Friday, February 25, 2011

I was 50% right

On the post Dissection of a microdrill Bright Knight commented:

I believe that conscious feedback is important. If you get the solution to a problem wrong, it helps to work out why you got it wrong, and what you could have done to get it right. (Being too slow can be “getting it wrong” in this context.) Finding simple chess tactics is mostly an automatic unconscious process (unless you are a beginner). If you get it right very quickly, so quickly that you did not have time to work it out, that proves that you are doing it on autopilot. If you do not solve a simple problem quickly, you have not demonstrated even temporary success at the skill. You need to find the solution quickly, and have as much time as you need to reflect if you do not. If you fail to get the solution, a fast time limit is bad, because finding the solution for yourself is much more instructive than being told the answer - but spending too long on a problem is bad time management. The number of repetitions is critical. If you do not constantly repeat a complex skill it will disappear. If you want to retain the skill to spot a tactical pattern, you have to practice that skill for the rest of your chess playing life (ideally every time you are just about to lose it). (Being too slow constitutes “losing the skill” here.) If you do not repeat a skill, you do not know whether you still have it. The spacing between repetitions is critical. If you space them too widely you will remember very little, and if you space them too closely you will forget most of what you have learned very quickly.

I can't stress enough that we are talking about two quite different processes here:
  • Acquisition of knowledge.
  • Acquisition of skill.
The methods that apply for one process do not apply for the other. Spaced repetition belongs to knowledge acquisition. It isn't suitable for skill acquisition.
Knowledge tend to fade away overtime. So it has to be repeated every now and then. Once acquired, skill doesn't dissappear overtime. Exactly as you never forget how to ride a bike or how to swim.

I told you that there are two critical factors for scandrills:
  • Conscious feedback.
  • Speed.
After doing the board vision drills of Fritz for over more than a week I now know that conscious feedback belongs to the area of knowledge acquisition and not to the area of skill acquisition. On the other hand it has become perfectly clear that there is a second critical factor for skill indeed: simplicity. If the task isn't simple, it can't be transferred to skill. So:
  • Speed
  • Simplicity
Besides the vision drills of Fritz I have looked carefully into the article of NM Dan Heisman about the seeds of tactical destruction. The next skills I want to develop a training for must have a relation to the seeds he mentions.


  1. Good post and quite interesting. I agree that the extending time spacing probably works better for knowledge than skill.

    I disagree that conscious feedback isn't related to skill development though. There's been some great research on "deliberate practice", where there's a big emphasis on a conscious feedback loop, that shows how effective it is at skill acquisition in a large number of fields. I came across the model when learning the cello, but also use it for tactics as well.

    BrightKnight's approach is also what I try to do: Revising missed tactics to a) see what the answer was, and b) what was the missed pattern or hint that points to the right answer. When I've spent some time to explicitly reinforce the input pattern and the output pattern, I feel that I've been more likely to see the tactic more quickly and accurately next time.

  2. And on re-reading, I see that your example was board vision, which might not benefit as much from conscious reflection compared to pure repetition.

    I think this is probably because there's not too many ways to get it wrong while getting more practiced at it. For more nuanced skills like playing an instrument in tune, athletic skills, or finding tactics, it's more important to practice it right, and that requires conscious feedback.

    There's a saying that practice makes perfect, but it's not true. Practice makes permanent, perfect practice makes perfect.

  3. You've spent alot of time solving chess problems. So is the ability to find the first move in a problem (start of a plan) skill based or knowledge based? I'm not sure I buy into the difference is as extreme between knowledge and skill as what you say. In fact, I would argue than when you talk skill vs knowledge you are talking about things you do at a near unconcious level vs those things that you do very conciously. So I would say that some things for you which is concious Knowledge ( difficult 5 move chess problems) is for Kasparov skill (uncon.). That is is possible to chunk larger series of tasks as if it is a one step task.

  4. Have you been reading/read about the field of "Deliberate Practice' and the work on Anders ?

    I think you would find it interesting and may have something to say about what you are doing.

    Good Luck with the skill building.

  5. It should read Anders Ericsson.

    (Your word verification word for me was chess 8))

  6. Consciousness only is aware of what happens in the short term memory. That is an area with a lot of restrictions:

    *Limited placeholders for information (7 - 12)
    *Fades away within seconds.
    *Information is processed in sequential order.
    *Slow (>3sec)

    If you speed matters up, you simply cannot use STM anymore. Hence you must go beyond conscious feedback. Exactly as a tennisplayer has to hit a ball that moves too fast to see at the moment of impact.

    Playing Troyis is not about consciousness but about autopilot. The same is true about board vision exercises. The conscious feedback is more circumstantial. The task itself is too fast to be aware of, but you can see the result. Since it are typically simple tasks that lend themselves to be automated, the resulting feedback is often simply "right or wrong".


    Page 40-

  8. Sorry my last link was wrong, here the right one:

    Page 40-

  9. @Uwe,
    good to see that the first article quite backs my empiric findings with scientific evidence albeit they have more problems to formulate it clearly. But that has probably to do with that they don't use themselves as guinee pig so they haven't the slightest idea what they are actually talking about;)

    I will dig in the second article.

    @Tak, I probably will adress your questions in a new post. I haven't read Ericson's work. Jet.

  10. The second article seems to be in sync with my findings too pretty well. They found already one critical factor of the two: simplicity. They call it low level tasks. They seem to miss the second critical factor though, the necessity of speed;)

  11. The main area of practical application of the expanding repetition (aka Spaced Repetition) learning systems has been learning foreign languages. Speaking a foreign language fluently (let alone thinking in it) is a skill, albeit with a large memory component. You simply do not have time to recall the foreign word equivalents one by one, apply the grammar rules and translate. It has to be an automatic process. You have to have a program in your head that speaks for you.

  12. The text gives some hints wich skills might be usefull for better OTB-Chess, for example:

    Ericsson and Oliver found that the chess master could very rapidly extract the number of pieces of a given color that attacked a randomly selected square in a memorized position.

    So the the board vision drills of Fritz might be realy helpful!

  13. @Bright,
    that are two different processes. Learning the vocabulary and speaking the language. One is acquiring knowledge, the other a skill. Spaced repetition is applicable for learning the vocabulary.

  14. The proof of the pudding is in the eating here. I suggest that you find a set of problems that you have never solved before, but are at the right level for you (a couple of dozen will do). Solve them three times on the same day, timing your solution times for each problem on each pass. Your solution times should improve. Come back in a month and do another timed pass. If you are slower, there is something wrong with your theory!