Friday, August 18, 2006

Chug chug

I'm from an old school that doesn't allow me to have a big mouth when I have no results. I know, in these modern times that is rather old fashioned, but you can't learn an old dog new tricks. So that is why I blog so little lately.

But we can always do some crude calculations.
The table below is derived from a bellcurve at CTS giving the rating distribution of the problemset.
The first column is your rating.
The second column tells you how much problems you have to commit to your long term memory to gain another 50 ratingpoints.
The third column indicates the amount of problems you have to commit for 1 ratingpoint increase at CTS.


Rating

#probs to commit
#probs per rtg.point
1600170034
1650
150030
1700
150030
1750
120024
1800
100020
1850
800
16
1900
700
14
1950
500
10
2000
500
10
2050
300
6
2100
300
6
rest
200
4
total
10,200



As you see the #problems per rtg.point diminishes. That means that the effort to grow diminishes. On the other hand the reliability of the rating will diminish at the edge of the bell curve. Measurements in the past have lead to the hypothetical conclusion that there are 30-33 problems needed per rating point OTB.

Boy, I wished I had a reason for a big mouth.

8 comments:

  1. i think that you meant that the effort to grow increases. i like your thoughts though.

    ReplyDelete
  2. General, no i said what I meant to say. In terms of amount of problems you have to commit to LTM, the effort decreases per ratingpoint extra. Of course the problems become more difficult, but since you are always presented with problems you can solve in 10 seconds AT AVERAGE, you will not notice that (maybe only the first time you encounter new problems, but the mind tends to adapt) Overall I experience the same difficulty of problems as a year ago, but now the problems are rated 120 points higher at average.

    ReplyDelete
  3. Rating is one thing, and success rate another. The success graph at http://chess.emrald.net/ctsTactHome.php has a peak at 75%. Let us assume that most tacticians try to optimize their rating. It seems that two or three losses in every ten problems seem to be best to achieve this. If you try to get 90% correct, your rating may be lower than optimal. For me, at the moment, it seems to be about 50 points lower than before, when I used a 80% strategy.

    ReplyDelete
  4. Mouse,
    I use a somewhat different strategy. When I have committed all 23,803 problems to memory, my succesrate will become near 100% anyway, so why worry now? Ok, it might be optimistic, but that's my character:)

    ReplyDelete
  5. Tempo,

    When you say 30-33 problems in LTM per OTB rating point, do you suppose that this is linear? I tend to believe that it is, without proof, just intuition.

    Clearly if OTB rating is linear with LTM storage but CTS rating is not, then CTS rating can not be linear with OTB rating. Since it takes fewer problems per rating point at CTS and the same per rating point OTB, CTS rating should grow faster than OTB rating. But I think this is false since highly titled players are only rated 2100 on CTS.

    I don't really know what any of that means. The only thing I am mildly sure of is that I am getting better at CTS and I have the impression it will help my overall game.

    ReplyDelete
  6. Loomis,
    that is an interesting question.
    The INTAKE of problems is near linear. I don't think there is a physiological reason why this couldn't be the case untill at least you are 70 years old. (Look at Korchnoi.) If the motivation isn't falling. But what result will those internalized patterns yield as OUTPUT?

    There are a lot of factors at play. For instance the quality of the patterns. And the frequency the patterns appear in OTB play.
    There are tactical patterns, positional patterns, endgame patterns. Even opening patterns.
    First you will develop with one-movers, then you need two-movers, than three-movers.
    The total amount of all possible three-movers exceeds of course the amount of all possible two-movers. Is it necessary to know all three-movers? Probably not. You have to make a selection. How do you choose?
    Problably there are three-movers you don't need to now and four-, five- and six-movers you DO need to know.

    The matter is: can you find the reason you lose a game? Can you define groups of games that you lose for the same kind of reason?
    Can you you find a database with exercises that adresses exactly the kind of patterns you need?

    Since we are equally bad in all aspects of the game, almost every problemset will do for now:) But there will come a moment that it will become problematic to find the right problemset.

    So the output will not be linear. On the other hand, if you look at the ratinggraph of fastgrowing youngsters, you will often find a rather sudden plateauing. Which looks unnatural. I tend to think that at about the age of 18, the "fluidity" of the brain comes to a hold, and they have to switch over from effortless automatic learning to a conscious effort of pattern intake. Since almost nobody seems know this, the development grinds to a hold.

    About the comparability of CTS and OTB rating: you have to make a distinction between your INITIAL rating at CTS and what happens if you are increasing at CTS. Initially persons start at linear distances from each other, based on their OTB rating, but than CTS rating grows apart from OTB rating.

    ReplyDelete
  7. For example (see table):
    If you start at 1600 CTS and you absorb 10,000 problems, your rating will be 2100 CTS (gain 500). For OTB you should expect a gain of 10,000/33= ca. 300 points.

    Mind you, I'm not talking about #problems solved, but #problems committed to LTM.

    ReplyDelete
  8. i must keep my mouth shut also. i am a small chess person now.

    but really, really appreciative of tempos post, and everyones astute comments. what a great group of fine minds!

    i am back to CTS after a week or more off. my head no longer wobbles from its injury. warmest regards, dk

    ReplyDelete