By W. Casan Scott
I’ve been skeptical of quarterback metrics for quite some time now. Most agree, including myself, that quarterback is by far the most difficult position to analyze on the field. In the past, measures of passing yards, completions, completion percentage, touchdowns, and interceptions were how we graded a quarterback’s play. This evolved into a quarterback rating, which is essentially a combination of these measures at its core. ESPN eventually advanced the rating into the QBR, which incorporates situation-specific corrections.
The QBR has been a giant step forward in determining a quarterback’s true value. A huge hurdle remaining is differentiating the effect that a receiving corps have on a quarterback’s success. Using premium statistics from ProFootballFocus.com, I explore tendencies and patterns between quarterbacks and receivers.
This data-set includes quarterback and wide-receiver statistics from 2010-2014. To begin, I looked at how often a quarterback favored a certain receiver. Jay Cutler targeted Brandon Marshall 33% of the routes he ran during the 2012 season, the most during this study period. The top 10 quarterback-receiver tandems based on targets per route run can be seen below:
I was also interested in how accurate a quarterback was throwing to a certain receiver. Passing completion percentage is used in nearly every quarterback rating metric, but what effect does a particular receiver have on a quarterback’s ability to throw a catchable ball? For instance, if a receiver comes out of break and creates separation more effectively than others, does this allow the quarterback a more catchable throw (i.e. either a catch or drop vs a missed throw, batted ball, etc.)?
My interest in analytics began with the NFL Combine, so naturally I wondered whether short-area quickness improves separation, and whether improved separation make passing easier. The top 10 QB-WR duos in terms of catchable-balls thrown per target, can be seen below. Of these top 10 receivers, all but one were 5’11” or shorter. Edelman and Welker had pro-day short-shuttle times of 3.91 and 4.01 seconds respectively, which would have put them in the 98th and 90th percentiles in NFL Combine history for wide-receivers (that is saying A LOT). Cobb, Welker, Edelman, and Amendola all played multiple positions in college, whether it was receiver, returner, or even quarterback. So although these top 10 duos are made up of elite QBs (Rodgers, Brady, and Manning), there are also patterns in the athleticism of the receivers.
I was also interested in which receiver had the most reliable hands. During this 5 season period (2010-2014), some receivers were extraordinarily consistent in catching balls. Players known for making amazing acrobatic catches like Dez Bryant, Fitzgerald, and Odell Beckham Jr., were also the least likely to drop routine catchable balls.
What I learned was that if a ball is “catchable” (i.e. either caught or dropped), it is almost always caught by elite receivers. But, how often do quarterbacks throw “catchable balls” to their receivers? And does a quarterback’s completion percentage reflect exactly how good (or bad) he is at getting the ball in the hands of his receiver?
The figure below shows that this is not necessarily the case. Surveying the leading passers for each NFL team over the last 5 years, the QB’s completion percentage did a rather poor job of predicting how truly accurate he was at throwing catchable balls to receivers (r2 = 0.34. Interestingly, elite passers like Brady and Rodgers led quarterbacks in the percentage of catchable balls thrown per WR target (77% and 75% respectively), exceeding their completion % by a large margin (63% and 67% respectively).
I next explored how reliable receivers influenced a quarterback’s rating. The answer was not too much (r2 = 0.18). Cleveland had the least reliable receivers, catching just over 87% of the catchable balls thrown.
Lastly, I found that catchable-balls per target was not reflected as well as completion percentage was in a quarterback’s rating (NFL, not ESPN QBR). While completion percentage explained roughly 75% of the difference among NFL QB’s ratings, catchable-balls thrown per WR target only accounted for 53%. Higher percentage throws to TEs and RBs may inflate completion percentage and may lead to a quarterback being over-valued.
ProFootballFocus.com, PFF, established its own quarterback metrics, the PFF QB Rating and the PFF Accuracy Percentage, which accounts for spikes, throw aways, drops, and yards after the catch. I believe this metric is a huge step forward in separating receiver influence from quarterback performance. The figure bellows helps illustrate how completion percentage relates to the PFF Accuracy Percentage, in regards to drop rate and the proportion of passing yards that come solely through air.
Creating a metric to describe the performance and value of a quarterback is an inexact science that is plagued by a myriad of co-limiting variables. QB analytics are often criticized for their weak ability to predict trends. However, analytics did choose Teddy Bridgewater over Johnny Manziel, and Marcus Mariota over Jameis Winston. I believe it will just take time for statistical measures to catch up to the true needs for a QB analysis. There is a saying in environmental chemistry: “You only find what you are looking for. And you only find it if your instrument can measure it.” Groups like Profootballfocus.com are changing the way football is measured, and I feel they are paving the way in re-thinking on field metrics.