Thursday, February 21, 2013

Bustification of trial and error

After 3 days exercising the KNpKN endgame, my score went up to 80% succesrate. Enough to play this endgame with confidence in practice. Given the complications, I suspected that I would need at least 2-3 weeks to master this endgame, just as I needed for KRpKR.

Why did I learn this endgame so much faster?

I made a diagram of each position I didn't grasp straight away. By taking my time to define the goals and the relevant squares I was able to really grasp what is going on in a position. The diagrams were put in Anki and repeated. The hundreds of possibilities that arise in a trial and error approach were thus limited to only a few common themes. These themes or fighting methods or manoeuvers proved to be transferable from one position to another.

What can we learn from this?
Trial and error is totally busted as method. We already had hypothesized that, but now it is proven.
Every investigation of an unknown endgame has the following steps:
  • Defining the maingoal of the position. For instance: guard the pawn towards the promotion square or create a passer etc..
  • Define the subgoals for both sides that arise in reaction to the main goal. For instance: sac the knight for the latest pawn. Fork King and pawn. Block the promotion square with the king etc.. What one side tries to accomplish must the other side prevent.
  • Define the squares that need to be dominated in order to reach the goals.
  • Investigate the moves that lead to the required domination.
There are the following drawbacks with trial and error:
  • You need way too much time with positions that should be pruned beforehand. It took me 3 full weeks to master the KRpKR endgame by solving CT problems. If I had taken my time to work positions totally out with diagrams, it would only have costed 3-4 days.
  • Second is that recall of the final methods is weak and error prone. Since I haven't been exact, I only have a global idea how to approach that endgame.
Troyis redux.
When I got better at Troyis by just playing it, I worked on my complex motorskills. This is the kind of thing we can train unconscious on autopilot by just doing. The result of playing with Troyis was that I could do this exercise as fast as a grandmaster. So automatic training has a clear function when it comes to handling the dynamics of a position fast. It works on complex motorskills. Complex motorskills are geared around the dynamic handling of pieces in the mind. You need to be able to do that. But mere playing is already enough to exercise these skills.
When playing Troyis, I hypothesized that I could become even better and faster by inventing a strategy how to play it. I didn't test that at the time, since, well, it was only Troyis.
But now I tested the very same idea with the KNpKN endgame.

Defining the manoeuvers.
When you define the manoeuvers that belong to a certain endgame with the aid of diagrams, you are in fact pouring knowledge into patterns. When recalling these patterns with a high level of knowledge, you cut down on thinking time. Since the thinking already has been done during study time. The pruning of the irrelevant is already integrated into the pattern.

The past and the future.
In the past I have spilled my time and energy way too much with trying to improve my motorskills. What I must do now in stead is to pour all kinds of chess knowledge into patterns with the aid of diagrams and learn to recognize those patterns. Which is what I'm doing now.

No comments:

Post a Comment