The offseason offers an opportunity to reflect on the pressing sabremeteric questions of the times: free agent value, fielding metrics, pitchf/x release points. All worthy and important pursuits. Here, though, I hope you will indulge me as I address something much more trivial: the network (or “graph,” as Zuck would say) of baseball players’ twitter accounts.
You see, some baseball players tweet, and, not surprisingly, they follow other baseball players. But what does the network of these connections look like? Are teammates more likely to follow one another? Surely, but how much more likely? If player A follows player B, how likely is it that player B follows player A? And, more generally, how connected is the network — in other words how likely is it that any one player follows another?
Here is a representation of that network. The nodes are twitter accounts and they are connected by directed links, the arrow going from the follower to the followed. The image is interactive so you can hover over and click on nodes to get more info. Sorry it is a canvas element so it will not work in IE8 or older.
All told the network has a 5% connectivity. That is 5% of potential connections exist. But these links are not even across the network and their distribution among players is fairly leptokurtic; some guys have tons of followers (Nick Swisher, C.J. Wilson) and some follow almost everyone (Rob Bowen, Drew Storen), while most have just a handful of followers and follows.
There is a high degree of reciprocity. If player A follows player B then player B will follow player A 34% of the time, almost seven times the baseline follow rate. Teammates follow each another at a 42% rate, over eight times the baseline follow rate. Nick Swisher, for example, only follows his two tweeting teammates, Joba Chamberlain and CC Sabathia.
So there you have it, the network of baseball players’ twitter accounts.
For those who want the geeky details: I got the list of MLB player twitter accounts from @MLBplayers, so I am missing anyone they are. I used Twitter’s API to see who was following whom; positioned the nodes with the R package network; and sketched up the interactive graph with processing.js.