Perhaps due to JF's influence, I pay more attention to the words used in Go. The obvious questions for researchers are

"What are the general sorts of sequences of narratives and words used?"
"What are general concepts and why are they necessary/sufficient? Which of standard concepts are redundant to better play, and which are missing?" i.e. what is the connection to optimal play.

I have already mentioned the most basic of how groups/eyespace are deeply connected to optimal play, but most are more loosely connected.

Drawing inspiration from https://katagotraining.org/sgfplayer/ra ... es/745568/, here are some ideas that may lead to answers. Note my bias and belief in influence functions/weak groups as the next step in formal theory beyond tsumego. I think there must be something more concrete than qualitative words like this is urgent, wider, etc.

I can't pretend to have followed much of what you have written so far, dhu. Leaving aside the occasional mathematicalese, the urge to slap a number on everything is not one I share. But you appear to have at least an equal interest in the words of go, and I am willing to admit to an equal urge to slap a label on everything. I therefore will share some thoughts I have had recently. They seem to bear on what you want to talk about, even if you end up talking about it with the aid of an abacus, rather than an ABC.

The background is that I have often approached linguistic problems by creating and analysing a corpus. I have done this in quite a large way for classical Chinese go texts. I have done it in part for Japanese (especially) and Korean, though never to enough degree to call it scientific. Still, over time, one does acquire a certain sensitivity to the words being used, and I sense that there is a major shift under way in the vocabulary is being used, at least in Japanese texts. All under the influence of the AI revolution, of course.

Changes can be of several kinds. One kind is the more regular use of a word that is not new in go, but which has recently acquired more prominence. An example is "overconcentration". Much AI play can explained away by efforts to overconcentrate the opponent. This was picked up early on by pros, and is certainly still a major theme, but has become more nuanced. This area is still fluid.

The opposite kind of change is also happening. There has been an argument that josekis don't exist any more, as they only last a week under AI. A move to use "opening sequences" instead seems to be under way.

Another kind of change is an attempt to what might be called changing the imagery. An example is the old standby of "kill two birds with one stone" (I think that proverb may even have entered English from Japanese), which often occurs as meritorius good play in standard commentaries. But under AI, it seems the idea we have to use now is "kill three birds with one stone".

A further change is where a concept emerges and is talked about while a new name is debated. The concept may be new or newly emphasised. I think the best example is the old Chinese concept of zhao ying (call & response) where a stone is played in a central(ish) position acting as a lighthouse to guide friendly groups to safety, which may include connection, of course. It was very important in old Chinese go because of group tax, of course. It never made it to the level of a concept name in Japanese, though it did occur from time to time under various various guises, such as "lend a helping hand to", "offer succour to" and so on. It's never been rare but it's nevr been common either. But in one recent commentary I came across it SIX times (expressed in six different ways). It is of course a standby of AI play.

As a spin-off from this, perhaps, at the tactical level there seems to be renewed emphasis on magaris. I say "renewed" because of course Shuei was famous for his magaris (i.e. L shapes) in the centre, where they naturally act like zhao ying lighthouses. To cases in the past week spring to mind because of the unusual imagery: a magari worth a thousand pieces of silver, and a Tennozan magari.

Yet another kind of change occurring is redefinition of terms. It think the best example here is a recent noticeable increase in the use of teatsui instead of atsui for thick. The te- prefix adds a nuamce of hand-crafted, done with care, etc. In more idiomatic English we could talk of thickness (atsui) and proper thickness (teatsui). The difference is best seen in what was for most of us the eye-opening Direct 3-3 introduced by AlphaGo. When the outside player seals in the invader, AI tends to leave the position "unfinished" by omitting a final hanetsugi that, hitherto, everyone, including pros, played by rote. It has now been realised that this hanetsugi produces a position (for the outside player) that used to be called just thickness but would now be called proper thickness. Now, the wall produced before the hanetsugi is seen as just thickness and is attackable. The significance of this seems to be that, in the AI age, a sequence that produced thickness is evaluated as not good (even if not bad), but a sequence that produces proper teatsui sequence is seen as not bad, and may even be good.

But one now has to be chary about using the word "attacking." New phrases such as chasing and harrying seem to be coming into favour instead, and attacking itself seems to be in the process of being redefined.

That must be seen in the context of another change, maybe the most important. The main idea now of those who have good technique (i.e. pros) is to keep the initiative. Not sente. You have the initiative if you control where the next play is, whether it's by you or the opponent. This applies throughout the game, but is seen and talked about a lot in the opening. Even Sumire has talked about, making an extension that was one point wider than the traditional norm. The old thinking was that the wide extension invited invasion - bad. The new AI-influenced thinking is precisely to invite the invasion and so have the initiative - good! In the game in question, against Rina, Rina eschewed the invasion (reasoning that Sumire had studied it) and tried to take the initiative in another part of the board. In another, quite different commentary a phrase was highlighted, very unusually, in bold: Keeping the initiative is the key point. I have not seen a standard term yet for this initiative-stressing type of play, but of course it existed centuries ago in Chinese: jin 紧. I have not come up with a good English equivalent. Assertiveness may be a passable one. Whatever you call it, it seems to be characteristic of AI play and Japanese pros seem to be experimenting with it a lot.

All these ideas (and others I have) are speculative. I now have neither the time nor the inclination to create a proper corpus. Likewise, while I have toyed with the idea of making my case stronger by writing it up here in detail, with examples, the state of discussion here makes that a fool's errand. In any case, even if I happen to be on the right lines at present, the whole business is so fluid that some new observation or discovery next week could turn everything on its head. And I don't follow the Chinese or Korean scenes closely any more, so I can't speak reliably for what is happening there.

Still, my ramblings here may offer YOU some food for thought.

thanks.

I think I have heard of zhao ying, but I think I've heard hu ying (mutual support?) more often. I can't say I've ever thought in terms of Go Seigen groups, but the topic of magari is certainly difficult enough and valuable enough for me to have invested quite some time on between 5d-6d (practical play, not general theory), and I was praised by Tanguy for one in the game I that I won against him.

Much of what I have written lately on L19 is speculative even when correct. The formal things I am sure of I have written up in papers on my wordpress.

Good point on jin. That is quite a general concept. I think Ilya mentioned it criticising an idea I had, though I think he said "tight". I interpreted it as tighter moves are necessarily to hold onto the fighting advantages, even (or especially) when you have weak stones. It tends to appear in commentaries as

jin3 cou4 紧凑

where cou by itself suggests leaning on the opponent, though it is often negative or critical. cou4 diao4 zi4 seems to imply falling into the opponent's plan/trap.

Based on things I heard about Japanese criticising Western ideas as focused on static objects and ignoring dynamic objects (I can't remember where from or if it was go related), perhaps there are things missing such as "using up the value of strength":

In an unsettled shape, you refuse to let the opponent cut without a fight. However, if the opponent prepares with a move, then you have used up the value of your local control (i.e. opp weakness) to defend for one move, so that if the opponent attacks on the next move, perhaps you should just give way.

In general, perhaps we don't talk enough about balancing weak points or moyos against each other. For example, one way to reduce a moyo is to make a large moyo next door so that if your opponent expands you may get large territory by responding and if your opponent invades, you get valuable support at the boundary helping reduce or counter invade.

I think there is also a subtle concept of ji1 min3 机敏 (lit. alert and resourceful) which tends to mean well timed exchanges squeezing the most from the opponent's weak points. In a formal theory, I would describe this as you have many forcing moves on the opponent's weak point but it isn't big enough that your opponent has already defended. So with two moves in a row, you have a very big threat, but there isn't much yet if your opponent just responds. In general, this is equivalent to "there is bad aji". Saying the opponent's weak point tends to imply the opponent can fix it all with just one move (perhaps even actively). However, with the initiative, you press on one side and the opponent struggles to counter with the strongest local move since they risk a double attack of yours. But their normal response doesn't fully defend their weak point, so it means you can still lean on the other side, and perhaps even take gote for a valuable settling of the area around the weak point. The point is that if you aren't as skilled, you will just play the gote without profiting first, so a successful harassing is much easier for a commentator to notice than for a player to carry out (with all the reading/judgement to support it). It is like playing a small sente move. Even endgame theory can miss this because mathematicians are used to independent endgames, so you don't need to play a small sente as the opponent won't defend (and sente gains nothing until it does). However in opening/middlegame, many values of endgame are connected. Perhaps the sente only exists for a short period of time when the weak point exists, and pretty soon it will disappear not because the opponent spends a move to secure the weak point, but because they play a big move that indirectly defends against it. This is related to probes.

I suspect it may be true that many key subtle concepts are in fact (as ji min) related to pointing out that the board isn't 1d, and 2d has some extra features.

Whereas more general concepts are related to how weak points move around a group and compensate each other.

In this light, you have interpreted jin more in terms of the values of moves rather than shapes on the board as usual. But perhaps that is Bill's influence, and CGT's influence in general, since it connects to successes in economics etc. and it isn't so easy to explain what exactly it corresponds to on the board. The 1d line of numbers is easier to discuss than the 2d space of Go.

If I can try to connect the 2d space to 1d numbers, I would say "jin" means: you must play moves that have a large threat on the opponent's weak groups especially if that threat will remain long into the future or the opponent can only defend by playing slow moves that help you fix your weak points. Such is much better than the opposite "song" 松 (loose/slack) which moves more safely into empty space (which certainly helps attack your opponent's weak point), but which is at least half a move away from a direct threat because it isn't close enough to your opponent's weakness. This may give the opponent
1) space to save their stones
2) time to ignore as your threat is smaller
3) a combination of the above, which means they may have "ji min" techniques before they commit to heavily rescuing their stones or a tenuki.

In contrast "jin" moves will, at least for the next move (until the opponent defends), secure all your groups around the opponent's cutting stones, so that the opponent can't profit by threatening them from the outside, since you are one move away from capturing the cutting stones, and are willing to play a move to capture them since it saves all your groups. After a jin move, you only have one weak point, and that is the opponent saving their cutting stones. The opponent is in a bind if they don't want to spend a move solidly saving their weak cutting stones since that is such a useful move locally that if they don't play it, then you are thick(able). It is the only aji they have locally.

My interpretation is that ji min is closely related to "overconcentration", though ji min is more general and overconcentration refers to shape. If both hold, then you need to be able to explain why the area is valuable enough that the "ji min" player expects the opponent to defend even at the cost of overconcentration (it must be valuable though perhaps that is more territory and not just the weak point of a large group that I have written it as), as well as why they think they gain from such an exchange. Normally the "ji min" player can probably quite easily get 1 eye in the overconcentrated area, but getting two is too awkward so it becomes bad direction to play inside. Instead, there is more profit on the outside by threatening to get both inside and outside while attacking the opponent.

I hope that one day all common good shapes will be understood in terms of this sort of theory.

In terms of attacks:
Perhaps even shape mistakes can be thought in terms of missing the jimin subtlety.
Moves that threaten to threaten a weakness (allowing the opponent to defend too easy on the outside (rather than "inside") as hasty ajikeshi. (perhaps even the low approach to a 3-4 has some of this flavour)
Slack moves or those that lose a large group as thoughtless (lacking reading) or careless.

Another concept perhaps is "defanging" the opponent's strength by avoiding the need to ever fight near where they have already spent a "slow" move to attack.

See - it is possible to have a profound debate on L19. More posts could (and SHOULD) be like this. It's a two-way process that is by far the best for learning, unlike one-way "Look at me, look at me" videos or one-way "listen to me, listen to me" tweets. I'm a big fan of Andre Rieu. I like him because he doesn't just play music but he MAKES music, and he does that by involving the audience in a big way. It's two way big tine.

I have printed out Daniel's reply to study more leisurely tonight. But for the moment I will make some comments in scattergun fashion.

1. Hu ying doesn't occur in classical texts at all. I can see why modern players might prefer the term, however. One Chinese explanation of the origin of zhao ying (where zhao means shine) is that a teacher is like a candle, shining light all around, while all the students are transfixed by the light. This sounds a bit like folk etymology to me and has the drawback that the candle burns out. That bleakly reminds me of Li Shangyin's poem in which he says 蜡炬成灰泪始乾 (The candle will drip with tears until it turns to ashes grey). I preferred to think of the shining as referring to a lighthouse which provides succour for one's other groups scattered over the vast, turbulent ocean of the centre of the board (and at the end of the various journeys the beam still shines!). I still like the idea and think "lighthouse stone" whenever I see such a move. But I came to the term for freakish reasons. One reason was that Grace Darling was a childhood heroine of mine. The other is that a lady acting as annual president of a society I belong to was born in a lighthouse and made lighthouses the theme of her year in office. But other images can be used. A good one I got from a Chinese (i.e. a native speaker) is that a zhao ying move is like a mother calling out to her children playing outside. They just call back so that she knows they are there and they know she is there - there is no need to actually get together yet. From that I have chosen the term "call & response" in Go Wisdom.

2. Jin is by far the most common technical term in classical Chinese go. That's how important it is. The most common following character is cou, but the binome is nevertheless rather uncommon and doesn't seem to add much to the meaning. The dictionary translation is 'tight' but I reject that, partly on aesthetic grounds (it's cloth-eared in a go context) but also because it's not properly descriptive of the moves it describes (in anything, 'close-fitting' might do, but 'tight' is, well, too tight. I mentioned assertiveness as a possible rendering, but another way of looking at it is as 'controlling'. When we speak of someone who has a controlling manner, we don't actually mean he is in control, but rather that he is trying to be in control. I think that better fits a two-player game like go.

3. Daniel mentions probes. I have remarked elsewhere that when pros criticise the play of other and usually junior pros, the commonest comment seems to be that they ought to have made a probe. I assume amateurs are deemed to weak to have to worry about probes - they need to get their technique in order first. But in AI play, what I detect is that bots play moves that are not quite probes. The point about a probe is that it asks the opponent what he is about to do. In other words, it is giving the initiative to the opponent, which can go counter to the overrising need to keep the initiative. What AI (and the pros who follow it) now seem to do instead is to scatter bait. The locus classicus of this idea is Kamakura Game 1, where Go Seigen created bait consisting of a line of three stones (i.e. three moves). He reasoned that if Kitani wanted to capture them (as he did) he would have to use up to eight moves to do so. What seems to be happening now is that pros are leaving bait of (by my very rough calculation) up to eight probe-like moves, not necessarily joined up. Apart from making them very hard to capture (i.e. using many moves), these stones also have aji, and the so the new imperative of "one stone to skill THREE birds" is also satisfied. Interestingly, this number of about eight stones seems to correspond well to an old proverb devised by, I believe, a Russian player. It went something like "if an area is worth 15 points you can tenuki." Sumire seems very adept at this concept by the way, but she is far from alone.

4. Daniel mentions moyos. My impression is that use of this term is decreasing. I sense the reason is that they focus too much territory. Counting territory as a means of evaluation has never been a big favorite of pro players, I gather. The "balance of territories" is something that is useful maybe at the amateur level, but pros rely more on what I call the "balance of mistakes." That's my term but not my concept. I got it from Takemiya. Pros can assume close-to-perfect play most of the time, and so a quick way for them to notice when a game goes out of balance is to spot a mistake in technique (in simplest terms, say, an empty triangle. They also have enough experience to measure (though roughly) what each mistake is worth and so can draw up a balance sheet just using these. I think what AI has done for pros in this regard is to make them aware of a wider range of technical mistakes, or perhaps better: a wider range of ways evaluating mistakes (e.g. losing sente may not be a real mistake if it gives to the initiative further down the line; or, as just mentioned, being more robust in being prepared to sacrifice large numbers of bait stones).

5. I haven't registered ji min as a technical term (and it certainly doe snot appear classically), but perhaps it's a way of referring to the just-mentioned sort of moves that look like probes but which are not (and which also do not go so far as to make sabaki - AI seems to like to keep the tension in many positions).

hmm. I think you might be surprised how often even top pros like to directly rescue weak stones rather than playing what you might call lighthouse stones instead that the AI suggests. Though perhaps I am interpreting in terms of my jimin ideas. It can seem as if they think the point of being pro is to fight directly with all their shape skill and reading, or perhaps as practice/exercise.

I think we are talking about things related (jimin and your not quite probes). I was misleading when I said the player takes gote afterwards because sente is also an option. If they take gote then it is a classic probe, but if sente it seems to be what you refer to. If such are still to be called probes they are long term probes in the sense that it takes some time before they are worth following up directly. And yet they are still the AI moves that surprise us because the AI flits across the board from one side to the other.

The repeated complaint I've heard pros say is that they can't understand how to use such forcing moves, which is why they don't play it. If they copy and misplay afterwards, that is even worse than if they hadn't copied it. They of course also complain that isn't that ajikeshi (i.e. there is the aji of spending more moves to get from 1 eye to 2 in the overconcentrated area), so why play it so early?

Note that on my interpretation of jin in a fight, it of course takes some skill to recognise which of the opponent's weak groups is the most critical to be tight on. You need to choose the right target or else it is bad "fight direction". Most valuable is a group cutting many weak groups. But the only possible target is one that you have a direct threat against which the opponent has no way to fight back and must submissively respond to save their stones.

and yet huying literally means call and response. I don't think I ever looked it up until this conversation though. nb Pleco dictionary has that part of the poem but chooses 幹 instead which doesn't match either their translation or yours so I assume you are right. I admit I've never heard of that poet. Probably they mixed it up with the two sounds of the simplified 干.

Yes, I've come across this sort of comment, too - and not just re forcing moves. The pros keep saying they (and we amateurs) should play moves you think you understand. Seems so obvious, so why do so many amateurs ignore them!?

PS Li Shangyin's poem is in the Tang 300 (and he's well represented there with other poems). He's got a little bit of a cult in the west because he wrote love poems, which are rather unusual in the old Chinese corpus.

I'm reading this as: Favoring moves that may be very slightly sub-optimal to lure the opponent into variations that are for one reason or another predictable to the player. Similarly in Chess if one is "within preparation" while the opponent is not, one has the initiative simply by being more familiar with what is happening.

Is that a correct interpretation?

Perhaps it is that sort of move. But since I like maths, I was thinking more in terms of explaining why such moves are sometimes optimal. In Go, they can simplify the game in a sense since there is one main big weak point. But both sides should know that. However, the fighting over such a point can still be very complicated with the opponent's options to fight and cut your attacking stones. Arguably this can make it more complicated since other moves might submit with less of a fight.

Your explanation seems more to do with reading time, preparation bias, and is too "human" for me to understand in terms of maths.

random thought: if the second move in an area is smaller than the first, that actually supports the argument that you prefer to settle one area will with one move than a middling move that tries to get something partial on two areas.

confession: my maths degree went badly so my knowledge is lacking in many areas. In particular I suspect that someone with a better understanding of how to apply entropy, optimisation, probability to Go may have a much easier time of finding the local equations. I don't think the Go part is so difficult. Overall there is just a lot of averaging (i.e. integrating measure). If else may be key for tsumego, but averages of combinatorials are key elsewhere.

I can't give a definitive answer, but I'd feel fairly certain in saying no. I say that partly on the basis that chess-type traps are rare in go - the board's just too big. There have certainly been cases of joseki traps, but I can't think of any fuseki traps, and the Sumire-Rina game was the wrong sort of position for that.

But more importantly, I think there's a big difference between the use of the initiative in chess and go. Putting it perhaps too simply, you use the initiative in chess to force the opponent into a pitfall (which may be a tactical one but could be resolving into a won endgame). But in go you use the initiative to ensnare the opponent in to a spider's web. To put that in another but probably still too simplified way, in chess you use the initiative to dictate WHAT the opponent plays. In go you use it to dictate WHEREABOUTS the opponent plays. Maybe we can use the contrast: mousetrap versus butterfly net? But with the AI proviso that with one net you have to capture THREE butterflies .

To add another sidelight, there seems to be a strong impression (among those few pros who talk about it) that old Edo players score quite well when analysed by AI - not the famous moves so much, but the general run of play. I think this may be because of the lack of komi. Black automatically gets a strong initiative which good players feel confident they can keep. Shusaku may have famously said "I was Black" but actually just about every top player said something very similar.

To go back to probes (which as I say, don't feel quite like probes in AI play, and I think Daniel is agreeing with this), an important point is that you can only really play a probe if you have the initiative. But my hunch is that when such modern probes are being played they are not being evaluated in the old way (e.g. allowing sabaki or other wise settling a group) but instead a novel unit of measure is being used which I will call an INNIT (Brits will get the point), i.e. a unit of initiative. The innits may measure, for example, the degree of weakness, or number of weaknesses, in weak groups, with the important proviso that the web of measurement covers the whole board. A new-style probe, which I will call a brope to distinguish it, can be assessed by how much it affects all the weaknesses (or, alternatively, by how many weaknesses it impacts on). Even this may be too complex for a human, but even a simplified version may allow play to be more AI-like (in that one respect, mark you). At the very least it would ensure the whole board is being taken into consideration.

This is all partly tongue-in-cheek, of course, but there's a strand of seriousness in their somewhere.

The concept makes intuitive sense but still seems vague to me.
Is it the basic idea in Go of taking points while making exchanges that leave more problems for the opponent than they do for you? Over time, if you get the better of more of these exchanges, you would end up with increasing amounts of "initiative" leaving your opponent with a relatively large amount of problems to fix compared to you. Using "problems" somewhat loosely here. Am I closer?

If so can you explain how the example you mentioned at first, about the approach move inviting the invasion, would tie into this? That example to me felt more like a "trap" in that it may not have been the optimal move, but an attempt to commit the opponent to something the player is prepared for. Perhaps I misunderstood it.

I got something out of it, for what it's worth

In Go we talk about gaining points. There are rarely moves as forcing as in chess that threaten to capture a piece/king. But all is relative. We say we use the initiative (when we have the move) to either settle an area with a move, or play something forcing while keeping the initiative. I think another way to see John's not quite probes is as playing moves on the boundary of potential/territory that push out the envelope even if it isn't really secure yet. However with this flimsy barrier, it may tip the balance so that the opponent is unable to threaten your main weak points. Instead, they can only attack the barrier instead which isn't as valuable, and you are willing to sacrifice in order to defend your most valuable weak points.

In Go there is rarely such a thing as an uninvadable or unreduceable area. If it is so solid as to be completely secure you are playing much too slowly and the opponent will ignore it to take more space. It takes 4 moves to surround one point securely and even then it isn't secure as it isn't alive with two eyes. However this is true for the opponent too. So normally leaving some potential for the opponent perhaps to make one eye but not two is a good idea, especially because if you add another move you might even be able to prevent one eye while still moving into claiming more empty space.

Instead we (try to) think in terms of points. An area might not be your territory yet, but it will cost your opponent the initiative. Because they may take many moves to live, you may have the "initiative" that many times by attacking their shape with forcing moves and profiting. More subtly, strong players will make it so that such initiative are also valuable moves for themselves rather than worthless dame.

In this case the "approach" move A is optimal because it is approaching the opponent's valuable but lone corner stone C while threatening to build a region R. Even though R can be invaded, this is difficult because by attacking A it can't make much eyespace except in the centre (which isn't valuable), but if it tries to keep the corner, it can only do so by playing next to C, which C can respond to as a double threat that continues to both attack A and build R, until your invasion stone in R might be too weak to live. That is the plan anyway. Normally A will need to sacrifice the corner and try to connect up to R. But the corners are very valuable, and even if they connect, though your large side has been very reduced, they still aren't alive, so you still have attacking moves.

If I can bring up weak point theory again (somewhat digression). The point is that weak points are the most valuable things on the board, but often you don't gain much by playing them directly as when you attack the opponent defends. Instead you try to balance them against the next most valuable things on the board (empty space potential near the boundaries of groups rather than territory, eyespace on the sides/corners). Often space close to weak points are also valuable as the defender playing there can remove the attack on them until the attacker playing the threat is ajikeshi since the forcing moves are under attack themselves.

Symmetry can be explained like this. If the opponent defends their weak point safely (by making eyes), you have no more forcing moves but may have more space, eyespace access locally as they have played submissively. Then your approach of their weak point also owes an eyespace defense for the same reason. It is fairly valuable (even if less valuable than their weak point), but there are no more weak points so that if your approach is threatened, then moves to save it become smaller. Likely you back off if your opponent does too. If the opponent leaves more weak points in order to better surround your approach, then if the best weak points are close enough to your approach, then you also play close to better surround their weak points. Otherwise, you may consider either backing off or sacrificing to attack their weak point from another direction.

The effect of "corners take 2/3 of the moves to surround as sides" means that often you play solid shapes on the corners (one space jump, small knight's move) to maximise value and minimise opponent's aji, while on the side you play at least 2 space jumps normally, and often play directly into the centre to access more space or aji against the opponent's corner rather than making territory.

I think that generally bad shape occurs not from playing randomly but from players strong enough to notice there is a weak point. Sometimes if the shape is really weak and valuable, "bad shape" is best shape because being able to connect at all or having access is positive. But otherwise bad shape arises when the defender (of a group or potential territory, perhaps from aji of leaning on the opponent) doesn't play close enough to the most valuable weak points and the attacker can still threaten whatever the defender is trying to defend with a valuable move for them.

Daniel: I very much like what you say in this latest, but I'd like to quibble over a couple of points.

1.

Yes, WE do. But pros very rarely do. They are much, much more likely to talk about a gain in terms of being thicker (atsui). This is the latest middle-game/endgame usage of the term, which is rather different from the opening/early middle-game usage, but you can get round that by inventing an alternative term for yourself. Let me give an example. It's not a specially good one. It just happens to reflect something I've just been watching. It was an André Rieu documentary. He began his working life as a mainly unsuccessful and so poorly paid professional musician. Of course he was always thinking about where money (points) would come from. But over time he discovered that by making audiences feel bodily happy (not just happy ears, in other words), he could make a lot more money. That led to him thinking about even more ways to make audiences happy, which had various unexpected spin-offs (such as making his orchestra happy, who then in turn make audiences even happier, and so a virtuous circle is set up. André (apparently) acquired a mindset of thinking of ventures, or music selections, in terms of happiness units rather than money units, confident that in the end one converts to the other. I think this is something like what go pros do. They think in some sort of units other than points, equally confident that they will convert to points at the end.

I'm not sure exactly what these units might be, but I do know the term atsui is used about them a LOT. I think, however, they probably also encompass weak points/groups, which, as you point out are abidingly important. If you think of weaknesses as thinness (which of course the Japanese do), then they may be "counted" as negative thickness, and so we end up with a single unit of measurement which we might call a Tat (Thickness And Thinness). So how do you get more tats? Easy. Throw away your combs!

2. We say we use the initiative (when we have the move)
Again we do do this, but I don't think we should. I think we are being detrimentally influenced by chess. If you play a move that maps out a huge moyo (that is, takes gote) that will win the game easily if the opponent lets you keep it, he HAS to do something about it. You have the initiative but you do NOT have the move. You have the initiative because you are forcing the opponent to play in a limited number of ways or in a limited number of areas. You have no way of forcing him to play any particular move. He is a guinea pig in a cage. He can prowl anywhere within that cage. You can't force him to go on the treadmill or to take a drink of water or to scratch his tum or to play dead. But you can guarantee he can't take a walk in the park. I think AI use of this type of initiative (controlling assertiveness) is what distinguishes AI play, and I detect signs that pros are beginning to come to terms with it.

The Sumire-Rina example I referred to was in the 33rd Women's Meijin (Game 2). Sumire played the triangled stone, an unusually wide extension, and Rina commented that invasion on the top side was possible. But, presumably expecting this sort of play from Sumire, she added that she had been looking the previous at the pattern where she plays A, as in the game. In other words, she refused to be drawn into Sumire's web. But Sumire upped the stakes by next playing B, yet Rina again did not want to lose the initiative (after all, she was Black) so she continued at C. It became a kiai battle (which Rina won).



This is far from the most powerful example of "controlling assertiveness" but is maybe the easiest to get a handle on.

The Tennozan magari I mentioned was the triangled move below (Rina, Black, v. Wu in the Hoban Cup):



The million-dollar magari, which was missed by Sumire against Wu, also in the Hoban Cup, was A in the diagram below:



Sumire's triangled move was criticised as too small. It should gave been A, but Wu grabbed that point at once and took command of the game.

These magaris are nothing new, of course. What seems new is the emphasis being put on them. That is, they have a greater tat value If you want to get ahead, get some tats.

regarding the first shape, I have studied it a little before. The approach is standard if the upper left is a small knight's move, but with the solid one space jump, White is heavier though has more space. edited: I think I misremembered, and perhaps the approach is the only AI move in this shape whereas if it is the small knight's move, W considers a solid standard defense. Oops.

I think B is unusual though, perhaps bad direction.

If I try to apply weak point theory, I would say that W's approach makes sense as B doesn't have a particularly valuable invasion. However, my experience/understanding is that B invading has enough space to take over the main side eyespace of both W groups as well as defend the upper right a little (with forcing moves).

Regarding forcing moves. The more rational way to think is that some moves are connected to large threats. You don't have to play them yourself, but rather often it isn't possible for the opponent to play much until they defend first.

I think I was subconsciously trying to assert that the not quite probes are confusing if we think in terms of groups, but not so much if we think in terms of the value of weak points.

I'll share a cute AI move that surprised me after a ladder breaker follow up. I have a slight sense of deja vu but I can't remember clearly. Perhaps Andrew Simons showed me something similar?

Click Here To Show Diagram Code

[go]$$W
$$ ----------------
$$ | . . . . . . .
$$ | . . . . . . .
$$ | . . . O X . .
$$ | . . . . X . .
$$ | . 1 . . . . .
$$ | . . . . . . .

$$ ---------------[/go]

imagine some standard 3-3 invasion joseki

You seem to be referring to the distinction between kikashi and sente kikashi. If so, you may wish to look up the Forcing Moves book by Brian Chandler.

not quite relevant, but I have written up a (not mathematical beyond Go thinking) note on what I mean by weak point theory. It is at dhu163 wordpress, Board graph, weak-point-theory-teaser

another way to think about the not quite probes is that it isn't that easy to make life or change control where the opponent is strong. Even with several forcing moves that threaten deeper incursions it won't yet secure a base. But playing them earlier than you want to commit to settling them can help in case the opponent gets thick enough to kill them from the outside rather than defending. This then allows you to take territory more aggressively elsewhere since even if the opponent gets thicker, you have already gotten them to play a simple backing off move with the probes, which helps live, or at least gives better prospects and options for full life rather than just scattered potential to make one eye that doesn't link up enough to make two.

@John. You may be interested in thinking about the AI popularised invasions into seemingly narrow sides (often by W) when W has stones nearby on both adjacent sides. Often it is put in the middle of a 3 space extension and the timing seems surprisingly early. sometimes it lives, sometimes it is sacrificed, but either way, W often gets strength on adjacent sides from it.

especially given that it is W, the player with less strength on the board that plays like this, I think it is closely related to Tian ji's horse racing strategy. Sacrifice of one or more stones is ok when you don't expect to get much anyway, and if it can help strengthen some of your positions to become strong enough to live and even develop, then it may even be optimal. This happens when the profit per move in an area is like a sigmoid curve (between 0 and 1, with decreasing returns per move as it is more under one side's control), i.e. convex when more your opponent's, and concave when more yours.

nb: The most common criticism I heard by Chinese pros of European and Japanese play (pre Iyama Yuta's generation) was that although creative or artistic, they weren't good at fighting, reading, semeai etc.
The most common comment on Korean play was they were overly focused on territory whereas Chinese players were more a mix, but overall more balanced. Many Chinese players are very fighting oriented (Gu Zihao, Jiang Weijie, Mi Yuting, etc). Ke Jie in contrast is known more for having a very good record as White, with solid play, winning without risk, hard to find weak points.

"Keeping the initiative [...] I have not seen a standard term yet for this"

This phrase is the term.

"Japanese criticising Western ideas as focused on static objects and ignoring dynamic objects"

If so, those Japanese overlook very much of my theory on dynamic objects and procedures about fighting, positional judgement and endgame! In particular, they overlook where my dynamic theory describes Asian pro play applied but not explained by them.

"endgame theory can miss this because mathematicians are used to independent endgames"

This is misleading. The mathematical endgame theory by Elwyn Berlekamp, Bill Spight and me says very much about global endgame decisions!

I have never seen such Japanese criticism. I think what Daniel is referring to is my own oft repeated point, going back to rec,games.go days, is that westerners get too hooked on the Japanese term katachi (static) and ignore suji (dynamic). In other the words, the Japanese connection comes in only because those terms are Japanese. But, as we have also discussed in the past, Korean haengma is an alternative way of looking at it. My journalistic formula is katachi + suji = haengma.



Daniel: This seems to be what I mean when I talk about probes that are not-quite-probes (quasi-probes, or whatever). I deliberately shy away from calling them invasions because that is another term that is misused and misunderstood. The Japanese source (uchikomi) does not mean 'invasion' as in D-Day invasion. In military terms it amounts to creating a 'salient', but it can perhaps be better described in go as driving in a wedge. Or making a quasi-probe

PS Why does no-one talk of haengma any more?

Stones as only static shapes were stupid. Haengma contributed to a dynamic perception and introduced global efficiency to Western thinking. Now, for me it has become second nature to always check for global efficiency, or rather to strive doing so. It is not necessary to consider the concept of haengma explicitly but rather one must always develop all one's stones well. Haengma has been a PR term but the more fundamental concepts, such as connection, life, development, efficiency, sacrifice, strategic choice and tactical reading, matter in practice.

I'm going back to the original post, wondering even.

"What is a concept?"

at least in Go?

