Monday, September 19, 2005

The road is harsh buth clear.

20,000 problems done at CTS.
That gives a satisfied feeling, I must say!

My rating is now averaging at 1520.
I'm quite confident that my rating will continue to grow in a somewhat linear way until. . .
Well maybe untill I manage to disturb those two Swiss guys at the top?
The reason is very simple. I'm starting to memorize problems and the amount of problems is finite.
To complicate matters, CTS has released about 2000 extra problems the past few weeks and they will release another 3000 coming weeks.
But that will not make an essential difference.

The main question is: will it transfer to a rating gain OTB?
To a certain degree it will.
I'll try to explain.
Last friday I watched two guys playing chess with each other.
One of them has a rating of 1320, the other has about 1520. Resp. 400 and 200 points below my rating.
I watched them very carefully and every now and then they made a move which was very good and which I hadn't foreseen. I realized that they both had stored different patterns in their mind than me. That's the reason why I always have to play careful when I play against them. Allthough I win almost everytime, they often succeed to surprise me.

So everybody has his own database of stored patterns and there can be a vast difference in quality and quantity. The way your database is adapted to your usual opponents determines your rating.
I often have difficulty to "come into my own play". But when I manage to do so, it is as if my play is skyrocketed. It feels like slamming your opponent KO. This summer it happened in 4 of the 18 games. One of my opponents said that he wasn't thrown off the board so brutal in decades.
This means that my database is somewhat unusual. Which is not so strange, given my training.
In only ca. 20% of the cases my database pays off fully.

So the main question translates to: "how good is the CTS problemset?"
CTS hasn't given a straight answer yet to how they generate their problemset.
Thus we have to try to do some educated guesswork.
I assume the following (with -of course- reservations):
They use a database of games. Whether this is a database of human players or computers, I don't know.
Then they use the chess engine BetaX to filter positions automatically from this database.
Of course they use certain criteria for filtering.
Which ones, I don't know.

The positions I encounter at CTS look plausible.
That's to say, there is a possibility that you see these positions in a real game.
The fact that Frizz, a Swiss IM of about 2500, scores high when starting with CTS, proves that he has these positions allready stored in his database for a great part.
So you have to learn them anyway.
But the observation at my club shows that there must be a lot of patterns OUTSIDE the CTS problemset that has to be learned too.
For example endgames are not dealt with. But also some types of tactical shots I have seen on George Renko's CD's I haven't seen at CTS.
So for the moment the path is clear. Learn all problems of CTS a tempo.
And then? Well, maybe we should have a talk with those guys of CTS to adjust their criteria.
Or maybe we have to program something ourselves. There are enough free and excellent chess engines out there to make use of.

circle 0: 1470
circle 1: 1500
circle 2: 1520
circle 3: 20,000 /70,000
Highest rating 1567


  1. Are CTS problems only one move problems? Does it ask for answers that are a first move in a combination? I'd find out myself but don't allow myself being distracted until I finish my CT-art

  2. Tempo,

    Many of the CTS problems are straight out of other published works. I know this because I recognize many of them.

    And to be honest, many of their problems are not tactics at all - at least not in the usual sense of the term.

    Further, some of the tactical problems are just plain nutty. I've seen some problems that leave you in a weak position no matter what you do and you should clearly resign rather than play on.

  3. Tak,
    Most problems are 2 to 6 plies deep.
    Often only the first move is asked.
    Complexity: most players can do 60 - 100 problems per hour.
    Given enough time, a player of rating 1300 (OTB) can do most problems.
    The real problem lies in the fact that you have to answer within 3 seconds, otherwise you are punished. Ratingwise that is.

    You have to alter your critical attitude. Otherwise you will get problems in a world that is dominated by Microsoft products. Beware of your blood pressure!
    It is sad and incomprehensible that not everybody is like us. But you have to learn to ignore the shortcomings of the rest of the world.
    If I have 3 seconds to win a piece, I can't break my head about the wreckage of my pawnstructure.
    In a real game I will never trade a queen for two rooks. Simple because I'm often in time trouble, and it is far more difficult to handle two rooks properly than a single queen.
    But I have learned to ignore the fact that I have to solve such problems every now and then. I don't even notice it anymore.

  4. Memorizing positions: I think this is the wrong way, the issue is recognizing patterns in unknown positions.

    CTS generator: All problems are generated by an engine, at least this is what they say. By chance it may happen that a position of a famous game arises.

    Frizz: He has been quite a heavy user on CTS some days ago, and then he smashed a 2500 rated GM in 23 moves! So it seems that even IMs can take profit from CTS training.

    2 rooks: they equal 10 pawn units vs. 9.75 of a queen. So in most cases it is ok to give a Q for 2R.