It took a long time to get the western go world to accept thickness != influence - and even now it's only been a partial success. Similar problems and progress with aji and sente.

Maybe we need to start serious work on 形勢判断, variously rendered as positional assessment or positional judgement, and sometimes (loosely) as counting.

Step 1 is to acknowledge what the Japanese means, so that if you ask a Japanese pro to do a 形勢判断, know what can you expect to get.

形勢 literally is 'shapes and power'. If you want to know more about these originally Chinese terms, an easy way for a westerner is to look at Sun Zi's Art of War, though do be aware that 勢 is a concept that still befuddles western military men. In go it's sufficient to consider it to be the power exuded by strong stones (i.e. sometimes thickness, sometimes influence, sometimes the way stones work together). But the derivation of the word also implies the Japanese pro will focus on weak groups, strong groups, efficiency and shapes above all else. This is amply confirmed by the Japanese literature.

判断 as a verb can be rendered various ways: assess, conclude, infer, esitmate, judge. It does not mean 'counting'. Indeed, the word 目算, which from the characters looks as if it means 'counting the points', is used in Japanese to refer to a calculation of probability, i.e. expectations of chances of success. Whether you regard this as a cultural quirk is up to you, but it does mean that a Japanese pro will not normally count as a first response.

Step 2 is to acknowledge that rather many western players have a strong interest in numbers - some obsessively so. They tend to ignore the Japanese go wisdom and try to pin numbers on everything. Whether you regard this as a valid approach is again up to you.

But Step 3 does not follow Step 2. Step 2 is a yellow-brick road taking you out to the fantasy of Emerald City.

If you keep your feet grounded in Kansas you jump to Step 3 straight from Step 1.

Step 3 is to acknowledge that the purpose of assessing a position is to find the next move. This simply calls for a little common sense.

CS tells us there are three scenarios: (1) You think you are ahead, in which case you should play a safe move; (2) you think you are behind, in which case you should play a risk-taking move; (3) you are unsure who is ahead, in which case you can try to get some back-up from counting, but the odds are (if you are a pro and your judgement based on other factors is known to be reliable), you will still conclude you are unsure. You will then play neither an overtly safe nor overtly risky move, but will try to play what you think is a reliable move, which may simply be a procrastinating move or even some move that simply reflects your personality.

In other words, you are assessing moves based on winrate, just like AlphaGo. It makes no practical difference whether you think you are ahead or behind by 10 points, 12 points or 15 points. It makes no difference whether you think the game is close within X points. The game is close whether the real value is at the end of the range or at the centre. It makes no difference if a pro says it's close to within 2 points and you, with much less skill at assessing weaknesses and so on, say it's close within 10 points. For you, and your next move, X = 10, no matter what the pro or another amateur with an abacus says.

So what we see in the Japanese literature, therefore, is a slew of books telling us not to get hung up on counting, and especially so in the first half of a game. We get whole books with 形勢判断 in the title and nary a count in the entire book. For example, Kobayashi Satoru's book on "Pro 形勢判断 using tewari". In other words, learn to assess positions by spotting ienefficiencies. Or Takemiya Masaki's book on "Takemiya's 形勢判断" which is subtitled "Striving not to surround territory." Although he is speaking a little bit with tongue in cheek, he firmly points out that counting belongs to the very end of the game once the last move has been played, and that instead the key points during te game are "the shapes, relative strengths and efficiency of stones." Or take Kum Sujun's book on the "Four Basics of how to Surround Territory", which manages not to count a single territory but instead stresses the need for efficiency and advises always to start with the weakest group - in short, essentially the same approach as Takemiya. Or how about "Decision-making at the Highest Level" in which Yoda Norimoto, Yamashita Keigo, Iyama Yuta and Kataoka Satoshi (plus the players in the example games) all (but separately) assess many positions and so derive their next move? And heaven forfend, they do this without a single count anywhere in the book. Instead they all point to, say, a poorly placed stone and so infer that that side is behind (how much behind is a matter of debate, of course). Think you can assess by counting thickness instead of (or as well as) territory? Try O Rissei's book on the "Pitfalls of Thickness". Think the territory you counted is safe? Try Yamada Kimio's book on how to mess up territories. I am just mentioning recent books. The same pattern occurs in every generation of books.

Was the past a different country? Ever seen a commentary by Go Seigen supported by diagrams full of little crosses?

Now of course we can point to the occasional book or article with these little crosses (going back at least to the 1920s, long before Cho Chikun - or "me" - was born). Are these counter-examples?

Not really. They serve a different purpose. Broadly speaking they are intended as a back-up as mentioned above, or apply only in the more tractably countable late stages of the game (mostly around move 150).

In the case of back-up, all that is needed is a rough count. In his book on the "The Easiest Way to 形勢判断" Kataoka emphasises that all you need is a comparison of territories - "this Black corner" is about the same size as that White corner", or "that Black area and that one together match that big White one." A very common, and highly recommended, variant of this approach is to count basically in units of 5 but qualify the result for each area as "exactly" or "a little over" or "a little under". The summed overs and unders tend to cancel each other out, which if you think about is a sort of Monte Carlo approach. And which new pro was it that had spectacular success with MC?

In the case of the later stages of the game, as best explained in O Meien's book on Absolute Counting, the purpose is still to assess who is ahead and so what the next move should be. But even for a pro as skilled at counting as O, there is still a sometimes considerable margin of error (variable because it depends on the size of the biggest move), and it is certainly greater than +/-1. Counting is used mainly because at that stage weaknesses and inefficiencies become too murky to distinguish usefully. And even if you become skilled at this, there is nothing there to feed back usefully to the early part of the game.

Naturally, we must add a rider: not all pros are the same. Some may even like numbers and prefer to count. For example, in the AlphaGo commentaries it has been observed that Kim Myungwan counted rather a lot. But I wouldn't take that at face value. I'd first entertain the possibility that in his residence in the west he has become used to westerners asking him what the score is and is just giving people what he thinks they want. But even if it turns out that he is by nature a compulsive counter, that doesn't constitute a counter-example. It doesn't alter the situation where the vast majority of pros who've talked about 形勢判断 take counting off the counter and tuck it away in a box on the top shelf, to be used only when the occasional guy needs a gefurtel. But guys who like gefurtels can have them if they really want them. Especially in Emerald City.

John,

I've heard that Lee Changho's greatest strength was endgame and counting early was part of that strength. However, Myungwan Kim said that all the strongest pros can match Lee Changho's endgame because they had no choice. Is it possible, that predicting ideal moves, and counting earlier and more accurately is quite common in today's game?

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

Positional judgement can assess territory, other aspects of the current position in its creation from the past or other aspects of the current position in its future potential. Some aspects can be the best and most easily assessed by numbers while for other aspects this is not so.

For example, a positional judgement of the final scoring position is the best and most easily assessed by the single number of the score. For example, a choice between a move that kills and another move that does not kill (for assessing the status death) is the best summarised as choosing the move that kills (no number would simplify this for a human player).

This is not about preference for numbers or number-free parts of Japanese go theory. It is about preferably choosing the, as Paul Thurrott would put it, right tool for the right job. If one or a few numbers express it all, then why choose a wild combination of number-free tools? If one or a few number-free tools express it all, then why choose a wild combination of numbers having little meaning?

I do not, however, buy the argument that a rough count would be better when strategic decision-making demands an accurate count. Rough counts are good enough only when fine decisions need not be made. This is the most obvious during the late endgame. In my experience, accurate judgement also is frequently useful during the opening and middle game.

To me what you say makes no sense.
If we assume that Positional Judgement (=PJ) is, as you suggest, the skill to evaluate a position and say who is better, then I see two possibilities:

(1) Your PJ is superior to Redmond's
In which case, you can apply your skill to each candidate move and win each game, pretty much. And there is really not a big chance of a "blunder."

(2) Redmond's tactical superiority can affect the evaluation to your disadvantage
In which case PJ should definitely take the tactics into account - otherwise it is incomplete and not very useful, no matter how you think it "superior."

Bottom line is still, tactics or no tactics - either you can tell who is ahead better than Redmond and you can beat him in a game, or your claims are false.

Maybe you only think you can evaluate a position because you are unable to see the many tactical possibilities which are apparent to Redmond and which can greatly affect the evaluation? Just a thought. But this is really how it looks to me. Thus the beginner story from my other post. The less you know, the more you think you know.

_________________
- Bantari
______________________________________________
WARNING: This post might contain Opinions!!

This is simple cop-out.

Reading the posts you had to read to make the responses you made, and making these responses took much more time and effort than what you were challenged to do. If you have time to argue incessantly, and write one post after another - you have time to perform a 20 sec analysis and do it better than Redmond. You can probably even spare 30 sec, maybe even 1 min.

_________________
- Bantari
______________________________________________
WARNING: This post might contain Opinions!!

Bantari, in positions without required complex reading, positional judgement is possible without complex reading. In positions with required complex reading but enough thinking time to do it despite its complexity, positional judgement is possible with complex reading. In positions with required complex reading but without enough thinking time to do it despite its complexity, accurate positional judgement of the whole position becomes impossible and the winner is mainly decided by better complex reading within the limited thinking time. (Translate this to the more specific case of comparing Redmond and me, if you like.)

Posting messages takes minutes and calculating a count for an easy position takes less than in minute, I know. However, there is nothing to prove about being able to calculate a count for an easy position because 1) I have demonstrated it many times before and 2) everybody should be able to do this. It is only interesting to demonstrate that I can positionally judge more complicated positions, too. And this requires significantly more time than posting messages and than I had available yesterday. If you do not bug me too much today, I might find time today besides the necessities of life.

I can't answer that question with confidence. Only a pro can. But I would point out that a Yi Ch'ang-ho game is one of the main examples in O Meien's book, and he explains Yi's skill very clearly, at least for the later part of the game. That implies that other pros too have now understood his skill and caught up.

Speculating from that, I'd guess that Korean and Chinese pros have made much more of an effort to match Yi in that area than the Japanese, and have succeeded, and this is part of the reason they have opened up a gap. Still speculating, the Japanese may have been disinclined or less able to make the effort because the souba element in their go (and life in general - accepting par for the course and compromises) is too strong.

Yet more speculation: modern short time limits would seem to inhibit counting earlier (and earlier counting still doesn't change the principle that you only need to know who is ahead/behind). But the greater practice at counting the late game in the fashion of Yi and O (which is more to do with counting the value of moves, rather like de-iri values, than putting an x of each point of territory) may feed back to the early part of the game for them, giving them, for example, a better feel for the eventual actual size and thus efficiency/winrate of moves.

But I can't see that this changes anything for the typical amateur, who spends even less time on each game than Mickey Mouse. For them, the traditional Japanese advice still seems most accessible and useful.

A typical fresh PJ takes up to ca. 3 minutes. Update incrementally most of the time and refresh only a few times per game, and PJ is no problem in games with 60+ minutes basic time. Professionals have enough thinking time, except for lightning games.

EDIT: Here is a sample judgement:

viewtopic.php?p=200990#p200990

I brought up the question of Redmond's positional judgement in the ongoing complexity thread, and Robert responded. Con permiso, I am copying part of that discussion here.







Edit: No, that's not right. It is true that traditional go evaluations are only partially ordered. Evaluations based upon perfect reading which include who has the move are ordered: win > tie > loss. Whether probability estimates are only partially ordered or not is a question of their semantics.

In any event, early estimates are imprecise.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

Continuing the theme of the need to understand 形勢判断 better, I just came acroos this useful example.



The young Honinbo Genjo erred with White 32, assuming it was a territorial move, but Yasui Chitoku sucked him into a fight which ended with White 60.

Three independent commentators (Iwamoto, Segoe, Shimamura) all emphasised this phase of the game as leading to a lopsided victory for Black, and pick up on the facts that White started the attack but ended up in gote and that B got good moves for his ko compensation (turning a potentially later attackable group on the lower side into the foundation of a huge moyo). All make straightforward assessment, without the need for counting, that Black is now well ahead, even though White has no weak groups.

But Shimamura, in his verdict, sums this up by saying White has completely lost his 形勢　(形勢を損じた - a fairly common expression). Although in go terms it would be good to be precise and say White has lost both his shape and his outward posture, we often usefully sum up a team game such as soccer by saying one side has 'lost its shape'. Again, instant assessments are perfectly possible.

In go, specifically I think it may also be useful sometimes to reverse things and look at a more normal position to try to identify the elements of 'unlost' shape. What is in my mind is the fact that shape (形) actually covers rather more than what westerners think of as 'good shape'. It's more large scale than just pretty patterns such as knight's moves. It should feel more 3-dimensional. In military usage, one rendering is 'dispositions'.

Anyway, whatever it is, you can see that Genjo lost it. Tilt. Game over.