Monday, December 7, 2009

Parallelisms, Chunks, and Visualizations

On move 4, I castled. He then proceeded to move two pieces and take two of my pieces. "You can't do that," I protested.

"Why not?" he retorted.

"Because chess is a turn-based game. Each player gets only one move," I explained.

"But you just moved both your king and rook," he said, an annoying passive aggressive tone. "I should get two moves also."

Angrily, I made a move to sweep his entire army from the board with my arm, but the pieces came alive and fenced with each other, not turn by turn, but simultaneously, mano a mano across the span of the board. After a few seconds, the entire front froze again and my vision held it for a moment...
And then I woke up.

In trying to understand more about de Groot, I came across a research paper by Reingold and Charness which referenced de Groot. Reingold and Charness researched chess playing abilities to get at the nature of what is at work. A review of past related papers showed:

1. "Experts" (ELO 2200-2400) have a larger "visual span" as compared with "Intermediates" (ELO 1400-1700) and "Novices" (ELO U1400). Visual span is the ability in speed and area to remember details of a chess position at a glance.

2. Experts (ELO 1950+) have fewer saccadic eye movements in check detection than Intermediates (ELO 950-1400) and Novices (ELO U950), implying greater peripheral vision for detecting checks.

3. Experts spend less time looking directly at pieces and more time looking at empty squares between them than intermediates and novices and when they do look at pieces, they spend more time on "relevant" pieces.

The experimental contributions of this particular paper were to try to suss out if there were time costs for interpreting incremental increases in the complexity of chess positions.

1. Experts didn't seem to be slowed by adding one piece to various check detection tests.

2. Cueing a piece improved the reaction times to correctly identifying check status of that piece for weaker players, but did not help experts. In fact, the cue actually seemed to hurt the experts' reaction times.

The authors suggest that their results indicate that experts have an ability to encode position information in a parallel as opposed to a serial manner. Parallel processing was the force multiplier used by Deep Blue to ultimately defeat reigning world champion Garry Kasparov on May 11, 1997.

A friend of mine pointed me to the "My Brilliant Brain" series about Susan Polgar available on YouTube. The series' central finding seemed to be that extraordinary mental feats of memory utilize "chunking" and that Susan Polgar's ability may reside in the fusiform face area of her brain. An example of chunking would be to use "a Black kingside fianchetto" instead of black king at g8, black bishop on g7, and black pawns on f7, g6, and h7. After one game, Susan said that pattern recognition and intuition guide her through a game, especially at fast time controls. A PET scan indicated that the fusiform face area in Polgar's brain was being utilized as a specialized chess information recognition center.

I think I'm going to try to chunk some of my tactical exercises at Chess Tempo and do them blindfolded to exercise my chess eye.

Here's a Chess Tempo exercise I failed:

Chess Tempo #71555
White is up a pawn. Black's queen is awfully deep in my territory - trap it? cxb4 can happen any time. Black's Bg7 and Re8 look pretty inactive thanks to my pawn shield. Black's knight with the Bc6 help can check on f3 forking White king and rook.
Bc4 hits queen and cuts off its escape along a2-g8 diagonal. Then Qxa3 loses a pawn. Re3 hits queen again. a1-a6 are all covered, a3-e3 attacked by the rook, b2 and b4 - no escape! If Nf3+, I'll have to move to g3. Rxf3 allows Qxf3. I'm going with Bc4.

I played Bc4. Chess Tempo responded Qxa3. I played Re3 thinking, "I have you now.". Chess Tempo played Nf3+. Kg3 looks best. Failed.

Through #6.overconfidence in #1.queen complexities, I #4.cut off analysis prematurely and #3.overlooked a defensive resource of a #9.backward #8.knight move.

Kg3 fails to Nxd2 Rxa3 Ne4+ forking king and queen and material becomes even. I didn't even consider Qxf3 Bxf3 Rxa3 to know that White ends up a piece ahead.

No comments: