of theshold.

I agree that there is sente addiction, though. Reductions and invasion, though, are often sente now / gote later affairs, so the initial feeling of sente-ness may be misleading.

Guo Juan 5p, for one, usually does teach that in the case of competing moyos, the side that is bigger should prefer to expand, whereas the side smaller should try to reduce or invade. But your amateur author may be talking about more than just moyos. Maybe you can clarify?

[Kirby]

Interesting post (ignoring "Mickey Mice", who I'm beginning to hate :-p). For some reason, this discussion reminds me of a quote from the popular, "Catching Scent of Victory". I don't recall the exact wording, but it was something along the lines of, 'Playing White 23 immediately, and playing the same move after an hour of thought may seem to produce the same result, but in actuality, there is a world of difference'.

I think that part of reading strength is not only producing a good result in a single move, but the consideration given to that move, even if it ends up being what appears, on the surface, to be obvious.

Regarding the point of evaluating different moves, I find that in my own play, sometimes my goals are off-track. Recently, a stronger player told me not to attack to kill, but to attack to maximize profit. Though this sounds easy to do, I do not follow this advice. My goal is to kill the opponent, periodically at the expense of the game result.

In this sense, even if I read some number of plys down, I will typically select the option that brings me closest to killing the opponent, even if it doesn't seem to work.

[daniel_the_smith]

My goal is to kill the opponent, periodically at the expense of the game result.

It's waaay more fun to kill a big group than to win the game.

[Solomon]

13 moves = 3-dan
15 moves = 5-dan

[lovely]

I was reading about Benjamin Teuber's time in Korea last night and came across something interesting.

From Day 7

The day began with Mr. Kwon's lessons. It was a game commentary of two famous professional players, Yamashita Keigo 9p and Ch'oe Ch'eol-han 9p. This game was played the day before, with Ch'oe winning by 1,5 points.

Certainly the technical aspect of the game was very interesting, but it was again the "other stuff", which we found more impressive. He emphasized the "fight" and the "endgame" in Go.

He said that modern Go no longer has any formal josekis or fusekis. Instead, the modern players decide to fight from the first move, especially the modern players like Yi Se-tol 9p or Ch'oe Ch'eol-han 9p. Therefore, he added, outstanding reading qualities are absolutely necessary to become great professional players. This reading skill give the modern (Korean) players the "power" which the Japanese players are desperately missing.

Looking at the latest Honinbo match and all the fighting between Hane and Yamashita, perhaps they've got a clue that go isn't all about shapes anymore. The 1 geup (we play on 7 stones (I've beaten him once by 2 points and lost our last game by 4), so I would venture to say he's a classic 1 geup (in that he's close to pro strength) and not just a 1 geup based on the other players strength) at the go salon I go to (Master Song) said something equally as interesting that really got me thinking: "Go is a game about fighting, not building houses!".

[LokBuddha]

but building house for the troop to rest isn't a bad idea

[Numsgil]

The "problem" is that if your opponent wants a fighting game it's hard to force the game to become a moyo game! Or a territory game. Whereas if your opponent wants to play a territory game it's not that hard to force it to become a fighting game.

Or if there is a way I haven't seen it

[Bill Spight]

The "problem" is that if your opponent wants a fighting game it's hard to force the game to become a moyo game! Or a territory game. Whereas if your opponent wants to play a territory game it's not that hard to force it to become a fighting game.

Or if there is a way I haven't seen it

Who says that a moyo game is not a fighting game?

In a sense, every game is a fighting game because you have to be prepared to fight. But that does not mean that you have to fight.

Tenuki is always an option.

[gowan]

I was reading about Benjamin Teuber's time in Korea last night and came across something interesting.

From Day 7

The day began with Mr. Kwon's lessons. It was a game commentary of two famous professional players, Yamashita Keigo 9p and Ch'oe Ch'eol-han 9p. This game was played the day before, with Ch'oe winning by 1,5 points.

Certainly the technical aspect of the game was very interesting, but it was again the "other stuff", which we found more impressive. He emphasized the "fight" and the "endgame" in Go.

