Wednesday, April 12, 2006

Skill and understanding

Funky sort of asked me about the difference between skill and understanding in chess.
I'll give it a try.
The brain has different "parts" that can perform different tasks at the same time.
You can easy see that for yourself.
It is possible to calculate the square root of 6084 while at the same you sing "Mary has a little lamb".
One part can only perform one task at the time. So it is not possible to calculate the square root of 6084 while at the same time you add 2347 to 5378 (try!).
If you can find out how much different things you can do at the same time, you will see how many "centers of gravity" the brain has.
The main parts are intellectual (where you calculate), emotional (feeling) and mechanical (singing a well known song). But there are a few more.
In chess the intellectual and mechanical parts of the brain play the most important role.
Thinking and understanding are typically done by the intellectual part.
Automatic skills typically belong to the mechanical part of the brain.

Have a look at the great article at Susan Polgars blog.
She played a blitz game (1 minute vs 5 minutes) against a boy, and won.
"Once you have a winning position," Susan said, "play with your hands, not your head. Trust your intuition."
This refers to the mechanical part of the brain. Seeing is doing.

The mastering of every skill starts out with the intellectual center. If you learn to play a piano, you first have to understand a lot of things. You have to understand how to read music etc..
What is typical about the intellectual part, is that it is so much slower than the mechanical part.
When you practice a lot, you have to rely less on your intellectual center since the performance of the task shifts to the mechanical center.
At the same time you speed up a lot.

The same is true for learning chess.
First you have to learn and understand a lot with the intellectual part.
Then this knowledge has to be transferred to the mechanical part by practicing.
That creates room for new understanding (you can only perform one task at the time with the same center, remember?).

Since the game of chess is so complicated, we haven't to be afraid of playing it all by mechanical skills. There will always remain plenty necessity to understand and think during your game.
But the more understanding you have transformed into automated skills, the higher your rating will be.
Sorry for this rather mechanical view of mastership. I know many will hate it.
Logic obligates me to say it this way.
Being romantic is not my forte.


  1. Awesome post TS. You should change fields to neuro-science! Or how about AI? You aren't in AI are you?


  2. An interesting topic. This is just a note from the peanut gallery on how the cognitive neuropsychologists describe what you are talking about.

    Typically, they break memory up into 'procedural memory' (e.g., remembering how to ride a bike), and 'declarative memory' (e.g., being able to consciously recollect and describe facts such as the color of your first bike). These two types of memory correspond to two different types of knowledge, knowing how to do something (e.g., I know how to throw a baseball or do a Judo throw), or knowing that something is the case (e.g., knowing that your first bike was a Schwinn).

    I like de la Maza because it is clear that for experts, much of chess is know-how, tapping into the vast and unconscious procedural memory system that is built up via extensive experience. It is not a special ability to think faster. So I, and I think any psychologist who has studied chess, would agree with your evaluation, Temposchlucker.

    In neuroscience, tracking down procedural memory is much harder than declarative. If you knock out the hippocampus, you get anterograde amnesia, in which you don't form new declarative memories (including remembering new events and facts). However, with such damage, you can still learn how to ride a bike. You won't remember that you learned how do do it (you won't know that you can ride a bike), but you will actually do it better over time.

    It seems procedural knowledge/memory is much more distributed in the brain, such that nobody has yet found a single location that eliminates it. That is, there is no procedural anterograde amnesia that we know of. In rats, you can remove all sorts of structures, and they can still learn to push the right lever in a given context (a form of procedural learning). It seems to be so adaptively important that there is tons of redundancy in the brain so that a single stroke or head trauma won't make you unable to learn how to do new things.

    Some (there aren't good statistics on this) patients with retrograde amnesia (loss of declarative memory for events before the event) retain a good deal of their chess skills. It would be an interesting experiment to try to teach someone with anterograde amnesia to play chess. I bet they wouldn't get very good, but would learn it. They can learn a new programming language.

    This is probably one of the best popular scientific treatments of the subject, by two giants in the field.