Force Feeding

    Keywords: Life & Death

Tamsin: As of July 2011, I'd like to disclaim this idea. The pyschological research and experience just do not support it. At best, it's like using a sledgehammer to crack a nut. Review is better when spaced out, and it's faster over time to study small amounts often than to cram like this. Anyway, I leave the old discussion below for your reference.

OLD TEXT

Tamsin: I came across some problems headed "Korean Baduk Academy" on Gobase the other day. The contributor, Pierre Audouard, said that in Korea students go through all the problems in a level in about 30-60 minutes and then start over, performing the exercise 4 or 5 times.

This entails doing up to 100 L&D problems in a single sitting. I have decided to try this over the next few weeks and see what effect it has upon me.

What I have noticed thus far is this: it feels like the patterns and motives are being burned into my mind by force. There is something much more impressive about seeing so many ideas repeatedly in such a short time than learning them in a more gentle way (e.g., just doing a few problems each day).

I call this approach to learning "force-feeding". It can be painful, in a way, but can it be successful because it is such an intense experience?

Please share your thoughts with me!

unkx80: This is pretty much an Asian way of learning things in general, especially the Chinese, in what we usually call "cramming". It is often practiced in schools, where students will literally do tons of exercises over and over again on a topic so that they will remember them. It can be very effective for Mathematics, and probably also in Go. But my own personal experience is that such cramming is often only good for memorizing facts before an examination, and after the examination I forget them all.

Migeru: As any mathematician will tell you, cramming is useless for learning mathematics. It may be useful for passing exams, but I, for one, have never pulled an all-nighter in my entire life. I usually go to a movie and go to bead early the night before an exam, and in the morning I feel great and do great.

Charles: Cramming is known in the UK! In fact young people who fail important exams may be sent to a 'crammer' school to make sure they pass a retake.

I know that Korean instruction does place a high value on life-and-death study. I think the reason lies in the teachability of this kind of material: you can learn more and more if you really need to.

Here Matthew Macfadyen has advocated something that may be slightly different: high-speed solving of simple problems. I'd be interested in anyone's positive experiences of this.

Tamsin: Ah, yes, of course it is cramming. I didn't recall the term in this case because before now I have tended to study things in according to the principle of "little and often". This, on the other hand, is a "lot and often". Whatever you call it, I think the key must be, as I said, to make learning a really intense experience, laden with memorability because of that.

As an aside, I have used mnemonics such as the "Roman rooms" and silly mental images to remember things and do find them quite effective sometimes, but I have to say, I don't believe anything is quite so effective as good old-fashioned repetition. Even a thickie like me starts to remember things when they've seen them a couple of hundred times.

unkx80: Yes I agree. I think one reason why cramming works for life and death questions is that because many of them follow some rather basic patterns. As we repeatedly attempt the various life and death questions, the patterns will emerge and these are the ones that stay in the memory, and these are also the patterns applicabile in actual games. And this actually improves on intuition and often good intuition can help prune the mental "search tree" when a given position can be mapped into one or more patterns already stored in the brain. I suppose this is why some players are described as "sharp": their mental response to the opponent's weaknesses is fast because they are well trained in pattern recognition. =)

Grauniad: Tamsin, it's a bit tedious going through the [ext] Korean Baduk Problems you mention over a slow dialup line. Have you found any way to download them in bulk or to view more than two at a time so one is not bothered by the network delay?

Tamsin: You can download them and use them later, but I don't know if it's possible to do it in bulk. Sorry to sound smug but I'm on an ADSL connection. A good alternative would be to divide 1001 Life and Death Problems into 10 sets of 100 problems and to subject each set to the force feeding process.

