I really like this visualization (via FlowingData) of the most common motions of chess pieces, according to the 2million+ games in the fdsf dataset. Here's the chart for the black knight, which does a good job of scouting the entire board:
and here's the black bishop:
An elegant design decision is not using straight lines to connect the starting and ending points for each piece (which in the case of the bishop especially, would cause many different trajectories to overlap). Instead, a curved line is drawn with a little random variation added to allow the frequencies across gamer to be expressed. Creator Steve Tung explained the process in more detail in an interview with i09:
I downloaded the Million Base chess database of ~2.2 million chess games http://www.top-5000.nl/pgn.htm (click the link ‘Million Base 2.2’). Then I ran the ~1.5GB database through a program called pgn-extract to reformat the data and a custom Python program to make a list of moves and the total number of times they take place for each chess piece.
For each entry on that list, a node.js program draws a faint line for every 500 times (rounding up) that the move occurs. Each curve’s start and end points correspond to the move, and its middle point is offset to give the curve both direction and height, with a little bit of noise added in so the curves don’t clump up too much.
That's all from us for this week. We'll be back on Monday live from the EARL conference in Boston. See you there!
Comments