There's a new paper that quantifies how well Kasparov plays under blitz conditions. His performance still matches that of the top dozen players in the world. This result tends to show that searching deeply may be even less important to the performance of top players than we'd realized, and pattern recognition even more important.

Summary: http://www.freakonomics.com/2011/06/20/ ... erception/

Paper: http://dspace.brunel.ac.uk/bitstream/24 ... search.pdf

One fact: Kasparov studied his opponents' games beforehand. I have no idea what difference that makes.

This isn't Go, so who knows how it generalizes. Anyone with a lot of money want to arrange a similar experiment for Go?

_________________
Occupy Babel!

As far as I know that makes a big difference. Because the opening is so important in chess, you get a very useful insight which opening your opponent knows and which variants he plays - so you can prepare those. On the other side if you can play something he has never played before or not that often, you might have an advantage.


post scriptum: But well, this is grandmaster level, so I don't know how much of an advantage you can get ^^

_________________
My "guide" to become stronger in Go

Well, to me this study shows the dangers lurking in interpreting statistical results.


First, 20 seconds a move would not traditionally be thought of as blitz. That's over 13 minutes for 40 moves. Blitz is normally 5 minutes or less for an entire game.

Second, a 2650 ELO level is almost 200 points weaker than Kasparov's peak rating. And those 200 points represent a large difference in strength. The higher ELO gets, the harder it is to improve. It is the law of diminishing returns.

So I would actually look at the results of this study and reach the opposite conclusion. That Kasparov was suprisingly weak (relative only to his normal enormous strength of course) when he had to play moves in 20 seconds, and this indicates the importance of calculation to his overall strength.

All that having been said, I have no doubt that pattern recognition and judgment play a huge role in top level chess. Stronger players are very good at pruning out unpromising moves quickly, and focusing their analysis on the most promising moves.


(I just note that Kasparov is famous for calculating a great deal during his games, so the idea that calculation does not play a big role in his strength seems rather unlikely.)

I wonder if the two are not independent from one another. For example, I wonder if searching deeply on a regular basis aids in promoting pattern recognition ability.

It's just a hypothesis, but it seems to me that, when I practice searching deeply, I am forced to visualize patterns in my mind - which could, potentially, help me to remember the patterns more easily.

_________________
be immersed

My first thought when I saw this was, "Hasn't all of this been studied before?" Then I looked at the paper. It's from 1996.

I don't remember where I saw this, but I vaguely recall reading an article or interview where someone, perhaps one of Kasparov's trainers, was discussing how he thinks. And I think he said that Kasparov's real strength isn't so much in his ability to calculate concrete variations (although he is of course a fearsome calculator) as much as it is that his understanding of dynamic features of positions is superior to most other players. He has a better sense of when it's good to give up material for the initiative, when it's good to wreck his pawn structure to open up lines for his pieces, and so on. So even when his games look like a mass of complicated tactics, there is a great deal of positional/strategic intuition underlying them.

On the other hand, Magnus Carlsen says that the biggest difference between him and Kasparov is that Kasparov calculates better while he (Carlsen) has better positional intuition.

There are also a lot of strong chess players who will tell you that positionally correct moves in chess are usually tactically correct as well -- if you know which features of the position are the important ones.

I don't think anyone would doubt that deep analysis helps to build pattern recognition, but there's an interesting question as to whether (for example) it's better to study a few positions deeply or to spend the same amount of time studying a wide variety of positions more superficially. (I tend to suspect the latter is better, up to a point.)

:20 per move isn't blitz. This is blitz:
http://www.youtube.com/watch?v=Bzrap8Vtyq8

This is true blitz

http://www.youtube.com/watch?v=CuI9JaHxNrc

Minor point: you want his average rating over the period those games were played, not his peak rating.

Today there are 97 players at ELO 2650 or higher. That's still striking.

_________________
Occupy Babel!