Tamsin: Another aspect of this approach is "overlearning". That is, once you have understood something, you continue to study it anyway.[1] This, in my view, paves the way to gaining a much deeper understanding. Put it this way: I can solve the problems on the Baduk Academy page without too much difficulty, but I do sometimes need to think for a while, and I do make mistakes and I do make the same mistakes more than once. In contrast, a Korean kakyuusei will solve these problems very quickly and with few, if any mistakes. Because he has "overlearned" the shapes and themes involved, he is able to recognise them and work on them much more quickly than an amateur, who may indeed already know and understand them, but not to the same extent. Compare this process to learning a script such as katakana or hangul: it does not take long to learn how to do it, but to achieve the ability to read it effortlessly requires considerable repetition over time, i.e., overlearning. I am a believer that although well-chosen study techniques are very important, even more crucial is the willingness to keep applying oneself, over and over and over.

You will note that I have mentioned on the Principal Study Technique page that I repeat sets of problems over 21-day cycles. This is because repetition over a 21-day period, without fail, has been proven necessary for changing established habits. Although I doubt very much that anything like this much repetition is necessary to gain benefit from new material (as opposed to altering established habits), I strongly suspect that the same time frame is optimal for a course of repetitive "force feeding".

ChipsChap: Sounds a lot like 'spaced interval recall' which is the heart of some learning techniques, where you repeat often to begin with and at various carefully calculated intervals later on. There is a commercial product called SuperMemo? for this (probably others too). I think I might take the Baduk academy problems and try this out.....

Tamsin: I also recall now that my force feeding approach resembles the method of chess study proposed by Michael de la Maza in Rapid Chess Improvement. His programme consists of "seven circles" of tactical drills repeated over and over again until they become impossible to forget. He started at 1300 USCF and shot up to 2000 USCF within a short time; he recently won a large cash prize in recent tournament. This all underlines my point about overlearning to achieve a much deeper level of understanding: e.g., everybody who plays chess knows what a fork looks like, but most people still miss them or at least fail to take them fully into account -- only by seriously overlearning can one spot forks of all kinds intuitively.

Bill: I read an article about cramming in a psychology journal last fall. People are still doing research on it. The myths about cramming are true. You can learn a lot by cramming, but you forget it quickly. Overlearning is different. Overlearning is good, cramming is bad.

One idea about why cramming is bad is that the intense, focused study does not provide enough obstacles to retention. Something that I have wondered about in regard to the study of go problems is whether different kinds of go problems are different enough to make learning more difficult and more durable. If so, instead of studying, say, 100 life and death problems, it may be more effective to study a total of 100 problems of different kinds: life and death problems, tesuji problems, endgame problems, fuseki problems, hamete problems, invasion problems, next move problems, etc. It may be that these different kinds of problems interfere with each other sufficiently to overcome the cramming effect.


[1]

Charles: I think this is suspect, for many aspects of go. I don't think one should 'hardwire' one's brain with joseki, for example. I think one of the obviously less desirable aspects of server go is the way people play quick, thoughtless games relying on tactical sharpness. Consider that the way to advance in go is to be playing in a balanced way. It is certainly a fault obvious in many European players, that their tactical ability has outrun their general strategic understanding [2].

Tamsin: Tactics is probably the easiest thing to improve, and probably the thing that makes the most difference. Usually, when playing with somebody stronger, it's by tactics that one is defeated. Besides, the Koreans seem to lay great importance to tactical training, and who's going to argue with them over the best way to learn and play baduk?

Charles: Well, an informed discussion would be good. It's a country with 10000 amateur 5 dans, to begin with. Are we talking about how they get so many of those; or how some of those are hot-housed further to pro level? The striking thing, for a visiting EGF 3 dan, is how few players are at that level: people are weaker, or stronger.

[2] Dieter: I respectfully disagree. I think most Western players of a more experienced level have read those books about fundamentals, thickness, Chinese opening and so on, thus gaining a broad but superficial understanding of strategy. They play for fun, hence read for fun and acquiring tactical sharpness through tsumego is not what most players call fun. So I think the strategic understanding generally outruns their tactical ability, to the extent that after a few moves they are fully capable to appreciate the undesirability of the position - if only they could have read that far. (they include me)

Charles: It isn't how I see it. Most EGF 3 dan and 4 dan players are relatively strong at fighting, but don't know what they are doing.

