Tuesday, September 12, 2006

Surprising discovery

Today I had a closer look at my performance at CTS.
I did a series of 70 problems. I had 15 errors. The average solving time from the other 55 problems was 8.4 seconds. During the solution of the problems I counted the problems that I recognized. Much to my surprise that was close to 100% of the problems!

So the problem is not that I have not memorized the patterns, since I have.
The problem is that my braincells still need 8.4 seconds to release the answer at average!
That's why my rating doesn't increase at CTS while I learned more patterns.

The question arises if this is a matter of stockpiling the patterns in the wrong memory or is it a pre-phase of storing them in the right memory?

16 comments:

  1. it's probably impossible to keep repeating and not commit them to LTM...

    maybe the neural mechanism of tactics recognition itself gets faster by training? so you'd gradually get better at tactics, somewhat irrespective of complete memorization? sounds plausible to me. self-organizing neural net optimizing itself to do the job better and better.

    ReplyDelete
  2. Worm,

    it's probably impossible to keep repeating and not commit them to LTM...

    yes, but to WHICH LTM?
    The amateurlike declaritive explicit conscious long term memory or the grandmasterlike procedural implicit unconscious long term memory?

    ReplyDelete
  3. I wouldn't be surprised if, in chess, the two were very important. The procedural (which is easy to train quickly: there is a famous study where people were able to learn 10,000 pictures, just to identify via quick recognition, in ONE day). Declarative memory, though takes at least a month for consolidation. It sounds like, since you have the quick recognition memory, it may take a month or so for the declarative to catch up so that it can guide your saccades etc..

    ReplyDelete
  4. Blue,
    I'm sure we have something important here. But I don't understand your comment. I hope you will take some time to explain it to me. Probably I have a wrong idea about the situation, please correct me.

    This is my line of thought:
    Learning chess is comparable with learning a complex motor skill.
    If you learn to play piano, first you have to think about the movements of your fingers, the pitch of the notes on the score etc.. I should say this is a typical activity of the declaritive memory. You have to formulate the things you want to do first. The release of information of the declaritive memory is slow, explicit and conscious. So you can only play very slow. Just the problem I seem to have with CTS. Although I recognize the position, it takes me an awfull long 8.4 seconds to release the solution.

    When you exercise the piano, the movements of the fingers and the reading of the score are taken over by procedural memory. Which releases the information much quicker once it is stored there.

    What you say seems to be the other way around. Where am I wrong? Have you more information about that famous study?

    ReplyDelete
  5. I was speculating that chess isn't as simple as playing a piano. For one, semantic/declarative memory isn't always slow and effortful. For instance, what is the English word for a female dog? Most English speakers will immediately think of the word 'bitch', which is in semantic memory (a form of declarative memory).

    Perhaps, with chess, there is a similar rapid mostly unconscious access to semantic memory, and this guides our saccades to relevant locations on the board.

    So there are two phases to successfully solving a problem. First, rapid pattern recognition (which is not obviously procedural in nature: when we recognize a person's face, we are probably activating our semantic memory of that person), and second a (time varying) period of inspecting the position to make sure it is indeed the right pattern. This involves memory, not of the global characteristics of the board, but other memories of what the key theme is (pin/deflection) and consequent inspection (via saccades) of key squares to make sure that the board in front of you is actually a good match to that stored pattern. Perhaps this second stage takes longer to consolidate in memory, longer to become automatic than the initial stage of learning to recognize global patterns.

    This is just a possibility I've been thinking of since I read a bit about the psychology of saccades in chess. The amateurs make saccades to irrelevant squares more often than masters, for instance.

    ReplyDelete
  6. Dam I thought I posted a response and it didn't show...

    I was speculating that chess is more complicated than just riding a bike, in that it consists of rapid processing that involves both procedural and declarative memory.

    Consider even recognizing a face: it has both types. We can recognize a face as a face very fast, even if we don't know who it is. However, we also recognize it as someone (e.g., your mom) very quickly too, even though that involves accessing a semantic/declarative memory that we have stored. Similarly, think of the word for a female dog. If you are really good with English, you don't have to think: the word 'bitch' comes up immediately, even though it is stored in semantic memory.

    So, in other words, the distinction between declarative and procedural memory is orthogonal to the distinction between rapid/effortless and effortful/conscious processing.

    So too it might be with chess. Based partly on the chess saccade study I mentioned, it seems our chess vision has two components. First, seeing a global configuration on the board that we vaguely recognize. This is very fast, gestalt-recognition, perhaps procedural in nature. Second, pulling up specific (semantic memory) components about what exactly we are recognizing: e.g., mate with the help of a bishop pinning a piece. This semantic memory triggers saccades very quickly to the right square: check to make sure the bishop is pinning the piece (say, the way we would identify a twin by saccading to the scar on Jane which Nancy doesn't have).

    It is known that semantic memory takes about a month in the hippocampal buffer before it is consolidated. Perhaps you've trained up on one component and the other needs to catch up.

    I have a paper version of the article called Learning 10,000 Pictures. I'm not sure where I put it. If I find it, I'll post the abstract.

    That paper is amazing. People recognize thousands of pictures with very little effort put forth in memorizing them. If this is true, why I am so bad at chess? What, exactly, is 'recognition'. Knowing you've seen it before? That's what they tested in the study. But that's not helpful in chess unless you know what you are supposed to do in that position.

    ReplyDelete
  7. Crap my initial post just showed up. Well, they are slightly different ways of saying the same thing, and with slightly different emphases so I'll leave them both.

    ReplyDelete
  8. Blue,
    thx for the explanations. However it is complicated matter, I think its important to know precisely how it works.
    For instance, I probably recognize the problems for 98% during the past months without knowing that not the recognition is my problem, but the speed of it is too slow. If you look at Clyk at CTS, he has done 90,000 problems at CTS, probably catched in the same trap as I'm trying to escape now. He tries to speed up at the cost of correctness, his average solution speed is 5.3 sec/prob vs my 8.4 sec/prob. It doesn't help him either.

    So it is easy to spill months (years) with slightly the wrong training. I use MDLM as a beacon how to do it. He used only 1029 probs for repetition.

    In my next posts I will reformulate the matter in my usual way, to look at it from different angles. I hope that Blue is willing to comment where I abuse science, and other readers too for inspiration.

    If we crack this nut, we can make a big step, I'm sure.

    ReplyDelete
  9. Why not to think that you improve in chess like in any other sport, like football for instance where both Argentina and Netherlands are good and know about improving and playing well :). Do you remember the article about "effortful study" ? I think that's the key.

    What I think effort is in CTS for example ? it is very different if you pasively wait for a pattern to pop up and don't do "saccades", as blue devil mentioned, but if you rush to find something, very fast, with dozens of saccades, looking where your pieces point to, the king safety, all the checks captures and threads, etc, etc... I mean, you really do an effort to find the move quickly and feel like your brain hurts, I think you will be improving. As in other sport, your brain must transpire. You will feel the effort feeling in your brain if you do it, and you won't feel nothing if you do it in a lazy way.

    That's my idea. Brain effort.

    here is the article.

    http://www.sciam.com/article.cfm?chanID=sa006&colID=1&articleID=00010347-101C-14C1-8F9E83414B7F4945

    ReplyDelete
  10. I agree with Shakmat. You will improve only with effort. That's why I have abandoned the repetition of known (done) problems at CTS. Just because the CTS problems are too easy once you know them. No effort, therefore, no training.

    ReplyDelete
  11. I think it doesn't matter if you are solving a mate in one or a mate in 9. In both cases if you really effort to do it as fast as you can, or faster then the last time, and as the article says, you are enjoying that, then you will improve. Regarding the repetitions think it is the same, if you effort and blablabla the same as mates examples.
    I read once that the chess training, like other sports, must be with so much effort that when you play OTB it has to be like a pleasant walk, and not a big effort all the time. That's training for the competition I think. You trained that much that all will be easy, except for some little things. But those little things could give you the win.

    ReplyDelete
  12. Of course it takes hard work to improve, that is no mystery to Temposchlucker. The name of the blog is chess improvement by effort. But it's not enough to work hard, we must also work smart. A basketball player who shoots a thousand half court shots a day is working hard, but not getting any better.

    If I understand Tempo correctly, his question is how to work smart? I think Blue Devil has said some very insightful things regarding how we see the board. My personal belief is that working smart means working on how we see the board. That is, when we look at the board, what do we see?

    A beginner doesn't see anything but individual pieces on individual squares. It is nothing special to the beginner that a piece sits completely undefended and subject to attack. An advanced player knows that this piece is in danger and can be taken advantage of. More importantly, the advanced player knows this before thinking about it consciously. Similarly, the advanced player recognizes checks, captures, forks, pins, skewers, etc. before thinking about them consciously.

    How does the advanced player do that? I think Blue Devil has hit the nail right on the head. The advanced player has trained his eye to scan the board for important features. This process has become a subconcious one and very efficient. My hypothesis is that we must realize what tactical features we don't recognize in this way and work on recognizing them until the brain decides they are important enough for efficient recognition and incorporates them into subconcious sight.

    ReplyDelete
  13. I have tried efforts at different degrees of intensity. Until I squinted. Of course. But I didn't get significant improvements in comparison to a more relaxed approach. A brain is clearly no muscle. The term effort is somewhat vague, since it seems to be only measurable by the amount of pain it causes or how tired you feel afterwards. Research showed a significant improvement when FOCUSSED though. Maybe focus and attention are more clear words than effort when it comes talk about intellectual improvement.

    ReplyDelete
  14. --with the least intent to grandstand' publicly in what is really another group of persons public conversation—- bur still pose the question:

    Not to be too tongue and check, but,

    What about higher percentage success instead of, or along with:

    faster problems (aka 'Rating'),

    or learning more problems ('memory', long or short or whatever it is)?

    That is to say:
    or more accuracy (percentage).
    dk

    ReplyDelete