I had a holiday break at the weekend, not far from a village called Woolfardisworthy. As part of the recharging process, I decided to do something outrageous. I took along a book on what most people here would call either yose or the endgame.

Regular readers will know I rail against misuse or misunderstanding of the term yose. It just means making boundary plays. I was therefore delighted (though not at all surprised) to see that the authors of this Japanese book shared my views. In fact, the very first word in their Chapter 1 was the heading "Boundaries" (境界). The last sentence of the preface also conveys the same emphasis: it extols the virtues of learning to know where the boundaries (境界) lie. The word yose does not get discussed until page 45, and even then gets the scantiest of mentions. The word for endgame (終局) does not get mentioned at all! The book itself is called "A small cyclopaedia of counting" (目数小事典).

But what I was surprised at was getting the solution to a mystery that has dogged me for years, and - if my reading of this forum and SL is correct - a mystery that is shared by an awful lot of amateurs. It may be, as astute readers will have deduced, that the solution came to me while I was in Miss Marple territory. Normally, I can't stand Agatha Christie stories, but maybe there was something in the local water, which they call cider, and it reached parts of my brain that I never knew I had.

The mystery is this: what the hell has deiri counting got to do with miai counting?

I'm not actually interested in the connection. The reason is bothers me is that O Meien wrote a book on "Absolute Counting" and he rather rubbished deiri counting, while his AC counting appeared to be miai counting dressed up in marketing frills. It's actually a rather enjoyable book, but I was puzzled why it seemed to vanish without trace within Japan. Was it because he was pushing a Chinese-based system instead of a Japanese one? He certainly hinted that that was part of his approach, and he himself is a native Chinese speaker. But against that, I do know that Chinese pros use deiri counting, too. In fact, I saw an example of that just before I went on holiday, in a book of very detailed commentaries by Chen Zude. He gave an example of comparing moves by counting and the unit of measurement was 目 and not 子.

But this rather old book I started reading told me the answer within a few pages. It appears that Japanese pros are perfectly familiar with both deiri counting and miai counting and use both, but miai counting is subsumed within the actual procedure they use, and so they just opt to use the term deiri counting. In other words. deiri counting doesn't really mean deiri counting. It means deiri counting + miai counting. If you are familiar with Japanese books on the endgame, you can spot the difference because deiri diagrams use X's and miai diagrams are decorated with triangles.

It may seem strange to some readers that a term does not necessarily mean what is says. But there are those of us who do not think it at all odd that the aforementioned Woolfardisworthy does not have the pronunciation Woolfardisworthy. It is Woolsery.

In fact, terms are a big part of the book in question. Over 60 are mentioned in the index (a rarity in Japanese books, so that alone tells you some thought was put into the book). Next to none of them appear on SL in the same understandable form, and some don't appear at all ("boundary" is a gaping hole there, incidentally), though it does appear that several are not entirely fixed in Japanese and different players may use different locutions for the same thing. The book does, however, give the correspondences (from a Japanese-language perspective, of course). Indeed, it doesn't bother with strict definitions at all but instead it gives example sentences of how a term may be used in practice, and some of these example sentences turn out to be a fantastic little array of either proverbs or useful nuggets - in either case very digestible. One was something along the lines of "if the count ends in 3, think about tedomari." The book doesn't say, but it seems these nuggets come from the mouths of pros. The book itself is written by the editorial staff of the old magazine Igo, which has a distinguished track record in go journalism. They are known to consult pros properly. The book is also published by Seibundo, which is well known for its high standards.

I didn't get far into the book. As soon as lots of numbers started appearing, my eyes glazed over. That was despite the fact that the authors are clearly NOT of the far-end-of-a-fart school of endgame study. Fractions are mentioned really only to show why they occur, but they are then treated in the form of "a wee bit over 4". I was pretty confident that, if I ever felt so inclined, this is a book that would teach me how to understand the basic-to-intermediate application of the counting procedures as used by pros (I gather the trick is to know when to prioritise timing over counting), and for me it is the only one I have ever seen that would do that. (For a higher level of understanding I would turn again to O Meien's book, but I think I need to get a better grip on the basics before doing that).

One reason (among many) that I would put my confidence in this book is that much of it is simply a list of 35 basic patterns which are analysed in detail, and the various kinds of counts associated with each are listed (in that sense, it is perhaps more like a dictionary 辞典 than a 事典). But the beauty of it is that all the patterns are very common in actual play (they range over fuseki/moyo to specifically endgame shapes), and all or most can easily be used to derive many further patterns, so that the tale that certain players have memorised over 1,000 patterns suddenly seems far from outlandish. There is still hope for people like me. The end is not nigh! Though I might need another holiday after writing all this,

The ISBN is 4-416-70461-5. I have no idea whether it is still available (I've had my copy, unread, for some 20 years), but quite apart from recommending a good book, my aim has been to offer encouragement to those like me who have been left confused by the FEOAF treatment of counting boundary plays.

For me the mystery is not the relation between deiri value and miai value but rather their defintions.
Let's take the "deiri value" used for a very long time. If the defintion is quite clear for a pure gote position what is the definition of deiri value for a complex position.
I do not talk about the calculation of the deiri value but I talk only about the definition of the term "deiri value".
To be a little provocative, by defintion, what is the deiri value of the initial position (empty board).
Assuming the first player will win (with no komi) by 7 points could you derive the deiri value of the initial position from your own defintion of deiri value?

I've recently been practicing "miai value" with several not so complex positions. It speaks more to the mathematician in me than the go player in me. I like the logic of it. I can imagine how someone who practices a lot, can improve their decision making in the late endgame. I know the L-group's status by heart. An endgame expert should be capable of knowing the miai values of standard endgame positions by heart.

In the earlier endgame, with multiple more complex positions, often affecting each other, one has to approximate anyway and then I don't think it really matters whether you apply the strict logic of miai value, or a less well defined traditional way of looking at the endgame. Trying to be accurate might actually be negative for your decision making and energy not well spent.

Even in very precise conditions, I still see shortcomings. Let's assume

- Black to play
- Black has 2 moves available with miai value 5. One of them is sente, the other is reverse sente.

By definition a move is sente if its follow-up is bigger than the value of the move itself. Let's assume the sente move has a 10 point follow-up. And the reverse sente move is a White sente with a 15 point follow-up.

Black plays his sente move. White is now faced with either playing her 5 point sente or covering for Black's 10 point follow-up. By miai value, she plays the 10-pointer. Then Black plays reverse sente. Now White can play a move with value <5.

If instead White plays her sente, then Black is faced with either a 10-point follow up or cover for White's 15 point follow-up. He chooses the latter. Then White covers Black's sente. Now Black can play a move with value <5.

This idea of mutual damage really depends on the value of the follow-ups, not on the value of the sente moves itself. Perhaps this is well covered in endgame theory books by O Meien, Antti Tormanen or Robert Jasiek, but it isn't encapsulated in the concept of miai value itself.

All in all, I find the miai value concept well defined but not really sufficient as a guidance for the endgame.

What is the problem Knotwilg? If the sequences you proposed are played at temperature 5 then your two cases leads to the same result because in the first case white plays first in the environment while in the second case it is black who plays first in the environment.

Clearly I'm no expert, but I will try to say what I observe.

But first, a reminder as regards what I will lazily call the Japanese method: this was devised to be practical. It was not something defined by pure mathematicians but was based on accountancy procedures. Both de-iri (income and expenditure) and miai (making comparisons, or correspondences) are Japanese accounting techniques. Both go back a very long way, and worked (if we judge by the play of Edo masters) but miai "suffered" from a treatment by Japanese mathematicians of the FEOAF tendency in thr 1950s. The result is now a you-know-what in the swimming pool, but that doesn't change the fundamental aim: a practical procedure. That means talking about temperatures and definitions is unhelpful to those who want to actually finish their games.

The main thing I observe that seems different between what I see in Japanese books and what I see here is that many people here seem to want to talk about deiri purely in terms of numbers. That is close to an alien concept in the Japanese view as far as I can make out. They do make a count (目数) as the first step in a "deiri assessment/calculation" (出入計算) of the move being considered, but the end-product is a not a number. It is an evaluation expressed as a combination of a status phrase and a number. E.g. to take an example from the book: 出入目数　　片先手　　約5目半 (deiri count: one-way sente, approx. 5½ points). Also, as far as I can tell, the order is important there, in that the timing of the move is based on its status, and timing is generally given some sort of priority over size. The number component (income - expenditure) is not referred to as the "deiri value" - it is called the 正価目数 (net count).

I can't see any example of "miai value" in the book. There is a lot of "miai calculation" (見合計算) but what is being calculated is not a single number but the "boundaries." The purpose of the procedure is specifically given as 境界策定 (determination of boundaries). Obviously, once you determine where the boundaries will be, you can make an assessment of the size of each side's associated territories. But it is only an assessment. If you imagine the common edge plays where there is a gap on the first line and either player can play 1 and 3 as a hane-and-connect but ending in gote. Since you don't want to end in gote unless you have to, but you still want to know roughly where your territory will be, you can imagine a simple sagari by both players. Because these sagaris are not actually played, Japanese books usually show them with a mark such as a triangle. Making these imaginary moves is the main purpose of miai calculation (remember: that measn determining boundaries, not computing numbers). The example above is too trivial to worry about, but there are cases where we have to combine deiri and miai calculations, and this is where the miai is a "subsumed" calculation. Assume you have found a place where you think the size of the play is large enough for you to want to play it, but if you do play it it does not actually seal off the entire boundary. A typical case is an exchange on the second or third line which leaves a potential follow-up move for either side (for example, the type of edge move just mentioned). You use a miai calculation to sketch out a temporary and imaginary boundary (想定境界) to give you a feel of what amount of actual territory you can look forward to. The Japanese terms for these two separate calculation stages are 前段 and 後段 (1st stage and 2nd stage).

Once you have sketched in all these imaginary lines, you have a basis for deciding on how to proceed, but at some point you also have to learn to take account of reverse sentes (i.e. timing), kos and "responding" (応手), i.e. whether to respond or to tenuki.

Quite how you do that is not fully explained. But I think the important point that is being conveyed is that you need to grasp is that you treat the whole process not as a computation, but as a assessment/calculation. In other words, you treat it just like reading a balance sheet. You follow the numbers but also make strategic decisions, and there's nothing to say strategy can't come first.

My sense of what happens here is that numbers guys spend a lot of time checking that the income and expenditure totals add up, and they crow if they find a penny error, and go on then to rubbish to whole exercise. The more practical approach for most people is surely to think like a CEO. "We are losing money/competitiveness/status. What do we do about that?" Getting rid of a swathe of staff might be the first deiri idea, but that has consequences. With fewer staff you might find it harder to keep customers satisfied. That's the miai part of the procedure. The follow-up. And so on and so forth. A judgement is needed first, not a computation. A computation is necessary as a final verification (検証 in the book, also given as 目数比較 - comparison of counts). What would surely not be a good idea would be to start the reappraisal of the future of your company by calculating what you spend on loo rolls. That really would be the ultimate FEOAF technique.

For Gerard: the book does give a deiri assessment for the empty board: "deiri count: approximation, double gote, 26 points."

I agree that deiri assessment/calculation should produce a combination of a status phrase and a number. It is really a crucial point because in the orther hand the miai value calculation should produce ONLY a number. That is the basic reason why I prefer using miai value rather than deiri value.

First of all my FEELING is that the deiri value is approximatively equal to 4 x komi.
If the book mentionned a net count = 26 I suspect that the book considered a komi = 6.5 which seems reasonable.
But, for a theoritical point of view a deiri assessment has nothing to do with the komi.
For a theoritical point of view, instead of taking the komi, you have to take the result of the god play. This result is expected to be the integer 6 or the integer 7 but it could not be equal 6.5
That means that the net count of the empty board position could be 24 or 28 but could not be 26.
Coming back to the mystery of the deiri definition, how can we say that the net count of the empty board is 4 x komi (or 4 x result of the god play) if you have no defintion of what a net count is?
In addition how can the book say that the empty board position is a double gote position? This is another mystery. What is a double gote position definition for a complex position?
It is really a great mystery for me : we are able to give an estimation of the number associated to a deiri assement but we are unable to give a definition of such number. How can we give an estimation of something not defined?

Do you then make an exception for double sente?

Maybe I'm struggling with the temperature concept then: doesn't it matter who has sente when the temperature drops below 5. In my example, Black got both 5 pointers and White had sente at dropped temperature. Versus both playing a 5 pointer and Black having sente at dropped temperature.

Your example would be more interesting like this: the sente move has a 15 point follow-up. The reverse sente move has a 10 point follow-up.

Black plays his sente move. White plays his sente move. What should Black do? Continue his sente move, or answer White's move?

I do not understand your point Hanspi.
let's take this tree which looks like a double sente situation

You can consider it is 2 points in double sente but you can also say it is a position with miai value = 21 (edit)
The two interpretations are possible and you have only to decide which is your own preference in practice.

Oops, when black plays his sente move black should answer to avoid the black follow up. Instead, it looks a bad move for white to answer by playing her sente move. If black played her sente move at temperature 5 then white seems not in a hurry to play her own sente move.

I believe the UK programme 'Allo 'Allo achieved some popularity in France, so I shall copy Michelle Dubois and say "LIsten very carefully. I shall say zis only wernce."

How can we define something that is not definable?

De-iri is not one thing, but two. De= outgoings, iri = incomings. It is therefore not amenable a computation. We say "deiri keisan" where keisan, if you look it up in a dictionary, may give the impression that a computation is involved, but actually the true meaning of the phrase is more like "a way of looking at something using go's equivalents of income and expenditure." It is possible to attach numbers to various elements, but not to derive numbers from the various elements used. Furthermore, although there may be a bottom line in the balance sheet, the balance sheet is not an absolute statement. It is a snapshot in time, and can change as circumstances change. It may be that an individual item in your balance sheet is far more important than the current bottom line. For example, if you office rent has shot up 300%, you are likely yo concentrate most on changing your landlord.

Similar remarks can be said about the "miai way of looking at things." It's an illusion that it's more accurate. It's just that it is easier to look at. There are no piddly things like moves types, follow-up moves, kos and so on to bother about, as there are in the messy "deiri way of looking at things". But being easier doesn't necessarily make it better. It seems that real pros do it with deiri (which subsumes also miai) and not miai. It seems from O Meien's book that they do this with a liberal degree of approximations, which in turn are based on long experience, and so this is something we cannot expect to replicate quickly just by reading a book or looking at a few problems - and even less by doing even more arithmetic.

There are many instances in the real world of using things which are essentially not definable but where we still attach numbers. A good one is distance. Distance matters if you are thinking of making a holiday trip, say. For me, London to Paris is 300 miles. London to Edinburgh is 400 miles. Using a miai way of looking at things, I would have to choose Paris on the basis of a simple comparison. But using the pros and cons of the deiri way of looking at things, I might put down a lot of extra elements. I can drive all the way to Edinburgh on the right side other road. I can call in to see relatives and friends on the way to Edinburgh. I don't need a passport for Edinburgh. Edinburgh doesn't have enormous queues. And so on. On the other hand, Paris is more beautiful. Paris has better cuisine. Paris is more exotic. And so on. There is a greedy option, of course, which I have taken recently: do both. But in general, all such decisions, in many areas of life, are best taken with a delicate deiri sauce with an underlying hint of miai, served by a professional chef, rather than a bland miai ketchup given out free at MacDonalds.

OK John let's accept deiri (or miai) are not defined. In that case why do we calculate the associated number? IOW for what purpose the deiri or miai value are used? If the purpose is to say that it is better to play first in local position P1 rather than in local position P2 then another problem arises: the calculation may be different for black or white point of view. We know that in the majority of cases black and white points of view are the same but it is not always true is it?

For the very same reason we look up the numbers that tell us how far London is to Paris and Edinburgh. They are then used as one element in making a decision as to which move to play next in a complex position. A game of go is not a sudoku puzzle.

I have tried to make the point that, in phrases such as "calculate the associated number", 'associate' (or 'attach') is more important than 'calculate' when there are other elements to consider. The numbers can be approximations (as O Meien has said).

O Meien said many other things that I don't see being made here. I said above, perhaps a bit rudely, that his preference for calling his view of making boundary plays Absolute Counting is miai dressed up for marketing purposes. There's nothing wrong with that, and given the confusion that reigns with the traditional terminology it may even be a huge plus point.

Yet, in fact, O's book is in very many respects very similar to the "Igo" book I am talking about here. He does talk a lot about deiri, for example, despite his title. Let me give some random quotations that I have just plucked from his book. I do this to show that he talks about different things or things in a different way from what I observe here, but also in ways that are similar to what is in the Igo book. In particular, despite his new terminology, he talks about the topic in ways that marry with the standard Japanese/pro way of looking at things. My own impression is that the Igo book stresses "boundaries" a lot more than usual and O stresses miai counting more than usual, but I imagine both do this not to alter the substance of the topic but rather to try and make a difficult topic more accessible for amateurs.

Anyway, to the quotes, which I have selected to show some of the less usual elements he discusses or the unusual stress he puts on the usual elements.

The first group shows he does use deiri counting but implies there may be a problem with pure deiri (hence, as the Igo book claims, the need to subsume miai into deiri):







The second group shows that he emphasises timing especially, among the other elements apart from numbers.







The next quote shows an example of extra elements even beyond timing:



The last group gives a couple of quotes that show his overriding aim is to illustrate how to think about the endgame, which includes a willingness (?need) to use approximations:





I think I have exceeded my numbers quotient for the year and so will say no more. More seriously, it is pointless trying to argue with me. I am no expert on the endgame and don't want to be one. I have not just read O's book but have translated it in full. That is why I can quote so easily from it (I can just search on my digital text). But my interest was precisely in "how other people think" and not in how to get numbers to play with. I have made no attempt to issue a translation because that would involve royalty negotiations and the end-result, given the state of the book market, would give me no more than a few cups of coffee (I am not exaggerating).

The best advice I can give is to read the various books yourself, and argue with the pros, not me. But for all practical purposes I'd still put my own money on the pros in any argument.

Thank you John to have taken some time to give us such interesting information. Sure every one will appreciate.

What a coincidence, I've been using that standard recently in my game reviews, I guess if it has pro approval it must be right!

0.75 any loss more than this is a style difference
1.5 any loss ore than this is an inaccuracy
3 is half the value of sente, anything more than this is a mistake
6 is the inherent value of sente, losing more than this is a big mistake
12 equals a pass in the opening, one handicap stone after the first, anything more than this is a blunder.

Pro approval is just one opinion. Rather, the first statement is right as approximation due to my theorem 50, which Bill Spight motivated, I formulated and proved for the temperature T:

"For the net profit P of starting, then alternating, in an ideal environment [with N value drops], N >> 0 => P ~= T/2." [26]

As to the second statement, a margin of error can be chosen so a particular one is not right or wrong but rather a matter of preference. Bill has suggested a reason why to use this particular margin of error, which I have formulated as follows:

"[...] the minimum value of having the turn is 0 and the maximum value of having the turn is T. Then 0 is the estimated minimum error of having the turn and T/2 is the estimated maximum error of having the turn. Now, we can also estimate the average error as T/4. The maximum error of this estimation is T/4. [...]" [27]

However, apart from playing around with more numbers, I am not convinced that using any margin of error at all provides more practical information than not using any and instead only using the primary value, that is, the value (which is a net profit) of starting in an environment.



7 and 14.


References: https://www.lifein19x19.com/forum/viewt ... 45#p143245

For me the explanations given by John are quite clear and make sense.
What can happen in practice? You have two local positions P1 and P2 and an environment E so that the board looks like: P1 U P2 U E
When you calculate the miai value mP1 of position P1 you assume the environment of P1 is an ideal environment but here the environment is P2 U E which could be a quite complex environment. That means that the miai value mP1 is only an information you can take into account and a lot of other informations have also to be taken into account to play the best move. I think it is exactly what John explained.

P1 and P2, if significantly larger than the temperature T(E), are the ensemble, which can be solved together with the consideration per variation of its player then starting in the environment E and gaining T(E)/2. Ignore if necessarily always the same player starts in E.

Oops I do not consider mP1 and mP2 are significantly larger than temperature T(E). On contrary I assume T(E) significantly larger than mP1 and mP2 and I am waiting temperature drops in order to play in P1 or P2.

I do not really understand why you consider P1 U P2 being the ensemble. The basic advantage of theory is precisely to handle separatly independant local positions like P1 and P2: instead of considering a difficult P1 U P2 local position, in my view, the idea of the theory is to handled separately local position P1 and local position P2.

I agree with John presentation. I am happy to calculate mP1 and mP2 (or at least an approximation) because I consider they are a (very) useful information but they are not all the picture. if mP1 and mP2 are not so different and taking into account that a real environment is never ideal then a strong player (a pro?) must consider more reading before choosing between a move in P1, in P2 or in the environment.