Recognition vs understanding

[RobertJasiek]

Kotov said that if you think too long about a move, then you forget your initial analysis, allowing the potential of a colossal blunder.

I cannot know if that always applies for Chess players. It does not always apply for go amateurs.

Chess training books are so much better at giving tips like this, they simply do not exist in (western) books.

There are indeed no relevant (for playing strength) Western language go psychology books yet.

[illluck]

Kotov said that if you think too long about a move, then you forget your initial analysis, allowing the potential of a colossal blunder.

I cannot know if that always applies for Chess players. It does not always apply for go amateurs.

"Does not always" is not very useful wording :p

There's a well-known Chinese saying that "Lengthy deliberations result in terrible moves" (长考出恶手) and it's often seen in game reports. I haven't really noticed it myself because I usually play pretty quickly and my play is full of terrible moves regardless. I do think that the idea of forgetting the initial analysis to be a good explanation, though.

[Joaz Banbeck]

the Fred Seki

What is this, please?

Fred <> Joe

[lemmata]

Hi John. It seems that what you are describing as "effortful struggles" that give rise to pattern recognition is essentially how we study life-and-death and tesuji. We do these problems repeatedly and gain pattern recognition skills that allow us to come up with candidate moves fairly quickly. The rules of L&D permit a few widely applicable generalizations, but, for the most part, it is difficult to come up with a general algorithm to determine life-and-death in most situations. Thus, repeated problem solving and resulting pattern recognition skills form important bases for in-game L&D determination. So I would say that the pattern-based teaching is already ocurring in the go world. Wouldn't you agree?

[RobertJasiek]

"Does not always" is not very useful wording

As long as neither side of the argument has evidence for relative frequency, there is no better wording:)

[RobertJasiek]

It seems that what you are describing as "effortful struggles" that give rise to pattern recognition is essentially how we study life-and-death and tesuji. We do these problems repeatedly and gain pattern recognition skills that allow us to come up with candidate moves fairly quickly. The rules of L&D permit a few widely applicable generalizations, but, for the most part, it is difficult to come up with a general algorithm to determine life-and-death in most situations. Thus, repeated problem solving and resulting pattern recognition skills form important bases for in-game L&D determination. So I would say that the pattern-based teaching is already ocurring in the go world. Wouldn't you agree?

While there is no doubt that pattern recognition does play a role, I disagree with "is essentially how we study life-and-death and tesuji" etc. We use every available means: patterns, reading, principles and others. I also disagree with "The rules of L&D permit a few widely applicable generalizations" (see Capturing Races 1 and Wolf's semeai papers for contrary evidence) but I would agree that so far a too small percentage and a too small number of widely applicable generalizations have been spelled out clearly. Overcoming that is on my TODO list, and I will need to do a lot of research to find more and better principles, especially such with conditions for applicability and such reducing reading complexity by specifying which moves can be ignored safely. Thomas Wolf and less frequently others are also working on that, but currently from computer programs' perspective.

Finding general algorithms is rather easy; the difficulty is more to restrict the amount of necessary calculation so that players or computers find the algorithms useful in practice.

One must, of course, overcome by far too weak professionals' advice of the kind "first reduce eyespace, then occupy a vital point", which can be right or wrong in different positions. Citation from my expected book First Fundamentals, p. 158:

"
The most important means of gaining life are to: increase eyespace, partition eyespace, move to a wide open space or connect to a helping group. Since any of these means can be correct in a given local position at a particular moment, one must always consider all of them and choose the most appropriate means.
"

This is a widely applicable generalisation and corrects the above mentioned bad advice still taught by too many professional players. For understanding to be better than recognition, the theory (as expressed, e.g., in principles) underlying understanding MUST BE CORRECT!

[entropi]

According to Hendriks, what is required to make good moves at the board is not understanding or planning but simply recognising.

Very interesting discussion. I wonder where reading belongs in this context. By reading I mean strictly "I play this move, he does that, then this..." kind of partly visual "calculations".

If recognising is indeed more important than understanding, I would interpret it that efficient use of your long term (visual) memory is more important than your analytical thinking skills. This would then probably mean that the importance of reading in the sense of "I play this move, he does that, then this..." is a bit overemphasized in Go community.

Would you agree with that? Or would you rather argue the opposite or would you say reading has nothing to do with this discussion, it is an orthogonal dimension of Go related thinking?

[kirkmc]

Interesting post, and good to see you back John.

