woensdag 20 februari 2008

High level scanning

The next position (Arkhipov-Kuznetsov, Russia 1980) is a forced win for white. You might want to solve it, and are invited to do so if you wish, but this not necessary to get the rest of this post.

What I will try to show you is what scanning and pattern recognition are about. In this post Temposchlucker refers to a article in which scientists reported that they had discovered that: "Grandmasters while studying chess get to see hundreds of thousands of different (chess board) positions. They store these 'chunks' in their brain. When they see a similar position, they retrieve the data from their memory. But what do they retreive exactly? To explain this I will introduce first some patterns that are relevant to solving this position. Patterns that any (grand)master will retreive when he is confronted with this position."

The next diagram is a standard mating position. Note that a piece on d7 is necessary (not a knight though) to prevent the king form escaping. Also note that it does not matter if the king is on g8, f8 or e8.

The following diagram is also a common mating pattern. It occurs frequently when the g-file is opened with force.

Armed with these chunks (and on equal terms with the focal gamma bursted grandmaster) we now return to the initial position.

When we start calculating a lot of us will now come up with the following variations:

1 Qh5 gxh5 2 Rg3+ Bg7 3 Rxg7+ Kf8 4 Rxh7 (see pattern 1)
1 Qh5 gxh5 2 Rg3+ Bg7 3 Rxg7+ Kh8 4 Rg3 (see pattern 2)
1 Qh5 h6 2 Qxh6 Bxh6 3 Rxh6 (see pattern 1)

Unfortunately however there is that darn Queen on c5 preventing this winning combination (1 Qh5? Qxh5). But a simple deflection does the trick here. After 1 b4! Qxb4 we are back to the variations given above.

This is what high level scanning should bring about. Focal gamma bursts that retrieve the right patterns and that make calculating look really easy! But to lift your scanning skills enough to reach this level, they have to be build up step by step.

Geen opmerkingen: