Football Analytics with Python & R

Chapter 1. Football Analytics

American football (also known as gridiron football or North American football and henceforth simply called football) is undergoing a drastic shift toward the quantitative. Prior to the last half of a decade or so, most of football analytics was confined to a few seminal pieces of work. Arguably the earliest example of analytics being used in football occurred when former Brigham Young University, Chicago Bears, Cincinnati Bengals, and San Diego Chargers quarterback Virgil Carter created the notion of an expected point as coauthor of the 1971 paper “Technical Note: Operations Research in Football” before he teamed with the legendary Bill Walsh as the first quarterback to execute what is now known as the West Coast offense.

The idea of an expected point is incredibly important in football, as the game by its very nature is discrete: a collection of a finite number of plays (also called downs) that require the offense to go a certain distance (in yards) before having to surrender the ball to the opposing team. If the line to gain is the opponent’s end zone, the offense scores a touchdown, which is worth, on average, seven points after a post-touchdown conversion. Hence, the expected point provides an estimated, or expected value for the number of points you would expect a team to score given the current game situation on that drive.

Football statistics have largely been confined to offensive players, and have been doled out in the currency of yards gained ...