Total quarterback rating (abbreviated as total QBR or simply QBR) is a proprietary statistic created by ESPN in 2011 to measure the performance of quarterbacks in American football. According to ESPN, "QBR incorporates all of a quarterback's contributions to winning, including how he impacts the game on passes, rushes, turnovers, and penalties. Since QBR is built from the play level, it accounts for a team's level of success or failure on every play to provide the proper context, then allocates credit to the quarterback and his teammates to produce a clearer measure of quarterback efficiency."
ESPN claims that it created QBR to be a more meaningful alternative to the traditional passer rating statistic. However, ESPN has never released the complete formulas and procedures to calculate QBR, resulting in criticism among fans and commentators alike questioning the results.
History and development
Total QBR was developed by a team at ESPN Stats & Information Group including Jeff Bennett, Dean Oliver, Alok Pattani, Albert Larcada, and Menlo College professor Ben Alamar. The group also received input from ESPN analysts Trent Dilfer, Jon Gruden, and Ron Jaworski. Total QBR was developed based on analysis of 60,000 NFL plays between 2008 and 2010, and was unveiled on August 5, 2011. The formula was modified in 2012 and again in 2013.
Characteristics
There are six steps to building QBR: Each QB "action play" (passes, rushes, sacks, scrambles, or penalties attributable to the QB) is measured in terms of the expected points added (EPA) Adjust for the difficulty of each play. EPA is adjusted based on the type and depth of a pass, and whether the QB was pressured. If there is a completion, he only is credited for the typical number of yards after the catch (passer rating takes all yards into effect) based on the type and depth of the pass There is a discount on garbage time, or a time where the score is out of reach near the end of the game. Opponent adjustment: More credit is given with tougher defenses and vice versa. QBR averages the adjusted EPA per play and transforms it to a 0 to 100 scale, with 50 being average.
Summary of computation
Raw QBR is calculated as the following:
,
where g() is a function that scales from 0-100, where 50 is average. Total QBR is the raw QBR adjusted for the strength of the opponent.
EPA is calculated based on the down, distance, and the yard line at snap, with each combination having its own point value. The point values are the average net point advantage the team on offense can expect given the particular down, distance, and field position. For example, a 1st and goal chance on the opponent's’ 1 yard line heavily favors the offense, yielding a positive point value. On the other hand, a 3rd and 9 on the team's own 3 yard line is heavily negative because it drastically favors the opponent.
The value of each play's outcome is measured by the snap-to-snap change in expected points. This is called Expected Points Added. The Expected Points Added (or lost) in each play are divided among the contributing players on the field based on the role of each player and the type of play. Deeper throws give a higher share of credit to the QB, while screen passes give relatively less credit to the QB and more to the receiver.
Plays that occur in “trash time” are discounted by as much as 30%. Trash time is measured based on the leverage of each play which is primarily a function of score, time, and field position. Important, critical plays that are likely to change the outcome have high leverage, while plays that occur after the game has largely been already decided have low leverage. QBR discounts low leverage plays, but does not boost credit for “clutch” plays.
After each play's Expected Points Added is adjusted for difficulty, division of credit, and trash time, it is averaged on a per play basis. This average is further adjusted to account for the strength of opponent. Performance against a stronger defense that tends to allow low adjusted EPA per play is adjusted upwards while performance against a weaker defense is adjusted downwards. The degree of adjustment is in direct proportion to the strength of the opponent.
Lastly, the resulting adjusted EPA per play is transformed to a 0 to 100 scale, where 50 is average. The result can be thought of as a percentile. For example, a QBR of 80 means that the QB's performance is better than 80% of the game performances by QB's since 2006. A game QBR of 80 would also mean that, given that QB's performance, his team would be expected to win that game on average 80% of the time.