Interesting Scientific Chess Article

Dec 27, 2010, 7:44 AM |

Hey everybody, check out this Article I just found:

In that study, scientists analyzed correlations between response time, player ratings, and move quality (as measured by crafty evaluations) in a set of about 2.8 million blitz games (3 0 games).

Some of the findings are interesting, and some are expected. For example, Figure 2D (shown below) shows the distribution of a particular piece move through out the game; you can see the classic generalizations of opening with a pawn, knights before bishops, castle early, and use your rooks and king in the endgame born out through these statistics. In the figure, frequency at which moves are made with the indicated piece is shown on a scale from red (high) to blue (low). The X-axis is move #; the bottom row, in case you can't tell, represents the castling move.

As a scientist, I particularly enjoy seeing data like this; even though we all "know" that you should castle early, and that stronger players make less blunders, it's not really knowledge until its shown empirically. (Example games that highlight the themes, while good teaching tools, are essentially anecdotal evidence for these claims.)