I haven't been playing much go lately, but picked up a pretty good backgammon app for my iPad, and bought a book about backgammon to try and learn more about the game. One thing that stood out in the book - written several decades ago, but republished recently - is that, in the introduction, the person who worked with the late author made some comments about some of his points. One thing that stood out was that the author - and generally all backgammon authors - always stressed that the 5-point was the most important point to take early in a game. But a small comment in the introduction pointed out that thanks to software, it has since been found that taking the 7-point is far more effective. So here was a basic tenet of the game that was proven wrong by software that would play out gazillions of permutations of games.

Computers can find things that may go against the grain, and this certainly happened in chess. The whole point of recognition in chess, however, works well on a small board, but not so much in go. On the other hand, local recognition is essential in go: that's why everyone says on has to do tons of life and death problems.

Aside from that, playing through pro games is probably better than learning josekis, if one does it while really paying attention. I think the mindset that involves learning josekis is probably detrimental to becoming a stronger player. (My personal experience...>)

[gowan]

If I'm thinking correctly, recognition can be achieved through move-by-move reading and using principles but move-by-move reading and consciously thinking about principles are too slow to be practical. It is necessary for principles and reading to become internalized so the player "sees" things without conscious effort. This internalization can't take place just from explanations of what the principles are. Internalization takes place through repeated effortful practice (e.g. playing real games of go and paying attention to what happens as well as remembering what happened in other games). It seems to be similar to learning to speak a language. As far as I know, there has never been a complete specification of grammar rules for a natural language. I know of a linguistics research monograph titled The Major Syntactic Structures of English, which attempts to specify only major rules of grammar. The book is over 800 pages in length. Clearly no one can learn to speak English fluently by learning all the rules and consciously invoking them to form every utterance. People for whom English is their first language learn to speak fluently before they learn any explicit rules of grammar in school. Even people who use English fluently likely cannot explicitly describe very many grammar rules. Grammatical decisions are made on the grounds of whether it "sounds" right. I wonder whether this ability to judge correctness by how it sounds is the same as recognition. This same sort of internalization takes place in physical activities, sports, etc. Recognition helps us decide how to make a tennis shot, I think. We'll never succeed at tennis if we have to think explicitly about grip on the racket, foot placement, arm movement, and so on for every shot. So I guess the real question is how the internalization that is a prerequisite for recognition takes place and, for trainers or coaches, how to encourage and facilitate it. I am guessing that there might be some connection with the commonly accepted idea that to get really strong at go it is necessary to learn from a young age. We notice that children of immigrants to our countries learn the new language quickly and fluently and speak with no or almost no accent while their parents struggle with the new language and might have a strong accent even after decades of living in the new country. Kitani's pupils at his dojo were children and apparently they learned through playing and analysing the results. How many of us, when commenting on a game played by weaker players and seeing a move we don't like, think first of what move(s) we would make in that situation and then, maybe, think of some specific reason why our move is better. We might not be able to think of an explicit principle that applies to the situation. In other words, we recognise a correct move without being able to cite principles that prove it correct.

[ez4u]

First let me pose a general question: What is the proposed relationship between "recognition" and "intuition"? It seems to me that for those who believe that repetition and the resulting recognition are the main source of chess/Go strength this is a serious question. In this frame does "intuition" amount to anything more than plagiarism?

In addition, what is the source of innovation in this world? When we look at modern Go, does anyone seriously believe that the rise of Korea and China has been based on a more rigorous application of historical experience? Hasn't it rather been the more rigorous search for and application of new ideas that conflict with classical experience? How has the search been conducted? It certainly has not yet reached the point where we can credit the use of computers.

Recently (just after the Japanese team managed to finish at the bottom of the table in the Chinese League qualifier) I had the opportunity to ask Kato Atsushi 9p what it would take for Japan to regain some sort of competitiveness with China and Japan. Needless to say, we were drinking but Kato-sensei tends to be pretty frank in any case. He replied that when the Ki'in arranged for Cho U, Yamashita, Takao, and Iiyama to go to Ichigaya every day and study with the younger crew like Ichiriki and Ida there might be some hope; but otherwise forget it. The present approach in Japan of basically individual study cannot compete with what is going on in the other countries.

