Sunday, August 26, 2012

Organization of knowledge

When knowledge is organized, it is easier to retrieve it from memory.
The better it is organized, the more cues are formed, the easier the retrieval and the faster the retrieval.
Verbalization is one way to organize the knowledge of a chess position. But we already know, that a verbal description is not the most logical way to transfer knowledge from one problem to the next. Besides that, language itself is inherent limited in describing a chess position. It can be used to describe the core ideas, but the details become problematic. That is why I want to use a second method to describe the position: graphical representation. The use of two methods for the organization of chess knowledge has an extra advantage: it provides more cues, hence faster and better retrieval.

When you work out all variations of a given position rated 2000+ the amount of variations usually exceeds hundred moves by far. Of which every move has its own consequences and repercussions on the position. It is possible to compile the essence of the variations in a few chess diagrams with arrows and coloured squares added. In language you would need 100+ little descriptions what the moves are doing. That is not realistic.

I don't know what is the most essential that must be represented, since the possibilities are endless. Probably the mere compilation of the diagrams themselves is the best way to store knowledge to memory, but in the end you will have a file with diagrams which can be used for some repetition if that is needed.

I compiled a few diagrams from different points of view. I haven't decided yet what are the most usefull diagrams. I suspect when working with it you will see a shift overtime. In the beginning you need for instance a way to represent the overworked pieces. But once the recognition of overworked pieces has become automatic, there is no longer the need to document them, since you simply see them. You can find the original problem at CT here.

Diagram 1. The arrows show where the main attackers go
You see that the diagram is some sort of time machine. You see not only how the position is now, but you can see the result of moves too. This provides a glimpse into the future. A representation like this is much less demanding for working memory than the managing of individual moves.

Diagram 2. The coloured squares to show where the main attackers go.
Coloured squares in stead of arrows might give a quieter look. It will become problematic when the attackers converge, though.
Diagram 3. Invasion squares.
Here you can see where two attackers converge on the invasion square c4. The green squares represent the defenders of said square.

Diagram 4. Counter attacks.
In diagram 4 you see the knightfork Ne4 represented in yellow. The arrows show which counterattacks this move would unleash.
Diagram 5. From attacker to target
In diagram 5 you see tactical elements like the two pins and the knightfork in one glance. The blue squares are the focal points of the position.

Diagram 6. Immobilization.
When you think about the combination and the move order, diagram 6 comes in handy. The move Rd4 binds both the knight and the queen to the invasion square c4. It happens to free the black knight from the pin too. The move Qe2 immobilizes the knight even more and binds the white queen to the defence of the knight. After Rd4 and Qe2 the white queen and knight are totally paralyzed and ready for the knightfork Ne4.

Diagram 4, 5 and 6 together seem to represent the core components of the combination pretty well. If we are able to see these invisible structures fast, we would have a whole lot of more time to think about the details, with a working memory that is way less overloaded.


  1. Looks like the diagrams from the movie "a beautiful mind".

  2. To me, this position seems to be in the murky swamp between being a tactical theme and being an attacking theme. Ironically, I think this is one instance where it would be easier to find the move in an actual game than it would be to find on a tactics server. That's because on a tactics server the mindset is to be looking for the immediate pin or skewer or discovered attack, etc., but in real life we are much more accustomed to situations where we might not have an immediate tactic at hand and so we have to go through a planning stage, or in other words introducing tactical elements into the game. Of course I could be wrong...
    I found your blog a couple of weeks ago and enjoy it very much.

  3. @Munich,
    you mean that documentary from the BBC?

    thanks for the cheering!
    At first the combination looked very murky to me. But after making the diagrams it seems much clearer and "cleaner" now. Allthough the mind is still somewhat wobbly with it, so I doubt if I will find it in a game. Yet it feels like progress.

  4. Today I got some feedback from my brain. I think I understand it (=my brain) a tiny little better than before.

    Do you do "check mate in 1", "mate in 2", "mate in 3" puzzles in the difficult range, too?

    I expect here better results in understanding/ learning, than for the difficult range of non-mates.

    Since you did hard puzzles for a while now, I would be curious if you have a vague feeling that the check mate puzzles are more effective?

  5. I have done all mate in x int the range above 2000 (all 159 of them).
    I used to be very terrible in king chases, but this certainly helped. I have repeated them a fe times.

    I have done all discovered attacks 2000+ all 179 with a total recall of 100%

    I'm now busy with double attacks 2000+ (about 350 problems)

    I can't say which are more effective. I'm bad in everything.

  6. Of those 3 I believe the "discovery" set will be the least helpful set.

    Of the mates I believe the mates that are not much of a king chase but just a bit odd/tricky even thought they are only 2 or 3 movers - here I could imagine that they are very helpful.

    This is a new assumption of mine. It does not change what I previously said, but I have now an explanation about which ranges are helpful under which circumstances.

    A check mate is a clear target.
    Especially if it is achieved in a few moves (check mate in 2).
    You dont get distracted by other patterns.
    You very much focus on squares. You look at the activity of your pieces, and every square counts!
    You get a feeling for the power of your pieces.
    It doesnt matter that it is a king that gets hunted. Maybe trapped pieces would work as a training, too. But then again, a check mate is never tagged wrong, nor do other elements (distracting/misleading rules) matter in a check mate puzzle. It is probably more helpful if you now how many moves you have to check mate. If it is a check mate in 2, you really need to find this very check mate, and not the variation where you will check mate him in 5 moves. You need to check the squares very accurately.
    So better isolate the mates into different sets: a "check mate in 2" set, "check mate in 3", "check mate in 4 or more moves".

    After you solved such a puzzle, the repetition step starts.
    With reviewing these puzzles you will look at the only relevant sequence/variation. You dont look at other variations, but only at the variation that is played.
    So you will only memorize this variation/ sequence. It is not memorizing literally a move order. The very sequence makes sense, but is also likely the only thing you will be able to memorize in an SRS.

  7. @Munich,
    that's old school stuff. Has nothing to do with the subject.

  8. True, it has nothing to do with the subject. (Unless you interpret the titel "organization of knowledge" a bit more open than to just focus on the best solution for diagrams).

    The old school stuff is so old, that it got meanwhile forgotten - which makes it a very good "phoenix from the ashes"!

    I finished more or less my SRS of "all puzzles I failed at least once" in the CT Blitz rating range 1475-2000.
    Though this isnt the very hard candy like you do it, it is at least not an easy range. The average rating might be at ~1750.

    And this is what I found out what happend when I did my SRS set:

    All I remember (but at the same time very much understand!) is the main variation that is actually played. All side variations ("alts" and similar) I often only understand 1 or 2 plys deep. Unless I already know the side variation as a pattern. Like "...and in case he does not go there, then I have the smother check mate in 4 moves, so he cant go there."
    But often the side variation has no certain clear pattern, and so I simply ignore it (unfortunately, and not on purpose. I ignore it unconsiously).
    I found out that as soon as I find the main idea of the puzzle, that I remember the puzzle. The idea triggers my long term memory. Not the other way round! This means I dont remember the puzzle at know the solution, but it is the other way round - I find the solution and remember that I had seen the solution before.
    My memory give me the "confidence" of doing it right. I see the pattern pretty clear.
    But as soon as this memory is triggered, I only remember the main sequence (= have only the confidence for the main variation). Which is only a part of it. If the CT player would move different, I dont know the solution. But since I allready found and remembered the main idea, I simply now the right move - with ignoring the not played parts of the tactic.

    I didnt realise that I droped little by little some understanding of the puzzle. I got faster, but a huge part is due to dropping some understanding. This understanding is not needed for the solution, so my brain was filtering it away. Not important for the solution - not remembered.
    It took me long to understand that this happened in my brain, because it is not a concious process. It sneaks from behind. You have the feeling that you understood it all.

    How did I find out then? --> By coincidence. I explained a puzzle to my son. He asked back.... and I did not know! I didnt even think about it.
    All I knew was, that I knew the answer to his question some time ago, but I also knew it is not relevant, because the black king would not move there anyway.

    This is not literally memorizing because what was played I understood.

    After I found out my learned knowledge is just enough to get the puzzle green, but misses half of the understanding, only then I checked carefully. Suddenly I found lots of missing understanding. The SRS made me blind in a way.

    But not always!
    I found out that for check mates I had a full understanding. So I could see a clear difference in learned understanding between a 1800 rated non-check mate, and a 1800 rated check mate in 2.

    Is this really old school?
    Would you like to think about it?
    If you think about it - please share your reflection with me.
    Any opinion? I dont mind if your next post is about this (if you like). Or simply discuss it here. Even if it is off-topic.

  9. I give you an example of what you might not have learned when you did your SRS of DA 2000+.

    You remember this puzzle as soon as you see the possibility Nh6+.
    At that stage it is probably a bit difficult for you to see what came first: remembering the puzzle, or first seeing the Nh6+ possibility and only then you remembered that you have seen the puzzle before.
    Never mind.

    Here the whole, I need to point your nose on it, otherwise you might not even see it (I told you, it sneaks from behind, the dropping of knowledge/understanding).
    Lets stop at move 2.Nd6!
    What do you move if black response was: 2...Rf4?
    Do you remember now your understanding? Sure you looked at 2...Rf4 before (or maybe you dont remember that you knew it some time ago).
    This possibility might have dropped. It got lost!

    No doubt you can understand why black moves 2...Qe5, though. So here you have understanding. Qe5 is threatening to take the white rook g3. It is a counter attack.
    The best black response in this situation.

  10. @Munich,

    The problem is that when we talk we make use of different premisses. That's why we never can agree about something.
    This are the two premisses I challenge:

    Repetition is neccessary.
    One needs to learn a huge amount of patterns.

    Don't take it too literally, but see the idea behind it.

    When you focus on recognition, there is no need to repeat something. You have to recognize it.
    Prof. de Groot put us on the wrong foot with his 50.000 patterns or chunks you need to know. There are only about 7 patterns you need to know, which are used time and again. If you want to split hairs, maybe 20.
    What you must learn is to recognize the 7 patterns under all circumstances and fast. This means that you will have to identy what impediments prevent that and take those away. That your unconscious mind might end up with 50.000 chunks underwater is irrelevant.
    Since that is why we call it unconscious.

    Given this difference in premisses, we can never agree.
    The verdict about the premisses we can make when I have reached 2000 rating.

    Having made that clear beforehand, I now take a look at what you actually say.

    Due to the comments of my post "understanding meaning", I finally came to the conclusion that it is possible to bypass verbal understanding concerning tactics.
    Which means that you can encode the essentials of the position direct into you memory. That is what this post is about.
    In a sense you have found the same. You encode the solution, and you forget the underlying verbal reasoning.

    I think that this is the reason why grandmasters often are bad in explaining their moves verbally. If you ask them about the reason why they made a certain move, they answer in variations.

    I found that I often recognize the essence of the position before I recognize the geometry of the position with my new trainingsmethod.
    That reminds me of how blind players report that they don't see the pieces in their head, but that they are aware of the "essences" of what is going on in the position.

    Verbal reasoning is only a bleak surrogate that simulates this process in the conscious mind. Maybe you remember that I talked about visualization guided by narratives?
    The real deal is bypassing verbal reasoning and work with the "essence" in stead. I haven't found the right words yet, but I hope you understand what I mean. When you bypass verbalization, you get what I said about blind players.

    About the problem you mention:
    You told me that the cue that triggers your recognition is Nh6+.
    I usually miss the first move and pay no attention to it.
    When I repeated this problem because you mentioned it, the essence of the solution came to mind before I remembered which problem it was. Only when I had a closer look at the pieces, I remember it.

    You are right that the unimported sidelines are not remembered but reconstructed. If I see an important sideline at study time, I make an extra diagram for it.

    So both your and my method works, yet we will have to wait before we can decide which method is the most effective.

  11. I agree that I did these 2 premises.

    However, a few days ago, I found that they are not the one and only truth.

    Instead I guess that I partially found when my 2 premises "rule the world" - and when they dont.

    For those puzzles where pattern recognition and repetition dont work (although you need to know of course many pattern, but I assume we know enough of them) - for those I have not found a key how to learn them. I know that they are there, and I can see that your aim is to learn them.
    The trouble I have with this attempt is, that chess tempo is not giving the filter that is needed most for that kind of puzzles:
    Filtering for puzzles with certain material constellations. For instance puzzles with queens on board and puzzles without queens.
    puzzles with no knights, puzzles with the pair of bishop.

    I could imagine though, that the filter "check mate" is a stronger filter. How strong is it? I am not sure.

    Of the non-check mate tags, I would guess that "advanced pawn" might be more useful, than "fork or discovered attack".

  12. @Munich,
    The premisses are not so hard as you use to wield them nor is my denial of the premisses so hard as I make it sound. If you can accept that, maybe only during a short time for the sake of reasoning, your mind will be open enough to learn a thing or two.

    For every problem that I don't solve correctly or which takes me too much time, I make a diagram as shown in the post. And belief it or not, my recall has boosted.

    The making of a diagram forces you to define the essences. What are the attackers, what are the vital squares, what are the targets, where to invade. It changes the way you look.

    When I do the same set a day later, my recall of the problems of which I made a diagram is much better than of the problems where I did not. That is strange in a way, since the latter were problems I was able to solve within a reasonable time, unlike the former.

    It is all about encoding. When you leave a problem, and you have nothing encoded, you will take nothing to the next problem. A verbal description is not encoding. Making a diagram can't be done without encoding though.

    It has a remarkable similarity with the flashcards of Rolf Wetzell, when you think about it.

    Tags are not so usefull in the upper region since it are combinations which will have multiple tags by definition.

    When you make diagrams, the whole tag thing is not so important. You focus on the function of the pieces, not on their geometry.