Completion Percentage Is Flawed

As I watch Oklahoma face off against Clemson in the Russel Athletic Bowl (it feels wrong to say the name of a sponsor), and face off is a misnomer in this case due to the lack of competitiveness so far, I see Cole Stoudt, the Clemson quarterback, complete wide receiver screens ad nauseam against the Oklahoma defense.

In this particular case, Oklahoma’s defensive backs can’t make the adjustments at the line of scrimmage to be able to neutralize the Clemson wide receivers.

In other cases in general, the deliberate use of screen passes, dump offs to half backs, and check downs are used as more out of necessity, due to the limitations of a team’s quarterback, than as a strategic move.

Anyone who saw the Dolphins and Ryan Tannehill play this year could have had the same sentiments after they watched the cautious approach the Dolphins took with their young signal caller and then looked at his numbers at the end of the day.

Tannehill was able to conclude the season with the league’s fifth ranked completion percentage—all stats given in this article will be based off of quarterbacks that took fifty percent of their team’s snaps (27 in total)—but finished 21^st in average yards per attempt (6.88 YPA) as the Dolphins took a judicious approach with their young talent.

This isn’t to say that Tannehill didn’t have a good year, or that he won’t eventually develop into a good quarterback. It just points out one of the ways that completion percentage is flawed.

This also isn’t to suggest that the aforementioned statement is ground breaking news, but, as the season comes to a close, I thought we could use this as a prompt to reflect on ways that we can improve completion percentage.

Awards will be given out soon, and I’m not sure if these people are common or if the few people that I have heard convey this sentiment have just stuck in my head, but some have clamored for the recognition, if they’ve felt compelled enough to support his candidacy with such vigor, of Tony Romo and the possibility that he could be 2014 NFL MVP.

These advocates point to Romo’s completion percentage: 1^st in the NFL. They hold up his touchdowns: 4^th. Dallas won the NFC East. The Cowboys finished with their best record since 2007. All is right with the world. Praise Tony Romo. Praise the Cowboys. Praise Dallas.

That’s what some could/would/might/do say.

Don’t get me wrong. I can empathize with the Romo apologists. Broken ribs: he plays. Broken back: he plays. For those that feel it is necessary, of which I am not one, to characterize football players with the epithet of “warriors”, Tony Romo should, for those people that feel the need to do so, almost always be first on that list. The subject of whether or not he should play with those injuries is another conversation.

But (of course there’s a but), Romo was 23^rd, out of the 27 quarterbacks that took fifty percent of their team’s snaps, in pass attempts, which makes it much easier to sustain a high completion percentage, and accumulated only 3,705 yards—14^th out of 27.

So while Romo was able to accumulate the leagues best completion percentage, his completion percentage is unable to account for the fact that Drew Brees, who was second in completion percentage, attempted 224 more passes and threw for over 1,200 more yards.

The amount of attempts a quarterback has is another way that completion percentage as a stat is flawed.

So far we’ve identified two ways that completion percentage is flawed (i.e. it doesn’t take into consideration attempts and yards per attempt).

And in an effort to improve completion percentage, I’ve come up with tCOMP%—true completion percentage.

tCOMP% weights a quarterback’s completion percentage off their yards per attempt and their number of attempts.

To take it a step further, I’ve also taken out the number of dropped passes and spikes that each quarterback had in 2014 from their total number of attempts to get a more accurate number (in the chart below as aCOMP%), as opposed to COMP%, to weigh their tCOMP% off of.

[table id=82 /]

aATT is the amount of attempts a quarterback had minus their dropped passes and spikes.

tCOMP+ shows how much better or worse a quarterbacks true completion percentage is than league average—.67%— where 100 is average. For example Ben Roethlisberger has a 141 tCOMP+, which means his true completion percentage is 41 percent better than league average.

The biggest mover on this list is Andrew Luck, and anyone that saw Luck play this year would agree, as his tCOMP+ does, that his 19^th ranked completion percentage is a mirage of ineptitude.

This metric is by no means perfect; for one, it doesn’t take into account interceptions (Drew Brees and Matt Ryan are respectively 1^st and 4^th on this list, but Drew Brees was also tied for 3^rd in the NFL in interceptions, while Matt Ryan was tied for eighth in interceptions).

But hey, completion percentage doesn’t take interceptions into account either.

The point is that we’ve worked towards a more accurate depiction of a player’s performance relative to their peers.

What else should be considered?

Devin Jordan is obsessed with statistical analysis, non-fiction literature, and electronic music. If you enjoyed reading him, follow him on Twitter @devinjjordan.