Everything useful seems to be a function of expected score and board shape, thereby associating them. Or only a function of one (e.g. keima etc.) which helps organise thoughts and allow freer word associations such as "keima is best for attacking."

connection into chains arises from the rules. Eyespace comes from chains and capture. Territory comes from score, but is close regardless. Because eyespace, territory are slightly different forms of control (as eyespace must be kept empty with all boundaries defended, whereas territory may be filled in with area scoring). Every point needs two eyes to secure control regardless of other considerations (or at least two liberties that the opponent can't play in even if they may sometimes be approach liberties or shared seki liberties (unless there is torazu sanmoku or ko).

Development is like areas close to being territory but not. Normally any region neighbouring a territory is development potential for such a territory. There are degrees of control and accounting for how many moves are needed to build it (and hence what the opponent can get from letting you try to build it). There are mutual areas of eyespace (side, corners) etc, but with a wall, that can make the centre areas eyespace options too, and hence also potential territory. But such is xu1 (abstract, weak) since though there may be a lot of space, each may quite easily become the opponent's control too.

Big moves before urgent moves. Urgent moves concern eyespace. But if the group isn't very big, then it isn't really urgent. All must be relative to score in the end.

Other than how to count, the key theoretical difficulty is estimating open areas, and evaluating fights (e.g. semeai, long before they start). This is difficult because the board is 2d not 1d like the real line. Otherwise, we get mixed concepts of positions we are so used to that we forget why it was optimal to get into such positions. We get strategic ideas of playing at outside eyespace point of opp in order to threaten to live immediately inside by attacking rather than opp having the advantage in their area. This relates to overconcentrating, direction of play, ...

Strategy, tactics may include mitigating cutting points, using their threats in combination. Plans may include preparing for fights in open areas by deciding how much you expect to get given the weak points and what it is that you want to get, preparing expectation of the likely valuable points even if there are many possible variations, so that you know which are more valuable. If the opp plays differently to expectations, you can compare and judge accordingly: did you misestimate, or should you counter with a more valuable move or a more dominating move that both takes away their threat and the value of their move (especially for weak points). F

An example of planning thoughts with a wall. It is more difficult for the opponent to make eyespace near it, so if they invade they tend to invade high, and yet if you also lack eyespace you don't want to make it inefficiently in the centre and may still want to compete for further away side eyespace at their weak points even if 2nd line since it remains difficult for them to make centre eyespace and valueless if you are already alive (even if it is only 2nd line). This is probably the main reason L shapes (magaris centre or wall with side) occur. You have good influence over the centre, but lack eyespace and can get that by making another wall around your centre while better controlling a thin layer of your own eyespace (perhaps by the opponent's weak points).

Probes at good timing are quite important in even games. Delay showing hand by committing to an area that becomes the target of the opponent's plans. Leaving the threat of aji is good. though don't let it conflict with the value of moves. Perhaps the not quite probes are fighting, to do with getting small sente moves when they may not be sente later since the opponent can lean on your weak points if they play first.

Working out exactly how far is the correct extension is tricky, but seemingly very instinctive once you get around professional level (I feel as though I am getting more confident in such a skill).

Why is small knight's move almost only shimari for 4-4 but rare for 3-4? Well in a 4-4 the corner is big and also the main source of eyespace locally. A 4-4 is happy to control it better and still try to be severe on the outside. Other moves leak the corner too easily. With a 3-4, a small knight's move isn't necessary to kill off the corner sufficiently, and still hasn't killed of the opponent's ability to play big "penultimate" moves that threaten to retake the corner from attaching/shoulder hit on the outside. A 4-4 shimari still leaves the opp approach or undercut, big move aji with eyespace, but just responding gets more territory so it doesn't mind. A 3-4 shimari doesn't want to give anything to the opponent locally without a fight.

Has AI changed Go theory? Very much so. Although its opening theory isn't exactly proven, there is a logic to it once you study it enough. This logic might one day be proven. Fuzzy things no-one was confident of, people are now sure of, with some corrections.

I mentioned play as B. For B it seems that when choosing direction of play, much like my ELO/Tian qi theory, B should uniformly add some pressure of continuing to attack by adding more moves nearer to B strength while expect to be able to invade W even if W gets the first move in their area.

If opp plays small move to both reduce your potential there and support their weak group, then let them connect then and instead commit to territory from the other direction (assuming you can't kill anyway). Then their small move isn't reducing any potential of yours and then still might not have connected their weak group to it.

@RJ


just done a rapid read through of endgame 4. I've only skimmed endgame 5. review/notes are in my study journal.

My impression is that no-one seems to talk much about the board shapes. All the theory is about move value, sente and gote. There the dependence is a matter of some moves may have follow ups, similar size moves should be played in special orders to achieve the pinnacle of accuracy globally.

But I wanted to say that thinking of weak points in terms of endgame theory has (obviously) more subtle concepts. The dependence is more subtle like the maker-breaker game of Erdos rather than being able to assign definite value to moves (yet).