It seems to me that another problem with the recognition school is that of assessment. In every position we see two sides. Which stands better? What does better mean? Problems are easy, they are constructed that way. But what shall we do about game records? Is our assessment based on the final winner in the game? This seems rather nonsensical based on the dramatic ups and downs we see in games between even top pros. But if we are to form some other judgment, what criteria shall we use? For we have to judge, right? If we cannot decide who stands better, do we "intuitively" apply the lessons of White's play or Black's to our own games? In the novel "First Kyu" the protagonist, Wook, solves this conundrum by studying the games of Wu (Go Seigen?) and those of Takagawa from the single player's point of view. The opponents are just foils against whom the masters demonstrate their art. Wook does not seek simply to memorize the patterns but rather to understand the choices made. Obviously he is a traditionalist. He is overjoyed when Wu's play begins to make sense after he works through the 300-odd games in his book only 8 times!

I have also been reading some writings on chess recently to see what insight they can give me into the challenge of improving at Go. I particularly enjoy the "Novice Nook" columns at Chesscafe.com since they are much more aimed at the challenge of teaching and learning chess than specifically at the novice player. The current issue "Analysis Insights" contains the following interesting comment relevant to this thread when discussing a whole-board position given to numerous students for their analysis/assessment:

'While some who took time did solid analysis, with others it was much more like the dreaded "Hand-Waving"... That is, rather than attempting the analysis earlier, they just generalized the factors in the position (doubled pawns, king in front of pawns) and made a decision based on those factors. That overlooks the fact that principles are made to help you decide what you should do if you can't calculate the answer. If you can calculate the answer, not only is resorting to principles unnecessary, it can be very counterproductive. Someone could object and say they "need the principles, because they can't calculate the answer." This might be true, but no strong player would consider this problem difficult, and if you don't try to calculate the lines carefully in a problem, you will never learn to do so, and thus master a necessary skill. Your inability to calculate problems will become a self-fulfilling prophecy. Practice may not make perfect, but practice will surely make better.' [emphasis in the original]

In addition, I am currently reading "Lessons with a Grandmaster" by Joel Sneed and Boris Gulko. Sneed is a strong amateur and a professional psychologist. Gulko is the grandmaster. The book is a review of a number of Gulko's games in which he challenges Sneed to analyze various positions and explain his thinking. Gulko then criticizes the analyses and explains his professional POV. He contrasts the amateur and professional approach. Below are a few comments perhaps relevant to this thread...

'Boris: In a previous tournament that year against Azmaiparashvili, Radjabov played 4...Nf6. I thought that he might have been worried about my preparation for the game. This is an interesting psychological moment. He was young and I was experienced, and in general, more experienced players excel in quiet positions and endgames where experience plays a more important role than the calculation of concrete variations (which benefits young players)...
Joel: That's really interesting. It is consistent with research on the rise and fall of fluid and crystallized intelligence across the lifespan. Fluid intelligence (e.g. problem solving) tends to decline with age, whereas crystallized intelligence (knowledge) remains stable or perhaps may even improve slightly...'

'Boris: Your thoughts are generally correct but too general. Here it is important to find a concrete way. When you have a position with an advantage in time you have to find an object of attack...'

'Boris: The move b2-b3 is very difficult to see, but for me it was easy to find because of the classic game between Geller and Lerner (Minsk 1979). In a similar position, Geller played b2-b3 and developed the initiative by attacking the e5-pawn...'

'Joel: This looks to me like a draw. It is an even rook and pawn ending. Many of these positions are drawn even when one side is up a pawn. Are you trying to push your opponent into a mistake? Did you think you were a better endgame player?...
Boris: I exchange pieces I don't need and get rid of pieces he needs for counterplay. In this position Black actually has a big advantage. I can create several weaknesses in his camp and have opportunity to increase the pressure. In general, this position is very difficult for White. Rook endgame rules are relative, as are all rules in chess...'

'Boris: Ponomariov was a young star at the time with one serious problem, he had a narrow opening repertoire and against anything that wasn't 1 e4 he played the Tarrasch Defense. So my strategy was to get him into an unfamiliar position because I was more experienced and had more knowledge of different kinds of positions. I thought to myself: "anything but the Tarrasch".'

'Joel: It seems in both diagrams my initial idea is strong and then I go through some analysis and arrive at the wrong move. Can you say something or give some advice about the thinking process and how one thinks or makes decisions in a systematic way?
Boris: It is not a thinking problem. You are not self-confident enough. Sometimes you see good moves but don't trust yourself...'

'Boris: You spend too much time verbalizing and not enough time and energy calculating...'

