It is currently Thu Mar 28, 2024 8:58 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 65 posts ]  Go to page Previous  1, 2, 3, 4
Author Message
Offline
 Post subject: Re: Recognition vs understanding
Post #61 Posted: Wed Jul 25, 2012 11:37 am 
Oza

Posts: 3647
Liked others: 20
Was liked: 4626
Encouraged by the discussion I went ahead and finished Willy Hendricks's book Move First, Think Later (though I did skip the chess exercises). I even looked at some of the other books he referred to. I offer here a summary of what I think he shows, but the first thing to say perhaps is how similar I found the chess and go worlds to be. Which means the book can usefully be read by other go players, and indeed I recommend it. One special merit was that the author wrote without bombast and with great humour. One anecdote that illustrates both that and the similarities with the go world is when he remarks that the stronger the player the more likely they are to avoid clear cut answers (and he explains why). In contrast, weak chess amateurs are just as prone to make assertive pronouncements on top players' styles as our go kyus. One such player said, "In this position Garry Kasparov would play Bd4". Apart from smiling at recognition of the sub-species, Hendricks decided to look up a database to see whether Kasparov had encountered this position. He had. And he played Nf3!

There are two apsects for a serious chess player: playing and studying. Hendricks demolishes the old notion that when playing a good player devises a plan and fits his move to it. Rather, the player "sees" (or "recognises") the move to play. He may then check it, of course, but relies on seeing the move based on what he already knows. The more a player has studied effortfully (Hendricks doesn't use that word: his choice is to study with deliberateness) the more he knows and the more likely he is to see a good move.

Of course occasions arise when a player doesn't know enough yet and so can't see a good move. When a young player, who has not yet learnt to articulate his thoughts about a game, meets this impasse, he is liable to point to an area of the board and say something like, "I'm stuck over there". An older player can waffle on a lot more, but essentially he is equally stuck. The remedy is not "protocols" (by which Hendricks means not just structures, lists or systems, but also proverbs or other guidelines. The remedy is "content", i.e. more work on looking at actual positions.

Applying the remedy is of course shifting the focus from playing to studying.

The situation with playing being about seeing moves is now apparently regarded among those that matter in the cognitive sciences as clear cut (i.e. content rules over protocols) though there are still pockets of strong resistance. Apart from people like chess trainers who fear for their livelihoods (wrongly), there are those who feel philosophically uncomfortable with what the science shows. It appears that when you play chess, you may think you are driving the car, but in reality it is the "chess beast within". When you work by exposing your brain to thousands of chess positions, it somehow sifts out the "fundamentals" for itself. Under the conditions of playing a time-controlled game it is this neural network, not the conscious you, that is really making the moves. To make better moves you have to "feed", not tame, this chess beast.

But in studying different factors can overlay this. The conscious you may even be in more control here, but the current situation is that (unlike with the playing aspect) there is very little empirical knowledge as to the best way to study. Quantity is certainly a factor (as Hendricks says, quantity is a quality, too). But beyond that there is room for many ways to feed the beast. Hendricks, stressing there is no dogma in this, describes his own impressions of some of the various possible methods, based on his own experience of 25 years as a trainer but also on his extensive study (if you trust his bibliography - as I do).

First, talent and heredity do make a difference. It is not yet clear how, or to what extent. It may be small. General intelligence may, for example, come to the fore in how the infamous 10000 hours are used. But quantity of work (content) seems still to be the biggest factor.

Hendricks recommends playing over classical games as the (?main) way to improve. That way you are feeding the beast with high protein food.

Second best appears to be playing games yourself, but not blitz games. The real problem with feeding your beast with your own games is that you are giving it junk food - not just yours but your opponent's. This can be remedied to some extent by playing slower games (rapid as opposed to blitz is OK) but with two important caveats. The first is that you must analyse your games afterwards, preferably by physically taking notes. The second point is that you must include in your deliberations the variations you did not play (variations? what variations? I hear myself say!). However, he seems to suggest that this does not have to be a specially onerous exercise. To learn just one thing from each game you play may be enough.

In both the above forms of study (classical games, own games) the paradigm of the title - Move First, Think later - is evident.

The best form of study when you do this does not seem to be known yet, and the precise form may not even be important. Nor does the area of study seem to matter much, so long as you are not looking for instant encapsulations. Your task is simply to find qood quality food for the beast. It will sort out the fundamentals for you.