Perhaps we are talking about different groups. I think EGF 2 dans probably have read the books (which is good, in itself); but make the mistake of believing them, or at least taking them too literally, too dogmatically. For example they read the Kato book on the Chinese, which is over 25 years old now, and assume that is 'theory'. No, it was a cheap book to write.

Dieter: Perhaps, yes. I do not meet that much 3 or 4 dans, unfortunately. So I may rather be talking about the 6k-2d group.

dnerra: Well, I certainly agree that the strategic understanding of most EGF 3 or 4 dans (including me) is worse than yours, Charles. :-) But by what standards do you want to measure the importance of strategical vs tactical understanding? This could be a long discussion, but I'll make a more polemic claim instead, which should illustrate my point:

Most strategical (or the most costly) strategical mistakes by 3 dans aren't made because of a misunderstanding of strategical concepts, but because of tactically misreading the situation.

This includes attacking a seemingly weak group although it is completely alive locally, underestimating aji, shape misunderstandings, etc... Of course I use a broader sense of tactical ability than "skills in winning semeais" here.

Tamsin: I wonder if it might be possible to devise a programme of force feeding to overlearn strategic ideas. Then, would the resulting deep knowledge be a bedrock for strength? Or a straitjacket making it very hard for the player to spot exceptions where they occur? That said, I read somewhere in a chess commentary a comment very apposite here: something to the effect that the bulk of grandmaster errors are the results of finding false exceptions to a rule. That is, unless your judgement is on an extremely high plane, you're better off sticking to the fundamentals all the time than trying to find exceptions.

(Now, saying all this seems to go against my grain as an academic, in that I believe passionately in the value of independent, original thinking; but in structural games like go and chess, I am inclined to believe that it is better to rely on the principles laid down by the independent, original thinking of the great players who created them than on one's own.)

Charles: Well, Tamsin seems to have started an interesting discussion here!

I base my comments largely on conversations with Kim Seong-june, who has now returned to Korea, but who played for a number of years in the UK at an EGF 2500+ level. We wrote two books and many tournament booklets together, so I think I eventually understood his point of view. Something significant, to me: he did some teaching based on games from the British Championship, and at those sessions the British dan players mostly asked him questions about reading. He had to put his strategic thoughts in a BGJ article!

The whole business of trying to put middle game play on a routine basis - suggested by Tamsin's comments - does seem hard to me. One ends up listing difficulties: the joseki context idea becomes a monster when the centre matters; large-scale exchanges can freely occur; ko can become a matter of policy; aji (nerai, as we're encouraged to look at it) cannot I think be handled simplistically; framework theory is ruddy enormous, to take something on which I have posted consistently here. And those comments all have some interaction.


Tamsin: It seems from the above that force feeding is more suitable for learning the tactical and technical elements of the game rather than the broader strategic issues. Being of a pragmatic mindset, I would like to develop my experiment with force feeding into a systematic study programme for improving your go. I will work primarily on myself and hope that I have yet the requisite potential to improve; for those who wish to follow me, I promise nothing but blood, sweat and tears, not necessarily in that order. I am convinced that even fast progress takes a fair time to be noticed and that no matter how hard you work, you will need to be patient. Anyway, here is my first tentative suggestion for a force feeding study programme:

1) One set of 100 Life and Death problems. Work through this set once a day, without fail, for 21 days.

2) One set of 50 shape problems (a good source is Making Good Shape. Work through this set in the same way as outlined above.

3) One set of 3 joseki. Study each joseki and try to understand the meaning of each move. Explore alternatives. Spend between 30 and 60 minutes on this. Again, repeat the same study every day for 21 days.

4) One chapter of Nagahara's Strategic Concepts of Go. Read once a day for 21 days.

4) After the cycle of 21 days is complete, give yourself a short holiday.

5) Find new sets of material and begin afresh.

6) Find time peridiocally to revise stuff from earlier cycles.

Remember, if it ain't hurtin', it ain't workin'. (So much the better if you enjoy this kind of study!) Good luck!

Please suggest your own regimens below:

Harpreet: One thing we know from research on memory is that repetition is good for remembering things but spacing out the repetition is even better. So maybe several different sets of problems that you go through and by the time you come back to the beginning you have had some time to potentially forget.

Tamsin: That's why I suggested periodic revision in note 6. :-)


Tamsin: Having now completed my first major cycle, I can share my first observation. When you do a large set of problems repeatedly, you become aware of recurrent themes and techniques. These themes acquire a much stronger identity within your mind than they would have if you encountered them in isolation. Anyway, among other things, here are four distinct tesuji with which I am now richly familiar, but which previously I only knew in a vague way.

Pulley

Crush

Eye by Exhaustion

Stomach Eye

Geoff Wedig: The idea that recurrent themes become apparent is well known in learning. As a common example, a child, when learning verbs, first learns the past tense by rote. At some point, the common rule of 'add -ed' gets recognized, and all verbs get -ed even the exceptions (I go-ed, when they used to know 'went') Eventually, the rule and *when to apply it* gets reinforced, after many repetitions, and you just know that the past of go is not 'go-ed' but 'went'.

Velobici: Does force feeding work for all levels of difficulty? There are problems that one can solve on sight. Others may take a little reading. Still others may take an average of 3 minutes per problem to read out, examine the answer, re-read to verify the answer and re-solve. This results in only 20 problems an hour. (An hour is about all I can devote to this activity each day.) Or will this result in memorizing these 20 problems only, without any ability to apply the results to other positions?

Tamsin: I don't know yet if force feeding works. My rating (not yet rank) on IGS has twice reached a new peak in the last week or so, but whether I can break through to 1k there and hold on to it remains to be seen. One effect of repetition is to make you question even the simple things. Many problems are easy to solve on sight, but when you repeat them you start to wonder "what if?" and you begin exploring other possibilities, and I think that has to be a beneficial process.

KRITZ: Do you have a suggestion for a source of good force feeding problems? I suspect this would depend on my level...

Tamsin: If you're a beginner, then I guess Graded Go Problems for Beginners would be good fodder. If you're more experienced, then 1001 Life and Death Problems. Also, the Korean Problem Academy cover all levels from beginner to 5 dan.


22 August 2003

Tamsin: I am coming to the conclusion that it is not necessary to go to the extreme of doing the same set of problems every day for 21 days without fail. That kind of practice, I believe, is more necessary for overcoming and altering existing habits, such as changing a golf swing or improving one's violin-playing technique. However, a lot of repetition, particularly if mixed with revision, is probably much better than going though problem sets and book chapters only two or three times. Any comments?


PROGRESS REPORTS

26 July 2003

Tamsin: I can say confidently that I have got stronger recently. I made an effort to control my emotions and to think calmly about my moves, and that has certainly helped, but I'm also seeing a great many more tactical possibilities than I used to, and this is surely the result of my force feeding programme. Also, having read the chapter in Strategic Concepts of Go on aji repeatedly for 21 days, I find myself much better at leaving situations be until they're ripe.

31 August 2003

Tamsin: I reached 1k* on IGS for the first time. Also, at the time of writing, both of my KGS accounts are at 1k. I am now certain that force feeding works. Currently I am using the SegoeTesujiDictionary.

Henge?: I too began this process of "force feeding" sometime ago, concentrating on the Korean problem set under discussion. I have been through all the sets 4 times, except the last which I am about 3/4 of the way through. I see value in the rapid execution of these problems: there is no doubt in my mind that this process contributes towards the imprinting on the mind of common shapes and tactics. Having said that, I see more value in reading more completely -- exploring the full range of possibilities. Reading very quickly, I tend to answer the problems correctly, but I can see that sometimes I get the answer right, but for the wrong reason, because I didn't consider a potential counter to my move -- even if it doesn't work. My anwer was right, but I am taken aback by a counter I didn't see, and the chief value of this is to teach me to be more careful, to read more completely. It seems to me that to get this lesson I have to slow down. By way of summary, I think the objective is not to knock off a set in 30-60 minutes, but to knock off as many as one can in 30-60 minutes, reading them completely. Speed will come. If one spends enough time -- I think 30 minutes twice a day is best-one gets the pattern recognition benefits, but without developing the negative habits of shallow inspection associated with the point-and-shoot school of problem solving.

Henge?: In e-mail correspondence with compiler of Korean problems, learned that we have only a subset. Actual set for each level is closer to 700 problems -- in time we will see the Gobase set enlarged to approach a full set. With 700 problems per level, I think that it is clear that Korean program does not consist in doing one level in 30-60 minutes, but, rather, biting off 30-60 minute chunks of each level at a sitting.

Henge?: Ok, found another really super source of tsume-go: Cho Chikun's Encyclopedia of Life and Death. It's a CD with roughly 3000 problems on it, broken into various categories: beginner, elementary, intermediate, advanced, and strange puzzles. There are more than 700 problems in each of the main levels: elementary, intermediate, and advanced. Thus far I've done only the elementary set, and if it is any example, these problems are definitely a bit tougher on average than the first, and maybe second Korean set. For example, the very first set of the Korean problems focuses heavily on extremely basic shapes, three in a row, four in row, bent fours, etc. As useful as intense drilling on those shapes may be, that emphasis is almost entirely absent in the Encyclopedia of Life and Death. It more or less gives a nod to them in the 200 or so beginner level problems--most of which are truly designed for what I would consider a complete beginner (black to capture one stone--that sort of thing), and then moves beyond them for ever. By the next, "elementary" level we are seeing problems like this one, problem 405. While it isn't tough, it is clearly more challenging than the first--and perhaps second set of Korean problems.

The interface is pretty good, though it's DOS based, which tells you how old the program is. It gives me a start-up error every time I run it, but it does run.

The solutions given are more elaborate than the Korean set. It gives a fair bit of detail in the answer, even discriminating between solutions that are ideal and solutions that work but are sub-ideal. For one problem it marked my answer: "semi-correct," because while with my solution I lived, and with the same points as in the fully-correct answer, my solution did not as in the optimal one leave the opponent with a bit of aji. So the solutions are well thought out and quite instructional.

By way of summary . . . the experiment in force feeding continues and for the rest of you binge eaters, if you want a large source of good problems to carry around on your lap top, this is it. Enjoy!

HelcioAlexandre I enjoyed much of this discussion and started the experiment too. I'm doing mainly 2 things: solving as many life and death problems that I can and memorizing the games of GoSeigen book (I'm still memorizing the first one).

erislover Overlearning has helped me with two things. First, it has improved my general reading ability in terms of depth (that is, I can read deeper in the same amount of time). Second, using less time on reading tactical sequences leaves more time for strategic concerns where simple reading does not quite work.


Rberenguel: I'll try my own program, based upon the one of Tamsin, basically the baduk problems (now level 2 each day for a week, a hundred one day, the other the next and so on), chapters 1, 2, 3 of Tesuji (3 times each in a week), and one chapter of attack and defense (every day of the week). Maybe if I get tired of the chapter I'll go to the next one.... Guess what happens... But I think that won't hurt my level.

UnderTheStonesProblem13

I have been using Cho's L&D encyclopedia now for quite a while. I think it is the single best resource I have found for studying technique. That said, the program does not run on modern machines without help because it was written for DOS and, I infer, makes some fairly machine specific calls. I first started using it on an older Pentium running Windows 95, and it worked fine. Then when I upgraded machines to a 32 bit system I couldn't get it to work. I tried other people's newer machines... same thing. Hint: if you want to run this program on newer machines, you're going to have to use a DOS emulator. I use DOSBOX. With this emulator program I can run Cho's encylopedia on all my machines. Good thing too--I was keeping the older machine around for only this one program. Now I actually carry the program around on a memory stick so I can plug it into a USB port and use it on any machine anywhere. If you like L&D problems, this is a great program. Good luck.


Force Feeding last edited by 1.72.1.136 on July 6, 2011 - 13:33
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library