Spatial Analysis of Tennis

Recently, Damien Demaj - a cartographer working for ESRI, published a neat piece of work using geovisual analytics to explore the Gold medal match in mens tennis from the 2012 Olympics in London. This match was, of course, between the legendary Roger Federer (holder of 17 grand-slam titles) and hometown favorite Andy Murray. In case you've been living under a rock, the game was eventually won by Murray.

The maps Demaj created combined movement data of the individuals, along with locations of where winning shots were taken. (see below)

Portraying movement data in a single plot/map can be incredibly messy when dealing with large datasets. Even more messy, is the map that includes not only player movement and winning shot location, but that of all shots, and the trajectory path of each of those shots. (see below)

 

This of course is an even messier clump of data, making it difficult to discern if spatial patterns are evident. To desiminate the data in a more usable form, Demaj depicts the winning shots (often termed winners) with at least three strokes, of Murray (n = 18) and Federer (n = 13) indivdually .

From these maps, Demaj is able to make some interesting spatial inferences. First, Murray scored several down-the-line backhand winners. Murray also connected on several winners from deep in the court (near the service line). Federer, on the other hand, was most successful by getting Murray deep or wide from his service game (10 / 13 winners!). What a nice piece of work by D. Demaj, and an even cooler spatial sports dataset.

 

What other types of questions could be investigated with this type of data? Well, I would be interested in seeing if individual players had characteristic spatial patterns as to where they hit shots. Methods from point pattern analysis (e.g., kernel density estimation) could be used to derive smooth surface of shot location intensity (essentially a heat map of shot locations). This could be a useful long-term statistic to keep on players to determine if they have spatial preferences in shot making, especially if grouped by situational context (e.g., backhand vs. forehand). Alternatively, the player movements could be incorporated into a movement heat map to see if movement patterns emerge. Finally, I think an interesting spatial metric to explore with tennis is the distance-to-out-of-bounds of each shot. Such a metric may be able to identify those players capable of hitting the ball closer to out-of-bound (painting the lines so to speak!) as a measure of overall shot effectiveness.

 

Check out more from Damien Demaj by following him on twitter: @damiendemaj