Scrambling for a Quarterback: Observations from 10 Years of Drafting

Overview

So, while I was working on evaluating Parcells for the last piece, I was challenging myself to see if any actual trend showed up in the data itself. I wanted to know if I could come up with a checklist that was any better than the one offered by Parcells. Of course, this meant I was going to need a working definition of “good professional quarterback.” I decided to set a high bar and a low bar. For the high bar, I wanted the quarterback to have three traits—at least one Pro Bowl, a career ANY/A higher than Jay Cutler’s 5.81, and a median indexed passer rating higher than 100 (which is another way of saying that I wanted the player’s median passer rating to have been at least average for the league when he recorded it). This would make the quarterback in question better than anyone I have seen play for the Bears, which is the hope.

This would give me Patrick Mahomes, Russell Wilson, Deshaun Watson, Kirk Cousins, Lamar Jackson, Dak Prescott, Matthew Stafford, Andrew Luck, Derek Carr, Ryan Tannehill, Carson Wentz, and Tyrod Taylor.

For the low bar, I wanted the player to have started at least 40 games or (for players drafted where this would be lower) in at least 60% of the games available to them during their career. I also wanted this player to have a career ANY/A at least as high as Mitchell Trubisky’s 5.64 and a median indexed passer rating higher than 90. This would make the player a low-end starter. Interestingly, this second pool was largely defined by the “not quite” group rejected by the first pool. It is a list of guys who seem like fringe starters, has-beens, and never-will-bes. What disqualified them from the first pool is listed in parentheses:

Baker Mayfield (Pro Bowl), Marcus Mariota (Pro Bowl), Colin Kaepernick (Pro Bowl), Teddy Bridgewater (MIR), Andy Dalton (MIR), Jarod Goff (MIR), Cam Newton (MIR), Jameis Winston (MIR), Josh Allen (MIR), Robert Griffin III (ANY/A and MIR), and Mitchell Trubusky (ANY/A and MIR). With the exception of Allen and Mayfield—who has been playing in the stacked AFC and might have earned a Pro Bowl nod in the NFC—they largely represent quarterback purgatory—better than a team is likely to find to replace them, but not good (at least, not across their whole careers).

Observation #1: Early Isn’t Better

On average, the “good” quarterbacks were drafted 50th overall (but with a median draft position of 22nd). Likewise, they averaged being the 4th or 5th quarterback taken (again with their median position being the 3rd or 4th quarterback drafted). In short, these players were almost as likely to be taken after the first round (5) than in the first round (7). With only a couple of exceptions, these were not the anointed ones. They were the players who had potential and who—in all honesty—were able to land in pretty good situations.

Interestingly, the “marginal” quarterbacks were drafted 17th overall on average (with a median draft position of 2nd overall) and actually averaged being the 2nd or 3rd quarterback to be drafted (median position being the 2nd quarterback drafted). Seven of these players were taken with one of the first two picks in the draft, and five of them were the first quarterback drafted overall. Only one of these players was taken outside of the first 40 picks (Nick Foles), and nine of them were taken in the first round. This could mean that these players would have been better had they gone to a better team, but it could also be interpreted as evidence that they were marginal players who kept getting chances to improve because of their draft status. The reality is likely that they are a mix.

Observation #2: One Measure Can’t Rule Them All

The one thing that I had noticed (and it began as a joke) was that simply seeing how many yards a player had thrown for in college had just as strong of a correlation with NFL success as any test Parcells came up with. This was a low and meaningless correlation, but it was still enough of an observation to note that total passing yards did a pretty good job of capturing games played (0.64), and that correlation went up (0.69) when I multiplied the games played by the completion percentage. It wasn’t perfect, but it did mean that one number was able to capture a rough sense of how many games a player had been in, how often he completed his passes, and how good those passes were in the first place. On its own, however, it was a terrible predictor of NFL success (below 0.20 on almost any measure).

Parcells had a point, at least holistically, when he asked that a quarterback have at least a 2:1 touchdown-to-interception ratio. One thing that the “good” and “marginal” lists above have in common is that all of them except Matthew Stafford did at least have a 2:1 touchdown-to-interception ratio in college. However, even adding a penalty for interceptions (of various sizes) does little to increase the correlation. Thus, while it is possible to use total yards in some measure to guess about how much playing time a college quarterback enjoyed, it does little on its own to predict how well that playing time is going to equate into professional success. Call it the Dan Lefevour problem, if you will.

Still, this measure is worth considering in a broader context, but more on that later.

Observation #3: Losing Yards Matters

There are more ways to pick up yards in football besides simply tossing the ball. Especially in college, good quarterbacks make plays with their legs. Just as importantly, college records indicate sacks as negative run plays. This means that quarterbacks who have trouble avoiding pressure will end up with lower (and in some cases negative) running yards. For example, from 2009-2018, three quarterbacks drafted in the first round had negative rushing totals in college: Josh Rosen, Brandon Weedon, and Jared Goff. That sounds right. The inability of those players to avoid pressure certainly contributed to their struggles, and while Goff eventually reached two Pro Bowl nods, his first year saw him give up sacks on more than 11% of pass plays.

Other players who went on to at least have a season as a primary starter who nonetheless had negative rushing yard totals in college include Mike Glennon, Nick Foles, Zach Mettenberger, Tom Savage, Jimmy Clausen, Kirk Cousins, Matt Barkley, Trevor Sieman, and Cody Kessler. Wow. Yeah. Of that whole group (including the first-rounders), Cousins is the only one with a median indexed passer rating above 100 (or average for the league at the time). By and large, these are not good players, let alone franchise quarterbacks.

Observation #4: Interceptions Matter, Too

Good players do throw interceptions. In fact, they are simply part of the game. However, good players limit those interceptions, and a player who gets picked off a lot in college is likely to struggle just as much—if not more—with the same problem in the NFL. Looking at the quarterbacks who were drafted, the following players all averaged more than one interception per game in college: Nathan Enderle, Matt Barkley, Christian Hackenberg, Tanner Lee, Sean Mannion, Zac Dysert, Mike Teel, Jameis Winston, Josh Freeman, T.J. Yates, and Landry Jones.

By contrast, a player who never puts the ball in danger might be too conservative, or might be playing within a system that just doesn’t require them to make pro-style plays. Players who didn’t have at least one interception in every three games include A.J. McCarron, Greg McElroy, Cardale Jones, Tim Tebow, Bryce Petty, Mitchell Trubisky, and Stephen McGee. Carson Wentz and B.J. Coleman sit exactly at the 33.3% mark.

Scrambling for an Answer

Playing around with the numbers, my rough cut is that I want a player with at least 9000 passing yards, an average of at least 20 rushing yards per game, and an interception rate between .33 and .67 per game.

Those criteria cut the entire pool of 112 players down to ten: Baker Mayfield, