Graphics Reference
In-Depth Information
require some degree of uncertainty to make it more difficult for the
defending team to plan its response. As Major League Baseball pitcher C.J.
Wilson has said, “Pitchers fall into traps. They get predictable with pitch
sequences.”
Sports data has become more widely available, and you can use this data
to create a decision tree. For example, National Football League (NFL) fans
collect and organize play-by-play data and also create programs to parse,
extract, and summarize the data.
Note
For an example of a collection of play-by-play data, see
play-by-play-data.html
. For an example of a program used to parse,
extract, and summarize the data, see
https://github.com/10flow/
Based on this play data, you can construct a decision tree for each team to
see the sequence of decisions on whether to pass the ball or run the ball for
each down. Ideally, a team would not have a visible pattern (such as always
passing on the first down and always running on the second down).
Using a simple Python script, you can count the number of plays in each
sequence, as well as record the average number of yards gained. Each
successivenoderecordsthedataforthatsequence.Forexample,afirst-level
node may indicate a first-down run, and a second-level node may indicate
the previous decision followed by the current decision (for example, a
first-down run followed by a second-down pass, and so on). The output is a
tree that can then be visualized using Cytoscape, as shown in
Figure 10-10
.
In this example, each team has a decision tree summarizing pass versus run
over the 2011 NFL season. Each level in the tree is a successive down. The
first split indicates the decision at first down, the second split indicates the
next decision at second down, and so on. For each split, there can be a pass
(triangle) or run (circle), or neither (an octagon), which indicates a play that
was something else (such as a kick, a sack, or data that was undetermined).
The color indicates the average yards gained for that play type.
