Saturday, February 14, 2009

Picking up the gauntlet

According to the chess exam I'm a fairly balanced chessplayer. I only suck at calculation. I suspected that a bit, but now I know for sure. All my tactical efforts and all the methods I have used didn't adress that problem. I am not the only one. Blunderprone recognized the same problem after doing the circles of madness. That indicates an essential flaw in our methods.

The strategy I followed the last year was to avoid complex positions with heavy-duty-calculations. For instance I throwed all my gambits out of the window. I was pretty succesful with that strategy and it gained me more than 100 points.
Which is in a certain way an indication of what the influence of bad calculation is. I'm sure that I can make more progress on that path (avoiding complexity), maybe another 100 points or so. But in the end I will inevitably hit the wall: bad calculation. This means that if I want to make some serious progress I must fix this problem.

I have no idea what calculation actually is. What the exact problem with calculation is. Otherwise I wouldn't have deceived myself by thinking that I made progress with calculation. I made certainly progress while doing tactical problems. But I didn't made progress with calculation. So that is the first task, to identify what calculation is about and what it's place is in the total area of chess thinking.

There are two main areas of chess thinking.
The knowledge/insight/conscious/declarative/slow/sequentieel area and the skill/unconscious/procedural/automatic/fast/parallel part. To simplify the text I will identify these areas with the terms knowledge and skill.

If I compare that with a clarinetplayer, a clarinetplayer needs skill to operate his instrument. But skill alone is not enough. Besides skill he needs the knowledge about what he should play. Without that knowledge he simply continues to play his toneladders and arpeggios without a purpose. Knowledge provides the insight in what direction to go. What to play and how. The instrument is played automatically by skill.

What is the relative proportion of knowledge and skill in chess?
With knowledge but without skill you know what to do but you are not able to do it without losing material on the way or other unexpected results.
With skill but without knowledge you can schwindle your way out of difficult situations, but you make serious positional blunders when there is not much to do.
The best games by the best CC players that aren't OTB players are a few hundred points lower rated than the best games by super grandmasters who play OTB.
It is reasonable to assume that CC players are able to equalize the skill of a grandmaster by using 10 days per move in stead of 3 minutes per move. Which means that it is the knowledge of the super grandmaster that makes his games of a higher quality. He evaluates the positions better. Because he has more adequate knowledge.

Calculation operates on the cutting edge of both knowledge and skill.
In itself calculation is a conscious process, but in order to take place it is assisted by a whole bunch of skills which are performed unconscious. Every conscious process has a serial flow and makes heavily use of the Short Term Memory (STM). The STM has about 5-9 placeholders for information and information in it decays within a minute or less, so it has to be repeated constantly to prevent it from fading away.

A serial process seems hardly suited to process Kotov's tree of analysis. A tree can have lots of branches which sprout from various junctions. This means that you have to simulate parallel processing in a serial way.

There are a lot of tasks to perform. I identified the following tasks:
  • generating sensible candidate moves
  • visualisation of the moves
  • keeping stock of which branches you have processed
  • evaluation of the position at the end of a branch
  • keeping stock of those evaluations
  • keeping stock of the pieces that are traded in a line
  • try to make incomplete lines work
  • get new ideas
  • work backwards from ideas to make them work
  • keep stock of two simultaneous lines if there is both an attack and a counterattack going on
It is a miracle that we are uberhaupt able to perform conscious calculations with only 5-9 placeholders in STM, and with information decaying within seconds. No wonder that I'm continuous repeating the same moves over and over again in my mind!

What are the possibilities to improve calculation or to lessen the effect of bad calculation?:
  • Avoidance of complex calculations.
  • Improve the performance of my STM.
  • Transfer of tasks from conscious processing to unconscious processing. Which can happen parallel and doesn't make use of STM.
At the moment I'm working on the first option. Throwing calculation intensive openings out of the window worked fine for me lately. But I expect that the progress I can make this way is rather finite.
About the second option: I can't imagine that that is possible.
That leaves the third option. Given the fact that I already tried a whole load of methods to little avail the question remains: how?


  1. Visualisation might be a good point to start with. All good Player are good in blindfold playing to. Lazlo Polgars 3 daughters became WGM and one point of their education was blindfold chess. How good are you in blindfold chess? Reading Chessbooks without board can help with the improvement of visualisation, "CVT Online" offers some excercises

  2. Uwe,
    the effect of visualization tends to be overrated. The problems at which I fail are of such difficulty that I fail them even with my eyes open and making use of an analysis board. This I tend to call perfect visualization. Of course, if you have less than perfect visualization you will score even worse. But even perfect visualiszation isn't enough by far for these problems.

    I did blindfold chess, I did huge amounts of CVT and other exercises and I am not too bad in it, but even then, it doesn't fix the problem. Not even close.

    The problem lies in overloading the STM, which paralyzes thinking.

  3. What about forgetting calculation for a moment and first exploiting the rating space you might have to gain by capitalizing on your positional potential? Who knows how long it takes till you hit the calculation wall. Why tear down a wall thats not busting your head? Why not focus on the max rating gain per time unit spent?

  4. Papablanca,
    those are wise words. But my character doesn't allow me to be wise. I'm just too curious.

    Ofcourse I will try to continue with positional training (I'm solving the positional exercises of Aagaards excelling at positional chess). OTOH the Chess Exam has made quite clear where my real problem lies. At the moment there is no reliable method to train calculation and it will take quite some time to invent one. That is enjoyable work and chess study has to be fun, after all.

    The positional exercises don't walk away and usually I finish what I have began, no matter the delay. So don't worry.

  5. Styoko exercises is what Rowson, Kotov, Heisman, and many others suggest and it seems the most logical. Get better at X by doing X.

    As I said in previous post, Aaagaard's book on excelling at calculation has a ton of exercises ideal for this. He has a nice rating system for the problems: rating depends on which lines you actually visualized. This gives you a sense for how deeply you need to go but also if you are going broadly enough, often the biggest problem for patzers. Plus, as a bonus, before the exercises he explicitly discusses the topic of calculation for an entire book, such as breadth versus depth, the process of coming up with candidate moves in the first place, etc.

    Book here.

    I wrote about this topic here with some other techniques people suggest. Some good suggestions in comments too. I now prefer flat-out Styoko exercises to actually writing out my tree of variations.

  6. Blue,
    I already did some sort of Stoyko exercises in the past. There are certainly good elements in it.

    This post describes exactly what I think that the problem of calculation is. So first I want to find out if such exercise fixes exactly this problem. I have to investigate that first.

    I don't want to throw myself in exercises anymore without them being plausible to fix the problems as decribed in this post. I have done that enough in the past.

    At the moment I'm reading Aagaards excelling at calculation. I like his way of thinking very much. But already it becomes clear that I miss a few things in his descriptions (as in Kotov, Tisdall and Dvoretsky). It is not complete from a patzers point of view. But it is very helpful to form my own thoughts.

  7. Whats about asking Igor Khmelnitsky himself? he offers advice at his webpage. he knows what he meant with "calculation" and he is a recomended chess trainer? he should have experienxe with improvement based on his test.
    "If you would like me to review your scores and provide training suggestions -"

  8. Uwe,
    Why not think for myself first?
    Why let him have all the fun?

    Of course it is good to know that I have a last resort when everything else fails.

    Now the problem is formulated so clear I would be surprised if it proves to be impossible to find an answer myself. With the aid of suggestions of my fellow bloggers, of course.

  9. Calculation is visualizing candidate moves and their consequences. That's easy. The hard part is doing it well.

    I wouldn't say you should just throw yourself into exercises, thoughtlessly, like someone might do at Chess Tactics Server. You need good feedback. The feedback (rating system) in Aagaard's book is good it seems.

    Obviously even better would be to go over the problems with a good coach right when you finish. Rowson talks about this, how his best experience with a coach and coaching is having students do exercises and then right when done they go over them. It obviously left a stark impression on him.

    The more intelligent and sensitive to your particulars the feedback loop, the better.

  10. Maybe if you like Aagaard you may also like "Inside the Chess Mind"?

    My understanding is that GMs use medium/long term memory while calculating. STM is too small. Calculations can last for much more than a minute.

    The problem of selecting candidates is partly addressed by studying imbalances first (HTRYC).

    Stronger players better know how to deal with gambit openings anyway.