'Boris: Chess players need to make both positional considerations and calculate variations, but in the right proportion. Your analysis very often is too general and you don't try to find specific finesses in the position based on calculations. You didn't solve this problem because you tried to solve it with generalities not concrete calculations which, by the way, is the most important part of chess...'

It is interesting that in the book there are several exchanges criticizing Sneed's reliance on general considerations. However, as a reader I think that one is missing. There are times that Joel seems to start with a basic assessment that White, for example, stands better and then proceeds to explain/rationalize that idea in terms of the basic strategic ideas that he is familiar with. He then uses those ideas to filter his candidates for the next move. Gulko, on the other hand, always seems to assess the current situation based on an expected continuation (with explanations ready and waiting for why not the alternatives) and a concrete list of advantages/disadvantages at the end of that continuation. I do not believe this puts Gulko in the "Move First, Think Later" camp. However, it would seem to make him an excellent example of the "increased use of concrete analysis" in modern chess as described by John Watson.

As part of preparing this post I downloaded the free sample of the kindle version of "The Secrets of Modern Chess Strategy" by John Watson from Amazon. Everyone who is interested in this thread and has the ability to download the piece should do the same and read the beginning section of chapter 1 - "The Nature of Middlegame Theory". It is a wonderfully articulate summation of his thoughts on the past and present situation. Along the way it reminds us that Reti wrote, "The aim of the modern school is not to treat every position according to one general law, but according to the principle inherent in the position..." already in 1921 (published in English as "Modern Ideas in Chess" in 1923), which was four years before Nimzowitsch codified chess in "My System".

In turning to how he will deal with the subject in the rest of the book Watson writes:
"In light of the above, the very attempt to elucidate modern chess in general terms may seem old-fashioned and misguided. And yet, our modern literature of games collections, annotated games, and magazine articles reveals a large pool of profound and revealing comments by strong players about new and subtle ways of thinking about the game. Even more powerfully, their games speak to us. It is still possible to discover general wisdom in the mass of modern practice; we simply have to realize that the new ideas will be more qualified and specific than the bold and often discredited generalities of former times. Furthermore there is a dynamic interconnectedness in chess which needs be taken into account; thus, modern guidelines will often have more to do with techniques, sequences and procedures than with static rules..."

[John Fairbairn]

Stirring the pot a little, I read some more of Hendriks's book last night and he mentioned "hindsight bias". For me this was good timing as I had watched a DVD courtroom drama a couple of nights ago in which a barrister tried to refute a witness's story by demonstrating it had been internalised by false memories created by wishful thinking. A good way of explaining hindsight bias apparently is implicit in another name for it: I-knew-that-all-along bias. I gather it is linked with "motivational forgetting", where e.g. you forget you learned something from somewhere else and end up believing you thought of it for yourself.

I think it's safe to say that we all encounter many examples of this, and surely we are all subject to it ourselves. But it seems that some people can be much more prone to it than others and, further, a special danger among those people is that it leads to overconfidence in one's decisions or opinions. Hendriks seems to think that chess trainers are among those who can be specially prone to it. If I understand him properly, this may be due to their way of working. They gather many "good move" example positions for their students. They then analyse these as a collection and try to find general principles. They "succeed" and then become convinced that these principles explain how the good moves were found. But the trainer actually found the positions before the principles were discovered, so he couldn't have found them that way. In fact (according to Hendriks) his brain just "recognised" them as good on first encounter - trainers already being pretty good players, of course. Furthermore, Hendriks condemns trainers who use these principles to teach others, simply because the approach doesn't work. He calls it the "look and you will see" approach. Look and see are used in fuzzy ways but I think what is meant by "look" is "examine carefully" and "see" (specifically mentioned as a synonym of "recognise" by Hendriks") is something done instantly and without apparent thought.

Specifically, he mentions (with disapproval) as a prime examplar of this approach the Jeremy Silman mentioned favourably by topazq. Silman says you should examine a position for "imbalances" (strengths and weaknesses) and once you do that a good move (if it exists) will become evident. As will be obvious by now, Hendriks says this is nonsense, but confesses he has suffered from the same "delusion" himself.

Hendriks, being a new (and no doubt therefore specially zealous) advocate of the "first move, then plan, then judge" approach, also defends the method often known as "trial and error". He argues that not only is this method the correct one but that prejudice exists against it because it is actually misnamed. It is not a random process. It may appear so, but it is controlled by the brain, and it is more controlled the more work you have put in. What the work achieves is the ability to "recognise" (it may be useful to note that he later refines this to "recognising the similar").

Hendriks's book is not like one of those awful books you see on the Business or Self-Improvement shelves at airports, where many cod-proverb phrases are set in little boxes and large type with lots of screamers. Bold is used so sparingly (outside of headings) that I think there is only one example in the book. Assuming, therefore, that he regards this as specially important, I quote the sentence here:

"Positions are not examples illustrating more general principles - they constitute the actual learning material!"

But since he is also advocating the paradigm "look, plan, then judge" we have to ask what the judging part is. As I hope I've explained earler on, I'm inclined to think this does not mean application of "fundamental principles" because that title implies they come first. But it does not rule out the use of principles in some way. At a guess, it means its OK to use "collective wisdom" (proverbs and principles) but in a pick'n'mix way. I also think this view accords with a general view that proverbs are NOT nonsensical just because you can find another proverb that says the opposite. They become tools for the job, and the set of tools you bring out of the shed depends on the job in hand. You don't bring out every tool and use them in a hierarchical order - unless you don't mind ending up with an apple bix when you meant to mend the wobbly leg on the chair.

Also in that connection, I think that the comment above that (move) reading is a confirmation phase (afterwards) rather than a decision-making phase (before) is spot on, but I'd like to suggest another facet which applies more specically to study rather than play. It is evident from the forums that when many people hear the advice like "replay 1,000 pro games to become a 1-dan" they assume that what will happen is that good shape will subconsciously worm its way into the brain. Something like that must happen, I suppose, but from my experience the most valuable effect is when I see a move I would never have thought of. I even find myself actually saying, "Wow, I didn't know you could play there!" The effect is that a new move instantly enters my go "vocabulary". To pick up on another contribution above, I am usefully extending my vocabulary rather learning a not very useful grammatical rule about whether to put a comma after a subordinate clause. And not only will I recognise that new move in the future, I will actively hunt for situations where it can occur. Even better, if I'm feeling energetic, I will try to create situations where it can occur. I think this sort of thing is what is meant by "judge later". And, of course, the major difference between just absorbing good shape and this sort of discovery is that the latter is "effortful" and so is more likely to stick in my mind and/or be easier to recall for practical use. I can easily imagine that replicating that experience several thousand times, which can happen if you play over a lot of games, would not only make me strong but would enable me to come up with potentially good moves instantly.

But that involves work. At present I think the main problem for most of us is that we don't have the time, capacity or motivation to do the work, and so instead we grasp at the latest proverb or buy the latest book, or hope that genius will be absorbed via the ether when we watch the latest video. I'll be interested to see what Hendriks has to say about that. I'll also be watching carefully to see whether he is himself guilty hindsight bias. After all, the zeal of the recently converted might just as easily be explained as overconfidence resulting from that.

[RobertJasiek]

I wonder where reading belongs in this context.

Whereever necessary, to confirm suspicions retrieved by other means or to get first order information if other means are not suitable.

If recognising is indeed more important than understanding

No.

[RobertJasiek]

move-by-move reading

It depends.

and consciously thinking about principles are too slow

No. If one knows the principles and is a bit experienced with their application, then that is fast to very fast. What can be slow is the thinking about what best to think at all.

It is necessary for principles and reading to become internalized so the player "sees" things without conscious effort.

Not necessary, but it is very useful.

Internalization takes place through repeated effortful practice

Or through learning or both.

Clearly no one can learn to speak English fluently by learning all the rules and consciously invoking them to form every utterance.

It is so difficult because grammar is so very awkward.

In other words, we recognise a correct move without being able to cite principles that prove it correct.

This kind of "correct move" guessing leads to 5 different professionals insisting on 5 different moves each as the only correct move... (As it occurred when they commented on a game of mine.)

[RobertJasiek]

principles are made to help you decide what you should do if you can't calculate the answer. If you can calculate the answer, not only is resorting to principles unnecessary, it can be very counterproductive.

Wrong. Principles CAN be used that way, but good enough principles can also be used in a different way: to MAKE READING SUPERFLUOUS and so to come to the correct and true solution faster.

'Boris: Your thoughts are generally correct but too general. Here it is important to find a concrete way.

"too general" is not generally correct. A general view can be correct when applied to a particular position.

[RobertJasiek]

the principle inherent in the position...

What is meant with this?