Total QBR gets minor modifications
Different approaches to kneel-downs and spikes better reflect realities
Total QBR was introduced in 2011 to improve upon the NFL passer rating by accounting for more of what a quarterback does. A quarterback's passing statistics, fumbles, sacks, rushes, scrambles, number of yards passed in the air versus yards after catch by the receiver and whether receivers were dropping the ball -- these are all in Total QBR.
How something is accounted for in QBR is just as important as what's accounted for in QBR. The method behind QBR puts the quarterback's yards, turnovers, completions and incompletions in the context of when they happen -- down, distance, yards from the end zone, score and time remaining. For example, a quarterback's lost fumble on first-and-goal when down three points is a lot more relevant than a quarterback's lost fumble on fourth-and-1 from the 50-yard line when he's already been stopped behind the line of scrimmage and his team is down a couple of touchdowns. The first fumble probably costs his team points; the second really doesn't cost anything because the decision to go for it on fourth down already potentially sacrificed field position and the inability to gain a yard actually did so. (See our original description of QBR for more details.)
Using QBR, we are not only able to rank quarterbacks in a more complete way than passer rating, we can tell more nuanced stories about how football works. For example:
• When under pressure from a pass rush, NFL quarterbacks as a whole see their average QBR drop under 10, including the impact of sacks, bad throws and any ability to scramble. What distinguishes quarterbacks, though, is how they perform without pressure. The good quarterbacks do well and the bad quarterbacks poorly when there is no pressure. Some middle-of-the-road quarterbacks do pretty well when pressured (Donovan McNabb, Chad Henne, Tim Tebow), and plenty of good overall quarterbacks struggle when pressured (Tom Brady, Eli Manning).
• QBR relates strongly to a team's winning percentage. A team whose QBR in a game is 70 should win the game about 70 percent of the time. A team with a QBR for the whole season of 35 should, as a rule of thumb, win about 35 percent of its games, or 5-6 games in a 16-game season. The Broncos under Tim Tebow and Kyle Orton had a 2011 QBR of 35 and won eight games, suggesting that the rest of the team carried them to an extra three wins or that they were just lucky, given that they were outscored by 81 points over the season.
As we studied QBR, we found a couple of useful surgical changes that better account for quarterback contributions. None of these dramatically changes Total QBR rankings over the course of a season, rarely moving players up or down by more than two or three slots. But making these changes this offseason improved the ease of telling stories.
Here is a summary of the changes:
• Kneel-downs are no longer counted. A kneel-down did not hurt quarterbacks much in their expected points added (EPA), but it added a play to their total. That extra play lowered the efficiency of winning quarterbacks in a way that was inconsistent with performance.
• Spikes are no longer counted for the same general reason. Counting a spike to stop the clock as another play lowered a quarterback's efficiency even though it essentially provides value. Eliminating these from the QBR calculation solved the problem.
• Fumbles on sacks are now not solely assigned to the quarterback. They are divided with the offensive line in the same way that sacks are divided with the offensive line. A number of fumbles on sacks come from hits that a QB could do little to avoid, so dividing responsibility on the fumble should be consistent with the way sacks are treated.
• Fumble recoveries by the quarterback are rare, except when they recover their own fumbles or aborted snaps/handoffs. Given that we don't debit a fumbler fully for fumbling, we now don't reward a player entirely for recovering a fumble. Sometimes another player also could have recovered the fumble easily. This smooths variations associated with the randomness of fumbles and recoveries.
• Penalties, which are not a huge part of QBR, were reviewed and refined. For instance, a quarterback who gets a delay of game penalty while trying to draw the defense offside on fourth down is not debited for that penalty. Trying to draw a defense offside is a strategic call and rarely a mistake of any significance, so adding an extra play was not helpful.
None of these modifications dramatically affects rankings. The modification involving kneel-downs generally raises the QBR values of winning quarterbacks but doesn't really change their order. The redistribution of responsibility for fumbles on sacks can make a difference for quarterbacks who got sacked and fumbled a lot, but the variation isn't huge. Jay Cutler in 2010 is the biggest rank change because he was both on a team that won a lot, so he was kneeling (he had the most that year), and he was getting sacked and fumbling a lot (he had the most here, too).
These changes also don't affect QBR's status as a key statistic. As coaches preach winning the turnover battle or winning the passer rating battle, winning the QBR battle actually has a stronger relationship to winning. Winning the yards battle wins the game 70 percent of the time (since 2008). Winning the turnover battle wins the game 78 percent of the time. Winning the passer rating battle wins the game 79 percent of the time. But winning QBR wins the game 86 percent of the time.
Finally, we should add that QBR has not been built or modified in a vacuum. Throughout its construction and revision, we have had discussions with football analysts both inside ESPN and out, including Brian Burke at AdvancedNFLStats.com. (See its review of QBR.) This year, Total QBR will be posted not only on ESPN, but also at Pro-Football-Reference.com and FootballOutsiders.com. We appreciate their cooperation in providing what we think is useful quarterback information to football fans.
The NFL on ESPN.com