He said that modern Go no longer has any formal josekis or fusekis. Instead, the modern players decide to fight from the first move, especially the modern players like Yi Se-tol 9p or Ch'oe Ch'eol-han 9p. Therefore, he added, outstanding reading qualities are absolutely necessary to become great professional players. This reading skill give the modern (Korean) players the "power" which the Japanese players are desperately missing.

Looking at the latest Honinbo match and all the fighting between Hane and Yamashita, perhaps they've got a clue that go isn't all about shapes anymore. The 1 geup (we play on 7 stones (I've beaten him once by 2 points and lost our last game by 4), so I would venture to say he's a classic 1 geup (in that he's close to pro strength) and not just a 1 geup based on the other players strength) at the go salon I go to (Master Song) said something equally as interesting that really got me thinking: "Go is a game about fighting, not building houses!".

Whether you "fight" or not, go is still a game of building houses. The reason is that at the end of the game the player with the most territory (=houses) wins. Yes go is also a game about fighting, but as Bill Spight points out fighting happens in many ways: moyo vs. moyo is still a contest (=fight) as is a contest of small groups in which subtle maneuvers take place to end up with more houses, as is a chaotic battle. We've all seen martial arts movies where one fighter charges wildly at the other only to be felled by a single blow. I think a lot of amateurs think that good go equals wild fighting. Yi Chang'ho plays a quiet game yet defeats the wild fighters. In order to win wild fights you undoubtedly need reading skill but you also need a subtle mastery of go technique. Finally, saying that go isn't about shape shows ignorance because without making good shape (part of haengma) you can't win fights.

[Alakazam]

Gowan's post and others: I think saying "Go is a game of fighting" is just Korean for "stop being so submissive just because you don't want to read difficult things." The only reason fighting ever happens is because of disputes over houses. If you don't pay the rent, the landlord comes in with a katana. Or a broom. Or a tooth pick. Depends on their power. Lee Changho will bring in some food or a ps3 to keep you busy and satisfied, steal something small, possibly a penny, and leave.

The only reason killing ever happens is 1: sacrifice. for territory! 2: misread or neglecting one's group... ... ...because of needing territory!

My style is very territorial, but who says that's not a fighting style? In the past I was VERY territorial, and later fighting in the center was disadvantageous for me but I had more cash. This means I must have been confident in my fighting power because i gave my opponent the handicap of influence/thickness for the fighting. Fighting could be anything from reducing to killing to invading/...many things!

Many amateur players, especially a lot of older korean go club people (the ones below 6d) have a false sense about Go. Fighting spirit is not about fighting because la-di-da. Fighting is to get 0.5+ more out of the situation! Fighting when peace is an equal option is also fine, but stupid if peaceful option is better, which it still probably is with more than half of the moves of most games.

@Araban: love your post! haha

I've been waiting to say this, read everything on this page.

5 dan = 15 moves? WRONG.... Sorry. The majority of the post content was really interesting and i enjoyed reading everything, im even going to favourite this particilar thread page, but ths rank-moves thing was way off.

I won't argue about the moves shown for the kyu levels, which looks believable and is not something I remember well, but the Dan levels were way off. 1d are usually not great at reading, so sure. But 3d can usually read more than 13 moves... one of the biggest differences between 3d and 1d is reading, but even more so... um, 5d can read at least double of what is posted there (15 moves). it would be quite sad if a player went from 3d to 5d and only improved his reading by 2 moves. the thing is, if we take one of those people who go through the ranks fast, like 1d in one year for example. all 30 kyu levels then take a year, but how long does 2d take? usually at least half a year I believe. each dan rank is almost like another 20-30 kyu levels in terms of improving time. I've never gone up a Dan in less then 3-4 months, and even then it was because I happened to have an epiphany or two shortly after the previous 1-up. Although 5d when playing online, in my experience, usually dont read more than 15-20, we can read up to 35-50 if the situation is not too ambiguous. And you know, even then, im not confident i can solve 5-6d problems in real games... I struggle with them outside of games...

