Wednesday, June 07, 2006

Back on track, the experiment starts.

After 3,500 problems and 7 days I have an average rating of 1530 at CTS.
That´s exactly what I had when I left CTS a half year ago.
So efforts aren´t down the drain after all.

Now comes the interesting part at CTS.
I have done about 44,000 problems at CTS.
This means that I have repeated the problem window of 10,000 problems more than 4 times.
Usually after 4 circles is where calculating ends and pattern recognition starts.
And indeed I´m starting to recognize a lot of problems.
So let´s see where the 7 circles of madness leads me.

For the improvement from 1470 to 1530 I had to learn 2100 new patterns allready.
That took me 7 months and 44,000 problems.
So to master 1 new pattern you have to solve 21 problems at CTS.
This doesn´t look very effective.
The main reason for this is that the problemset is too big.
So it takes about 1.5 to 2 months to repeat the whole set of 10,000.
In the mean time my OTB rating improved from 1710 to 1750.

To improve further from 1530 to 1600 average rating at CTS I have calculated that I have to learn another 2300 new patterns.


  1. Wow this means taking a job serious. Good luck! CTS patterns are not from games, but from a random generator. They may occur in games, though, I remember one or two parallels in my own games and at CTS. But I also have the impression that CTS patterns are biased in some way. Often repeated motifs in different puzzles. Maybe this is one of the reasons I stopped using CTS. The other is that I have the impression that speed is not my problem, I rather try to avoid time trouble, and I do not intend to become a blitz champion.

  2. Mouse, everybody with 100 points more OTB is better than me in blitz too.
    So developing blitz-like skills seems to be unavoidable.

  3. Many people have read my posts on CTS. My handle on CTS is spacecowboy.

    Tempo, regarding your two statements:

    "For the improvement from 1470 to 1530 I had to learn 2100 new patterns allready."

    -- and --

    "To improve further from 1530 to 1600 average rating at CTS I have calculated that I have to learn another 2300 new patterns."

    Apparently you are privy to some arcane information that you are using to make your calculations. Perhaps you would like to teach me how to do this if it isn't too complicated to explain.

    It is apparent that what you really mean when you say you have to "... learn ..." xx number of new patterns is you have to achieve memory consolidation of xx number of new patterns.


  4. Space,
    To calculate the amount of problems I did the following:
    1 problem at CTS = 1 pattern

    At CTS:
    Problems/Rankings look at the table with high rating.
    Press "next 20"
    In de adressbar appears the offset, which you can manipulate. So you can find the offset of a 1530 and a 1470 problem. Subtract the offsets of both problems from each other and you know the amount of problems between them.
    Another method is to estimate it from the bell curve.

  5. Tempo,
    How can we be sure that y is a function of x? For readers who are not up to speed in calculus and serious mathmatical issues, I will give an example to illustrate my point.


    Using the bell shaped curve and the 'function' that temposchlucker has provided us with, I estimate that only about 240 new patterns are required to go from 2000 to 2100.

    My my Tempo! Isn't it strange to compare this small number, 240, with your previous statement that you had to learn 2100 new patterns for the improvement from 1470 to 1530?

  6. Space,
    it is not as strange as it looks. We are talking about CTS-rating points, which become less reliable at the edges of the bell curve. Lately Uraleech demonstrated exactly that.
    At the edges the relation between effort and CTS-rating improvement is far from linear.

  7. Tempo,
    Good point. Yes, it is important to consider that CTS ratings are unreliable near the edges of the bell curve. And with this in mind, I agree that it isn't so strange that your function performs the way it does. I mean your function shows us a disparity between the number of patterns required to to go from 2000 to 2100 compared to the number of patterns required to go from 1470 to 1530.

    In the above sense, it isn't at strange as it looks. But in the sense that I am going to describe below in the next paragraph, it is just as strange as it looks at the edges. By this, I mean it is just as unreliable as it looks at the edges. I don't cite your function as the source of the unreliable results at the edges. But I would go so far as to say your function is less meaningful when it is applied at the edges.

    uraleech demonstrated that ratings above (pick a number?) perhaps 2000 on CTS are unreliable. Furthermore, the meaning of the difference between a tactician rated 2000 on CTS and 2500 on CTS is not well defined. And similar to the statement you made, we can also say that at the edges, the relationship between tactical ability and CTS-rating is far from linear.