How is Total QBR calculated? We explain our (improved) QB rating

AP Photo/Bill Feig

The intent of Total QBR is to isolate each NFL quarterback's contribution to his team's fortunes as accurately as possible with the data available. It measures nearly every aspect of quarterback play, from passing to designed runs to scrambles to turnovers to penalties. QBR also strives to separate the performance of the individual quarterback from the rest of his team, all in an effort to rate the overall efficiency of each quarterback in the league.

One thing it hasn't done, however, is to account for the strength of the 11 players on the other side of the line of scrimmage.

But now, QBR accounts for opposing defenses: Quarterbacks who face tougher defenses will have their ratings adjusted upward in proportion to the difficulty of their opposition, and those who face weaker defenses will have their ratings adjusted downward. This is one more step toward gaining a true measure of overall quarterback performance -- and a big step.

Here's how the opponent adjustment works. Early in the season, we don't have enough information on how good each defense truly is. It usually takes several weeks of data to have a reasonable estimate of how tough each defense is. So the adjustment starts with an estimate for each defense, which is based on a model of how defensive performance varies from one season to the next. Then, as each week is played, we update that estimate with performance information from each game. The result is an accurate and relatively stable estimate of how tough each defense is for a quarterback to face. Several weeks into the season, the effect of the early estimate becomes very small, and by the end of the season it is virtually nonexistent.

Next, we compute an adjustment to each quarterback's performance based on those estimates of how tough each defense is. We also factor in home-field advantage. It's no surprise that quarterbacks playing at home tend to have an easier time than those playing on the road. These adjustments are made underneath the hood of QBR -- on each quarterback's share of his team's expected points added (EPA) per play, which factors in nearly all aspects of quarterback play, including passes, scrambles, designed runs, sacks, turnovers and even some penalties. Those EPA rates are also adjusted to correct for performance in "garbage time," when the game's outcome has effectively been decided.

Lastly, those adjusted EPA rates are scaled to produce the final opponent-adjusted QBR, which is a number from 0 to 100. This scale can be thought of as a percentile of game-level performance -- 50 is average and a 99 would be among the best games on record. Quarterbacks don't often get below a QBR of 5 (they'll get benched by that point). On a season level, a QBR of 75 or higher would be elite performance, and 25 would be considered replacement-level performance.

Early in the season, the opponent adjustment will have the most impact since many quarterbacks have faced an imbalanced set of opposing defenses. Some examples will help explain how the adjustments affect the ratings.

Through Week 3 in 2016, Ryan Tannehill's unadjusted raw QBR of 53.0 suggests he has been playing at a league-average level. But his adjusted QBR of 62.2 is significantly higher, highlighting how difficult his opponents have been so far. The Dolphins have faced the Seahawks and Patriots, both on the road, plus the Browns, whose defense so far has been middle of the pack.

Drew Brees exemplifies the opposite. Through Monday night, his opponent-adjusted QBR is 52.7, lower than his unadjusted raw QBR of 57.5. His Saints have faced the Raiders, Giants and Falcons, all three in the bottom-third of the league in terms of defensive QBR allowed, with two of three games at home.

By the end of the season, most starting quarterbacks have faced an evenly balanced set of opponents, so their adjusted and unadjusted QBRs will be very close. But there are exceptions, and those are the interesting cases. For example, in 2015, one reason the Panthers were able to finish the regular season 15-1 was that they faced a disproportionate number of weaker defenses. The difference between Cam Newton's unadjusted raw QBR of 70.3 and his adjusted QBR of 65.1 (retroactively calculated) reflects the relative weakness of his opponents.

On the game level, the adjustments are usually more stark. Matt Ryan's recent Monday night game at New Orleans was among the week's best performances with a raw QBR of 92.4. But because the Saints' defense is among the league's easiest for quarterbacks to face, Ryan's opponent-adjusted QBR for that game falls to 87.8.

Jacoby Brissett, the Patriots' rookie fill-in, had a storybook debut against the Houston Texans, who have one of the league's most formidable defenses. His raw QBR for the game was 62.1, but after opponent adjustment, it becomes 70.7. The highest-rated game since 2006, when QBR began, is a 99.8 by Colin Kaepernick in the 49ers' 42-10 win against the Jaguars in 2013.

Jacksonville's defense was near average that season, so Kaepernick's raw and adjusted QBR are the same. Kaepernick was 10-for-16 with a TD and no interceptions, but this is where QBR and the traditional passer rating diverge. Kaepernick absorbed no sacks, never fumbled, rushed for 54 yards on seven carries, and added two rushing TDs. Remember that QBR is a measure of efficiency, rather than total production, and this game is almost as efficient as it gets.

The new, adjusted QBR is now the number you'll find under the Total QBR column on the QBR page. The unadjusted QBR is now called "raw QBR" and is still available for all seasons and games. No player rating system can be perfect, but we think adjusting for opposing defenses is one more step forward for measuring total quarterback performance.