The reading difference between each Dan is going to be at least 5..

mind you, i dont know what ranks we're talking about. im thinking of 5d as kgs 5d or tygem 7-8d (which to me seem similar).

[HermanHiddema]

5 dan = 15 moves? WRONG.... Sorry. The majority of the post content was really interesting and i enjoyed reading everything, im even going to favourite this particilar thread page, but ths rank-moves thing was way off.

Although 5d when playing online, in my experience, usually dont read more than 15-20, we can read up to 35-50 if the situation is not too ambiguous. And you know, even then, im not confident i can solve 5-6d problems in real games... I struggle with them outside of games...

The reading difference between each Dan is going to be at least 5..

Nonsense. Yes, sure, anyone can read a ladder 50 moves deep, but that's not the point. This is about reading out all the reasonable variations to that level in complex fighting.

In Nakayama's "The Treasure Chest Enigma", he describes a scene where Kitani Minoru comments after the game about a certain move that he looked at it, and it turns into ko. None of the present professionals have a clue what ko he's talking about, until he shows, to their amazement, a sequence of 37 moves that leads to a ko in the corner.

So 37 moves is amazing to a group of professionals, but you can read 35-50? Right.

Here's a straightforward life and death problem. No variation is longer than 15 moves, as there is simply not enough space for that many moves. Tell me how long it took you?

$$B Black to play and kill.
$$ --------------
$$ | . . . . . . .
$$ | . X X X X . .
$$ | . . O . . . .
$$ | . . O X X . .
$$ | . . O O X . .
$$ | . . . O . . .
$$ | . . . O . X .
$$ | . O O O X . .
$$ | . X X X . X .
$$ | . . . , . . .
$$ | . X . . . . .
$$ | . . . . . . .

Click Here To Show Diagram Code

[go]$$B Black to play and kill.
$$ --------------
$$ | . . . . . . .
$$ | . X X X X . .
$$ | . . O . . . .
$$ | . . O X X . .
$$ | . . O O X . .
$$ | . . . O . . .
$$ | . . . O . X .
$$ | . O O O X . .
$$ | . X X X . X .
$$ | . . . , . . .
$$ | . X . . . . .
$$ | . . . . . . .[/go]

[John Fairbairn]

Alakazam: Herman's made the point in one way. Let me make it in another way. The article said the number of moves was "how deep we usually choose to read", not "can read". "Can read" in fact has another meaning, which seems applicable here

But the point about fighting is spot on. Many people choose to confuse fighting with power go: if you can cut, cut; if you can defend; invade; if you can make a 100-point territory, kill five stones. Fighting is a necessary skill for all go players. The only debates are about when and where to start it. Amongst amateurs, power go is a personality defect.

[RobertJasiek]

re John:

A need for three potential eyes is only a special case. Rather much more generally, there is a need for being at least 1-alive (see Joseki Vol. 2 Strategy, p.64), i.e. having the option of playing at least 1 time elsewhere before an attack has to be answered.

I do not agree that amateurs would invariable choose reduction over enhancement. Rather many kyu players have a tendency for either and do not dare to do the other or they fail to reflect the alternative at all. I teach that the bigger effect on territory differences per invested excess stone, the bigger of reduction versus enhancement should be chosen. - It is not the 5d to 6d killer concept: I apply it but apparently I am missing something else for 6d.

Overall, I do not see that in the west the concepts would be missing. They are there and even in much greater generality but they are not generally spread here yet.

re snorri:

Guo's approach that the bigger side should expand while the smaller side should reduce the opponent is also simplifying. Again, difference of both players' territory changes on a per excess move basis matters.

re All:

Reading depth of a 5d: John appears to have meant an average depth of interesting moves. 15 is not that bad a guess. Some moves are so obvious; reading depth 1 suffices. So an average of 15 might actually be about right. It does mean though that often a 5d reads MUCH deeper, although he does so, of course, SELECTIVELY.

[kirkmc]

I wonder if reading and the tendency to reduce/invade rather than expand are correlated.

I would say that counting should replace reading in your statement. I think that's a big difference between strong players and the rest of us; they can count, and they can know whether expanding or reducing is appropriate at any given point.