Football Analytics with Python & R

Chapter 2. Exploratory Data Analysis: Stable Versus Unstable Quarterback Statistics

In any field of study, a level of intuition (commonly known as a gut feeling) can exist that separates the truly great subject-matter experts from the average ones, the average ones from the early-career professionals, or the early-career professionals from the novices. In football, that skill is said to manifest itself in player evaluation, as some scouts are perceived to have a knack for identifying talent through great intuition earned over years of honing their craft. Player traits that translate from one situation to the next—whether from college football to the professional ranks, or from one coach’s scheme to another’s—require recognition and further investigation, while player outcomes that cannot be measured (at least using current data and tools) are discarded. Experts in player evaluation also know how to properly communicate the fruits of their labor in order to gain maximum collective benefit from it.

While traditional scouting and football analytics are often considered at odds with each other, the statistical evaluation of players requires essentially the same process. Great football analysts are able to, when evaluating a player’s data (or multiple players’ data), find the right data specs to interrogate, production metrics to use, situational factors to control for, and information to discard. How do you acquire such an acumen? The same way a scout does. Through years of deliberate ...