Wednesday, August 15, 2007

How does a grandmaster finds his way through the thickets?

To bell the cat

The difference between a good player and a bad player is the ability to calculate well in complex positions. Difference in knowledge plays no serious role since you can obtain any knowledge within weeks or even days. But the difference in calculation ability can't be overcome in a short time.

Yesterday I asked myself "How does a grandmaster finds his way through the thickets of variations in a complex position?" What I tend to do myself is clear: I make a list of nice looking candidate moves and by means of trial and error I try to reach a definite conclusion. Say, I see 5 reasonable moves. And my opponent has 5 reasonable answers for every move. Within 3 ply I have to investigate 125 reasonable possibilities. Even a grandmaster wouldn't be able to work this way.

And thus he must approach it differently. This is my take. He recognizes the first tactical motif. That limits his own possibility to one move. There are only very few possibilities to answer a tactical motif. Say two moves are related to the tactical motif. The rest can be neglected since that are random moves, which don't give an answer to the actual threat. That way you can hop thru a list of 9 tactical motifs (based on yesterdays problem) with usually 1-2 choices for yourself and 1-3 choices for your opponent.

So that eases the task considerably since all random moves are out of the equation. I know, I know, this is a very ridgid way of reasoning. But with a little goodwill you understand what I mean.
You have to know your tactical motifs extremely well, so that you can be confident that you have treated all relevant moves.

Writing down narratives helps me to see a complicated position in a much more simple daylight. But will it be enough?


  1. Even Kotov says to only think through variation trees for sharp positions/moves. Quiet positions/moves let you use more general principles. I discussed this in a post here , which was one of the most useful, IN PRACTICE, ideas I had encountered.

  2. i continue to check in several times a week, but hard to comment meaningfully AT THIS TIME, but i cheer you on in your engaging process, which you share, and so well. thank you. david

  3. David,
    thank you for cheering. It is appreciated and needed, since the thoughts don't come easy.

  4. Blue,
    that is true. Your question some time ago about what I was actually doing when the position was in quiesence made me realize that I was wasting my time with trial and error, the amateurs form of "analysis". That insight has changed my chess dramatically.

    But here you sacrifice matter, which you can only get back after 20 ply of variations. At that moment you want to add some real calculations to your prayers.

    The fact is that I can find any tactical motive with trial and error. But for indepth calculation you need to be more familiar with the motive than that. Just "exposing" the brain to a truckload of basic tactical motives is not enough, as I have proven with my 70,000 problems solved at CTS.

    So the experiment is now: "will the description of the higher structures behind the motives by creating narratives be sufficient to familiarize the pattern so that you can ban trial and error?"

    Without that familiairity, you don't have a choice how deep to calculate. It is just limited by your short term memory and the available time.

  5. That picture...that cat looks so pathetic. Poor thing.

    I have trouble, even when I make myself see the solution after one mere minute of staring, coming to understand many positions even in CTB after 5 minutes of working over the solution (and this is for tactics in which the variation trees are no more than 8 moves). Not only seeing the answer, but really trying to understand it, why this rather than that, why this defense doesn't work. It is exhausting. After five problems, just a half hour, I am already tired (much like if I gave a half-hour think in a real game, at which point I'm ready to get up and walk around a bit).

    So far it feels like a better approach, more fruitful, more likely to teach me some general principles (so far I am getting a better sense for open lines near the enemy king, and learning to really check out how much freedom he has to move around if I start an attack, versus how much freedom my pieces have to sweep in and destroy him).

    I can't imagine what it is like for the Polgar puzzles!!! I could easily see taking many hours. Even some of these CTB K/P problems I could spend two hours on if I really wanted to be thorough!!!

    Dude, do the circles with Polgar middlegame treatise. It will only take seven years!

  6. Blue,
    it is true that a typical problem from Polgars book take me a few days. And that it is very demanding and exhaustive. But I have the feeling that this "powertraining" hasn't to be done in more circles than one.

    The 28 problems I'm working thru now, I have done before about half a year ago. I have done them thoroughly by then, often I used one or more days to work out all the variations.

    Now I do them again. Actually I have forgotten almost everything. That shows that working on the automatic pilot, even if you use days for one single problem, doesn't pay off. It is a waste of time.

    In this experiment I have repeated 4 problems of those 28 old problems. But now I insist to explain the lines to myself thoroughly in words. That has a strange effect. First is simplifies the position, as if it is not so difficult after all. Secondly, I can't imagine that I will need to repeat it, since I'm sure I will recognize the essence even after a decade.

    I truly think that we have wrong associations by pattern recognition. We only have to conscious assimilate the cognitive-super-rabbit once. Then we can recognize rabbits everywhere, in clouds, in trees, in mountains and in your coffeecup.

    I'm convinced there is no need to do all 4158 calculation exercises once you are trained in the use of the tables of multiplication.

  7. What's up with the cat? Looks like his favorite opening has just been refuted.

    My take on circles and analysis is that the benefit of the circles is pattern recognition to learn to see and know relatively simple tactics. BAM! Mate in two or five.

    Working through harder problems and doing the analysis is preparation for game time and practicing the full analysis. During that process, one will be aided greatly by the store house of simpler tactics hardwired into your brain (circles).

  8. Glenn,
    that was my idea too in the past. But since I have done 7 x 10,000 exercises at CTS without any effect, I consider that idea as refuted.

  9. Tempo: I don't think it "refutes" the circles. For one, how often do you overlook simple tactics? Probably not often. If you are still consistently dropping pieces to 1-3 move tactics, then the circles have failed for you. I know such simple tactics probably don't come up directly very much at your level, so you probably don't get a chance to see this in action, but they probably guide your play.

    Also your methods were so unique that is hard to generalize to people doing more traditional circles. Did you ever learn a problemset down cold: I know you said you did seven circles with various programs, but was it to the point where you were doing all the problems 'by hand', no thinking, quick-as-recognizing-a-friend's face?

    It's a given that it was not like that with CTS. Impossible. What about CT-Art etc? That's really the point of the circles. Not just doing a bunch of puzzles. That's what everyone has always done since Reinfeld's books and the Russian chess gulag. But doing the same set until you bang 'em like a bonobo.

    And if you did, I wonder why it helped MDLM so much and not you? He added a thought process after he finished: he said he never saw results otherwise from the circles. It seems much of what you have been talking about lately is the development of a more sophisticated way of thinking about positions.

  10. It will be very interesting to see if all this self-explanation pays off! I guess I'll know sooner than you, as I'll be starting Circle 5.2 in about a week.