Monday, September 18, 2006

Going around in circles

It is shown that there is a (negative) statistical correlation between the #times I have repeated a problem and the solution speed. Which means that at average I solve a problem faster when I have seen the problem more times before. This was based on a wee problemset of 56 problems that I solved correct. I plan to investigate a bigger problemset in the future.

So massive repetition works. It doesn't work very efficient though.
The question arises, what is it what works in massive repetition of problems?

We can define a broad spectrum.
On the one end of the spectrum, there is a big database of 50,000 - 100,000 basic patterns you have to master. If you are confronted with a chess position you have to recognize one of these basic patterns. Since the amount of basic patterns is so high, you can say that the "pattern recognition factor" (an own invention) is low. Science did an educated guess that learning much basic patterns is a possible scenario how a grandmaster became good (based on the theory of chunks). There are two arguments against this.

The first is that a pattern recognition factor is usually high in any given situation. If you see a cloud, you recognize a rabbit in it, not a Sylvilagus mansuetus. You only have a database of about a hundred pictures of animals, but you can recognize them in even the most distorted clouds. You don't need a database of 50,000 - 100,000 animals.
The second argument is that it is not according to my own experience. The past few weeks I concentrated solely on memorization of old problems at CTS. My rating declined in stead of improving.

At the other end of the spectrum you break down the basic database until only a few characteristic geometrical patterns are left. The Mother Of All Patterns (MOAP). Here the pattern recognition factor is very high.

You can imagine a database of anywhere from about 10 to 100,000 basic patterns in the spectrum. What is the right amount to train?
MDLM gave a clue about a database of a few amount of basic patterns.
To train them he invented the microdrills.

I investigated the microdrills two times in the past. Both times I dismissed them as being of no use for a person of my level. As too simple. So I have never trained them serious on a regular basis.
But now I see reasons why it could work. Hence today I started my concentric circles and knight sight exercises.
I'm not good in it.
Which means that I have to THINK to get the job done.
And thinking is SLOW.
If you realize that about 30% of the problems at CTS are DOUBLE ATTACKS, and that the microdrills are designed for DOUBLE ATTACKS, and that I do the microdrills SLOW, there is reason enough to do this experiment. Better late than never, as an old Chinese saying goes.

Doing massive repetitions of problems clearly didn't lead to a faster doing of microdrills. Now let's see of doing microdrills leads to faster solving of double attack based problems.

Have I finally arrived at the end of the rainbow, is it the wrong end! %&*#!


  1. "Have I finally arrived at the end of the rainbow, is it the wrong end! %&*#!"


    Tempo your work ethic has been an inspiration, so there is no way anyone can say that this attempt failed due to lack of effort. Sometimes finding out what doesn't work is just as valuable as finding out what does.

    A long time ago a man told me something that has stuck with me
    ever since. He said "Focus on getting it right first and the speed will come to you."

  2. Concentric circles? Knight sight exercises? You do these again, just in the way MDLM has described them?

  3. Mouse,
    yes, but not again, since I never did them.

  4. I did, but I think it was a waste of time. It did not help me to improve board vision in real positions. I think CTS is a better tool for this purpose. Much better. But of course you have to take time there, until the squares «pop out».

  5. There's no need to choose, I can do both CTS and the microdrills. The truth lies probably somewhere in the middle of the spectrum, but I have to check for myself if the microdrills can be ruled out. As Sancho indicated, I work by the method of exclusion. SOMETHING is working when exercising at CTS. What is it? So let's pop!

  6. hey, who can we blame for popularizing odd or idiosyncratic pictures on blogger in chess training and improvment? flog them!

    ...see you at the end of the rainbow tempo, nice post... 20,000 tomorrow for sure, only five days off plan set in ealry July.


  7. perhaps too large a database doesn't allow a tactic to be repeated in a short interval. ie a tactic repeated 10 days in a row causes a stronger recognition than one done 1 X a week 10 weeks in a row. You might want to consider forming a data base from the 5 X5 problems in Ct-art. This would be similar to the patterns in Bloch book there were about 400 of them . (this can be only found in his earlier book the latest doesn't have the simplified patterns). the 5 X5 problem set are the problem within the problem.
    good luck with the upcoming chess season.

  8. I do micro-drills whenever I overlook a knight tactic. The last time was 40 games ago.

  9. Tak,
    the past 3 weeks I concentrated on just that, repeating small sets of old problems according to the system of spaced repetition. But it didn't work out as expected. That's why I'm looking for new ways.

  10. I think that since the Knight sight drills do not come naturally for you might be a sign there is great benefit in doing them. Perhaps this is the catalyst that will activate the CTS circles you did recently. MDLM was a great believer in this and did them before each tournament. Look forward to hearing about your results. Reading Likeforest comment about the knight sight drills reminds me that I have been missing a few Knight forks myself and could benefit from revisiting them...