In short: How chess games are resolved with respect to the ranking (horizontal) and number of moves (vertical). When red, a draw is most likely.
Description: Can visualizing some data convey the emotion behind it? Good chess players are master strategists and inevitably being given the option of accepting a draw when a slight advantage has not been quickly achieved can be attractive for the wider strategy of maintaining the player’s overall ranking. However, for the spectator this can result in a frustrating number of drawn games.
An image from this article shows “a (genuine) snapshot of an organizer and his wife during the press conference after two world-class grand-masters had agreed to a draw after thirteen moves.”. This fantastic photo shows layers of underlying emotion that any artist would do well to convey in a piece of artwork. Taking open data on the outcomes from 900,000 high-level chess games, can the data show what they’re thinking? And would the resulting image keep the necessary objectivity?
Each game is added as a single-lined rectangle extending from the central divide. The bigger the difference in ranking the furthest from the center. The height is the number of moves. White victories are added in white, black in black. Where the spectator is more likely to see a draw the color is set to a featureless red.
What results is something almost akin to a Venn diagram, but with depth. As expected where one side is much more highly ranked than the other they are often victorious and as the two ranks close, the resolution is less certain. What also emerges is the advantage of white moving first as the structure of the data is offset. The areas where a draw is likely are clear, they are either long drawn-out games or where the rankings are close. The tight central circle of red is the area where closely ranked players agree their draw without giving the game a chance.
An interesting problem when showing three-state data in what… should be… a two-state situation is choosing the color pallet. The image could have easily be shown in, for example, red-green-blue but to get across the true nature of the situation black and white had to be kept, with the exception of the “unnatural” situation of a draw or a player winning against the odds being the only color introduced.
Point of interest: (Aside from the smallest number of moves before a draw is offered being a scandalous THREE). In international chess, the first 40 moves have to be completed in 2 hours, with an extra hour to reach 60 moves. Both 40 and 60 moves are seen in the data as moves at which the player favored most by their ranking is less likely to be victorious.
Technical: Each chess games in the KingBase Lite database (900,000 games from 2000 onward for players with an ELO rating of >2200) plotted as a single (half-)rectangle adjoining a central boundary. The distance from the central boundary is the difference in rankings between players. The height, the number of moves. Each game is added in the color of the victor.
For example, in this image, two games are displayed: a white victory over black when the white player has a much higher ranking and completes his/her victory in a relatively small number of moves; and a black victory over white in a greater number of moves but with black only having a slightly higher rank. As the image is built up blends of wins/losses and draws come through. Where the number of draws is greater than the number of wins/losses the color is set to a flat red. Where a player is victorious despite the ranking difference the blue channel is weighed to highlight this. The original image is 3000k pixels, 1000k is uploaded here.
The Graph Is Dead 