by Bill Connelly
Though I guess we all are to some degree, I've always been a bit of a thief, one to try to take advantage of the work others have done.
When I was first diving into college football play-by-play data (and was not yet working with Football Outsiders), I pilfered the idea of success rate and adjusted it to fit NCAA standards. And once I was having fun with that efficiency measure, I decided the best approach was to tinker with ideas for explosiveness (which eventually became PPP) to make something like an OPS for football.
Last year, when I began to realize that I didn't like the redundancy involved in PPP, I gravitated toward baseball's isolated power concept and adjusted my ratings to only look at the explosiveness of successful plays.
Et cetera. If another sport has a good idea, I want to figure out how to use it for football.
That brings me to radars. The work of StatsBomb's Ted Knutson, among others, has made me appreciate the way that radars can combine two questions: what kind of player are you, and how good are you at being that kind of player?
With complete charting data (something that is incredibly cost-prohibitive), radars could take on extra descriptive value and significance. But with simple play-by-play data, I think they could still be of use.
Naturally, my first application of this type of graph comes at the quarterback position. I'll post an example first, then dive into what it might mean. Here's TCU's dynamic Trevone Boykin.
There are six categories of measures here: yards per completion, completion rate, sack avoidance (sack rates), interception avoidance (interception rates), run frequency (your ratio of non-sack rushes to pass attempts), and yards per carry. I decided to present these things in terms of percentiles* -- where you do measure in each item compared to other quarterbacks?
* Percentiles themselves are based only on quarterbacks with at least 40 combined passes and rushes.
(Once I've settled on something permanent for this design, I'll go about adding data labels for each category -- i.e., Boykin is averaging 14.4 yards per completion, has a completion rate of 65.9 percent, a sack rate of 2.5 percent, and an interception rate of 2.5 percent, and is averaging 6.5 yards per carry.)
In a way, the optics of this chart favor dual-threat quarterbacks, simply because if your run frequency is low, you're guaranteed to have one low spot on the chart. But since this is intended to look not only at if you're good, but also at what kind of quarterback you actually are, I felt it was important to include it.
Boykin's chart reveals a player who doesn't quite fit the dual-threat stereotype. Boykin is athletic enough to have played wide receiver for a bit, but we see here that he doesn't actually run that frequently compared to other college quarterbacks -- 98 rushes is quite a bit, but not necessarily compared to 364 pass attempts -- and he avoids sacks quite a bit more than other players who are reliant on their legs.
In all, Boykin has just been tremendous this year. His interception rate obviously perked up against Oklahoma State, but he's still filling in quite a bit of space on this chart.
Of course, looking at only one quarterback will only tell so much of a tale. Here's how Cal's Jared Goff, one of the most highly projected quarterbacks in this coming Draft, compares.
Goff, whose footwork is about as pretty as I've ever seen in person, is an efficiency guy in Sonny Dykes' version of the air raid offense. He's completing 64 percent of his passes this year, which fits him into the 76th percentile (I realize using percentiles to look at percentages is begging for confusion -- will keep tinkering with better ways to set this up), and he's reasonably decent at avoiding sacks and picks. He doesn't run much, but when he does, he picks up quite a few yards.
What about a guy who has been projected as a pretty high draft pick since high school ... but might still need a little more work?
Hackenberg is dealing with an offensive line that is still pretty spotty and a receiving corps that is still young (his top three targets are sophomores). He's getting sacked more than 10 percent of the time, and his completion rate is a paltry 54 percent. You can kind of see here what he wants to be -- he looks downfield a decent amount but makes mostly safe passes and isn't afraid to hang in the pocket. But if he were to declare for the draft and get selected in the first round or two, his selection would be on faith and potential alone. He still hasn't had that many successful games as a college quarterback.
Connor Cook is another player whose name has been associated with solid draft status for a while. He combines minimal mistakes with an eye downfield. His numbers haven't been amazing this year (State ranks only 29th in Passing S&P+), but a lot of that is probably coming from the fact that the Spartans' run game has been dismal.
A Mark Dantonio team will likely always have a run-first offense, but State ranks only 112th in Rushing S&P+ and, consequently, 105th in Standard Downs S&P+. Cook and a couple of nice receivers (Aaron Burbridge and Macgarrett Kings Jr.) have allowed them to recover nicely (13th in Passing Downs S&P+), but it appears the only goal of the Michigan State run game is to work some clock. It's not actually getting the Spartans anywhere.
Quick side-by-side comparison:
Cardale Jones' decision to return to school after last year's "third-stringer to national champion" run late in the season has backfired a bit. In the preseason, the thought was that he only needed more reps to prove himself to NFL scouts. Instead, more reps have exposed some flaws.
[ad placeholder 3]
Jones has had a ... neutral year. Ohio State is still undefeated (and Jones is still undefeated as a starter), but his radar here suggests he hasn't actually been good or bad at anything. He's basically smack in the middle of every category.
Ohio State's other starting quarterback, J.T. Barrett (who overtook Jones for the No. 1 post until a recent drunk driving arrest and suspension), is a wonderful runner -- one of the best in college football at the position -- and a steady, conservative passer. He doesn't take sacks, and he completes most of his passes, even if they don't really go anywhere. The Buckeyes' offense found more of a rhythm with Barrett behind center, and these charts pretty perfectly show the differences between the two quarterbacks.
We'll finish with one of the most confusing (to me) quarterbacks in college football.
[ad placeholder 4]
Josh Dobbs has spent a good portion of this season handcuffed behind the line and looking to pass instead of utilizing his legs. Including sacks, he averaged about nine carries for 29 yards per game in early contests against Bowling Green, Oklahoma, Western Carolina, and Arkansas. But in some of the Vols' better performances of the season -- a narrow loss to Florida, a win over Georgia, and a romp over Kentucky -- he averaged 14 carries and 69 yards.
The more Dobbs runs, the better Tennessee seems to perform on average, but a dire depth chart situation (there are nothing but freshmen behind Dobbs) and perhaps an identity issue at offensive coordinator (Mike Debord has spent much of the last 25 years as either an NFL assistant or part of a pro-style, quarterbacks-as-statues offense at Michigan) have kept him reined in at times.
Feedback is welcome! I wish I had more categories -- air yards, for instance -- but the lack of charting data makes that impossible.