In that regard, what you are doing is not really building up pattern recognition. The use of the word "recognition" itself is a bit misleading. That goes on, of course, but what is really important is what your brain does with the patterns. Apparently no-one has a clue yet how the beast does this, but if you learn properly, once you see a pattern (or a similar one), the beast does not just ring a bell and say "Found!" but it serves up also a plan or a combination to make use of that position. In computer terms it seems like recognising the position is simply matching a hash number. You need to use that number then to look up (instantly) what is stored in a dictionary about that position. Like a hash-driven dictionary, it doesn't matter much in which order the entries are made, nor is it necessary that the objects be typed - they can be simply objects, on which you can apply a cast later.

Hendricks is totally relaxed about learning openings (ergo josekis?) by rote. The chess world has the same type of people as the go world who berate others for learning openings without understanding them. But (a) the beast will do the work of understanding if given enough time and data (like any neural network) and (b) since every game has an opening you might as well have some means to help you avoid treading on mines all the time - otherwise you'd never get to the other interesting bits such as the endgame.

That aspect apart, though, Hendricks did not seem to mention memory as an important element - not even in connection with top players who are known to have exceptional memories. That omission surprised me, although it does seem that a superlative memory is not a sine qua non among top players.

I was also a little surprised to see no specific mention of any differences in how to train amateurs as opposed to budding professionals, though Hendricks does say enough to suggest he sees no meaningful difference.

The evidence is very unclear as regards doing problems. Every chess trainer seems to swear by them. In fact, when you boil it down, collecting problems relating to certain themes seems to be most of what a trainer does. The rest can be summed up as marketing (understandable but usually phoney in Hendricks's view) and psychology. Hendricks appears to believe the latter is possibly the main reason trainers survive - students need a confidence boost, and those trainers who sound the most confident (or boastful) are the ones who get the most students, irrespective of the quality of their technical work. The least competent technically may, however, human nature being what it is, be the most useful if they have the gift of the gab.

But on that note, there is strong but perplexing evidence that studying technical positions, even with a top notch trainer, is not as efficient as expected. In particular, students who do well in solving problems in class routinely miss the same positions in actual games. The likely reason is that knowing a problem is a problem changes the student's mindset. Once he knows it's a problem he looks for a tricky and/or unique answer. Hendricks tried to overcome this by setting problems randomly. He made a list of positions simply by picking a game from a database in some random way and then choosing a random move within that game. This certainly worked in giving realistic positions, but since it produced many positions where there were many good but different options it seemed to have not worked as a training tool. However, and a little irritatingly, he is vague on the outcome. I imagine this is because he is trying to stress how uncertain - stripping away marketing bluster of the trainers, of course - the whole field of chess study is.

A further important aspect of both play and study that Hendricks passes over too quickly, I think, is calculation. He does mention it, and also the "Kotov bashing" that now seems to be the norm. Hendricks makes a thought-provoking point about Kotov's systematic, planned approach to analysing variations by suggesting it had a cultural genesis: Kotov was a product of the Soviet Union which imposed and stuck to plans on every aspect of life, even when they were mostly clearly failing or inappropriate. It is quite clear that Kotov's method likewise bears little relationship to reality, though don't overlook that a major and necessary aspect of the Soviet culture was the ability to read between the official lines. If you do that with Kotov, there is a lot of sense in what he says. Hendricks does not pick this out. However, he does point the reader to a book by Jon Tisdall which is dominated by a re-appraisal of Kotov. I bought this book, too, and am now reading it. It looks imporessive, even if some of Tisdall's methodology (he was a grandmaster but is now a trainer) can now be shot down if you accept Hendricks's views.

But in deference to Hendricks, calculation in his purview must come within the "Think Later" part. You see a move first, then you check it. The chess beast doesn't seem to do a good job of calculation, so feeding it with variations seems pointless. This is one area where chess trainers can justify their existence. It may also be the area where talent manifests itself most. Since Hendricks skates over this part of study, there are few clues as to how best to do it, but the Tisdall book (Improve your chess NOW) seems to have lots of concrete ideas. Blindfold chess and stepping stones are two (not sure what these are yet), but there was another approach that was of special interest to me as I got an honourable mention. Tisdall, being a chess grandmaster, found he could no longer remember the feelings of awe or bewilderment felt by his chess students. Once he became a chess teacher, he decided to try to recapture that feeling by taking up shogi (using my books). He did indeed feel bewildered by shogi, but what the lessons are I have yet to read.

