[Editor's Note: As you might notice, this article was written by a computer program named ZEUS. While Gelf does not generally give bylines to inanimate objects, we felt that the research compiled by this impressive simulator was interesting enough to merit its own page. For an interview with the guys behind ZEUS about these rankings, please go here.]
Throughout the NFL regular season, a specially designed computer program called ZEUS analyzed the critical downs of every game. ZEUS works by simulating games in a virtual environment and tabulating the statistical results based on factors such as score, time on the clock, down, and yards-to-go. Additionally, a detailed customization of the opponent's strengths and weaknesses based on NFL reported team statistics is incorporated into the simulations. For every playcalling decision to be made, ZEUS simulates the game to completion (hundreds of thousands of times) based on all possible choices (e.g. "run," "short pass," "punt," etc.). The "correct" play call, obviously, is the one that results in the team having the highest chance of winning the game. Often, though, teams make the wrong decision, and it costs them.
Here is an example of how ZEUS works:
Suppose a team is facing fourth-and-one at midfield. The correct call between punting and running is the play that will provide the higher expectation of winning the game on average. If ZEUS suggests the run's expected "Game Winning Chance" (GWC) is two percentage points higher than the punt's, this means that over the course of 100 games, choosing the running play will net two more victories than opting for the punt. The optimum call does not provide a guaranteed result, of courseno decision in football ever does. It simply identifies the best path to victory. Over the course of a long season, critical call errors add up. The ZEUS index reveals just how much these preventable errors accumulate, and how much they can ultimately cost an NFL team.
The ZEUS Play-Calling Rankings below is an assessment of each NFL team's performance in making optimal decisions on fourth downs, points after touchdown, and kickoffs. ZEUS refers to these types of play-calling decisions as "critical" because they often have a great effect on the outcome of a game. Every critical decision for the 2007 regular season was evaluated by ZEUS for its impact on the team's expected winning chances in that particular game. The errors were then summed for the entire 16-game season and the resulting total is the cumulative cost in seasonal win expectation. On average, suboptimal play-calling decisions cost each team .85 wins over the course of the season.
A team facing fewer critical calls during the course of the season may have an artificially lower cost. For this reason, we have examined what we call Hyper-Critical calls during the 2007 regular season, which represent fourth-down decisions with five or fewer yards needed for the first down. The "Error Rate Percentage" reflects the frequency of incorrect decisions in these very important situations. On average, teams made incorrect playcalling decisions during these critical moments more than 50 percent of the time.
ZEUS was developed by Chuck Bower, a senior scientist in experimental particle astrophysics at Indiana University; Frank Frigo, a former backgammon world champion; and Bo Durickovic, the chief strategy officer for a government-services company. You can e-mail them at firstname.lastname@example.org.
Related in Gelf
An interview with Frigo and Bower about these rankings.