Monday, February 16, 2009

Defensive Fingerprint

Defensive Fingerprint attempts to objectively identify the style of a team’s defense. Inputs into the system are the departure from the D-1 norm of the following defensive characteristics…

- assist percentage (triple weight, higher means a more likely zone team)
- 3-point attempt percentage (triple weight, higher means a more likely zone team)
- free throw attempt percentage (double weight, higher means a more likely man team)
- turnover percentage (single weight, higher means a more likely man team)
- defensive rebounding percentage (variable weight depending on offensive rebounding percentage, higher means a more likely man team)

All of these things are thrown together producing a number which will forever remain a secret. The number is then used to book the defense into one of four categories in an attempt to describe what type of defense the team employs…

- Mostly man: The team probably plays man defense on the vast majority of its possessions.
- Inconclusive: The data does not support identifying defensive tendencies.
- Some zone: It’s likely that this team plays a significant number of possessions in a zone defense.
- Mostly zone: It’s likely that this team plays most of its possessions in a zone defense.

A system such as this will have quite a few misses on blindly guessing a defensive style. It could be improved by using stats adjusted for the quality of competition, but for now I’m using the raw stats. Also, consider that labeling a defense simply by zone or man is an oversimplification. Some zone defenses are aggressive enough to function statistically as man defenses, and vice versa. But if you notice a label that looks out of place, let me know. This system will undergo some iterations.