Both the books I have mentioned here offer much sustenance for go players and trainers. I am certain there are other interesting and similar books, and I look forward to hearing about them here (recall that I am not looking for ways to improve my chess - I am interested in the latest advances in teaching approaches or in cognition, but at a "for Dummies" level). I am now quite certain - I wasn't when I started looking again at chess - that the differences between go and chess have no substantial impact at this level, and so we ought to make use of the prolific chess literature.

I hope I haven't misrepresented Hendricks too much, though if I have I will be happy to acknowledge corrections. But if you disagree with any of the points made, do remember not to shoot the messenger. Like heralds of yore, messengers on L19 have a precarious time.

In the meantime I plan to continue my perusal of chess works. If there's enough interest, I'll consider writing down my go-related views on Tisdall's book eventually. I'd also like to encourage others to do the same with books that impressed them. I mean a proper write-up rather than just a mention.


This post by John Fairbairn was liked by 12 people: Boidhre, dfan, fireproof, gasana, gowan, Konijntje, Ortho, palapiku, Phelan, Solomon, tchan001, Zombie
Top
 Profile  
 
Offline
 Post subject: Re: Recognition vs understanding
Post #62 Posted: Wed Jul 25, 2012 12:02 pm 
Lives in sente

Posts: 774
Liked others: 137
Was liked: 155
I just want to point to this old post of Sorin Gherman: http://www.361points.com/articles/novice-to-expert/ . It is about the skill acquisition model of Hubert L. Dreyfus, which isn't specialised on board games, was developed as a critique of so called AI research of his time, and notably has expertise conceptualized as "just seeing" as opposed to earlier stages that rely on rules, but of course this seeing is informed by a lot of knowledge and experience, just that often it can't be verbalized by the expert himself. Reading the last posts of JF this looked decidedly related.


This post by tapir was liked by: ez4u
Top
 Profile  
 
Offline
 Post subject: Re: Recognition vs understanding
Post #63 Posted: Wed Jul 25, 2012 2:12 pm 
Gosei
User avatar

Posts: 1581
Location: Hong Kong
Liked others: 54
Was liked: 544
GD Posts: 1292
John Fairbairn wrote:
Hendricks is totally relaxed about learning openings (ergo josekis?) by rote. The chess world has the same type of people as the go world who berate others for learning openings without understanding them. But (a) the beast will do the work of understanding if given enough time and data (like any neural network) and (b) since every game has an opening you might as well have some means to help you avoid treading on mines all the time - otherwise you'd never get to the other interesting bits such as the endgame.

So perhaps one of the ways of feeding the go beast is to take one of those huge Japanese joseki encyclopedias such as the Igo Daijiten or the Nihon Kiin Great Joseki Encyclopedia and just to look at the joseki within to try to think over the shapes, tesuji, and life and death situations of the many moves thought up by the experts of the past. And perhaps to even further enhance your study by supplementing with a comparison against some of the more recent joseki books. Since most people will not understand the Asian languages anyways, the diagrams would serve as a huge source of high quality protein for the beast.

_________________
http://tchan001.wordpress.com
A blog on Asian go books, go sightings, and interesting tidbits
Go is such a beautiful game.

Top
 Profile  
 
Offline
 Post subject: Re: Recognition vs understanding
Post #64 Posted: Thu Jul 26, 2012 7:56 am 
Gosei
User avatar

Posts: 1449
Liked others: 1562
Was liked: 140
Rank: KGS 6k
GD Posts: 892
I don't think I have anything to add to this thread, but it is very very interesting to read! I'd like to thank John Fairbairn and others for the great material. :)

_________________
a1h1 [1d]: You just need to curse the gods and defend.
Good Go = Shape.
Associação Portuguesa de Go

Top
 Profile  
 
Offline
 Post subject: Re: Recognition vs understanding
Post #65 Posted: Mon Jul 30, 2012 3:52 am 
Oza

Posts: 3647
Liked others: 20
Was liked: 4626
Yes

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 65 posts ]  Go to page Previous  1, 2, 3, 4

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Bing [Bot], Google [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group