woensdag 25 februari 2009

More on calculation and patterns

Temposchluckers attempts to shed some light on calculation and patterns (and the following posts) has inspired me to share some thoughts of my own on this subject with the readers of my blog. This is not primaraly meant as a comment or criticism on Tempo's observations and conclusions, but more a reflection on my own dicoveries and during the past year.

I have already stated several times that I think that for a player of my strength (2050+), I pretty much suck at calculation and visualization. So if my rating has some validity at all (and I like to believe is does) I have to do other things relatively well for a player of my strength. It seems to me that I am pretty good at pattern recognition.

Of course pattern recognition, calculation and visualization heavily interact. And is is very difficult to divide between these three. I have asked myslef if it is at all possible to find the best move without any knowledge of the appropriate pattern? Imho this is hardly possible. One could make a case that in a position where it is mate in 1 move, the best move can be found simply by exhausting every possible move that gives a check. But one could also say this checkmate is a pattern that is eventually recognized.

Because of the number of possibilities, which rises exponentially with every move, sequences of more than 3 to 5 ply cannot be found by calculation and visualization alone. At least not by humans. Computers however can.

But for us mortals, calculation and visualization have to be guided. This is largely done by the search for patterns. A player who finds the relevant patterns does not have to calculate and visualize very as much as a player who doesn't. The latter is clueless. He will have to calculate a lot of move sequences to find as much as a decent (not losing) move, and have a small chance of finding the best.

Computers however can use brute force. And they are so powerful now, that their ability to prevent weak moves within their horizon, outweighs the importance of even the strongest human players to (occasionally) find the best move beyond that horizon.

So in order to improve there are two leads. The first is to improve my calculation and visualization, and the other is improving pattern recognition. In my next post I will tell you the choice I have made the past year (in which I gained over rating points, reaching an all time high).

5 opmerkingen:

chesstiger zei

I gamble that you will try to improve your calculation since visualisation of a +2050 player is already good enough. :-)

likesforests zei

Glad you are back, and looking forward to your next post on the subject.

Phaedrus zei

CT,
I will not yet give away what I have done to maintain the cliffhanger. But what I can tell you is that I am not able to distinguish calculation and Visualization very well. So in my approach this is a package.

LF,
Thank you for your kind words, they are much appreciated.

Temposchlucker zei

In the past both DLM and prof. de Groot have put me on the wrong foot with their statements about pattern recognition. I wrote a lot about that and allthough I know that my previous notion is wrong, I never could wrap my head around a new decent definition of pattern recognition. When you wrote about pattern recognition in the past, I had an old association with that term. Hence I usually disagreed with you.

Only now I start to see that your use of pattern recognition has a lot resemblance with what I called recognition of characteristics of the position.
(The use of narratives that I advocate is only a temporary means to transfer these patterns into procedural memory.)

According to this new definition of pattern recognition there are 3 elements of which the whole of calculation exists:

Pattern recognition.
Evalution and other conscious logic thoughts.
Visualization.

It it hard to believe that you are bad in logical thinking, so that leaves visualisation. You often already said that. I begin to belief that it might be true. If so, the problem should be easy to fix. Unless we are still missing an unknown factor.

Phaedrus zei

TS,
Communication is missing each others point as nearly as possible!

I think you are right when you say that my definition of "pattern recognition" is broader than the one you used. This may indeed have caused some confusion in that past between us.

I will leave the question open for now if I think that Visualization is the key for further improvement for me.

But if it is, I do not know if there is an easy fix. Yet if there is, and you would help me find it, I will carry you on my shoulders through the streets of your town to a bar where closing hour is only declared when the last customer stumbles home, or when the beer has run out!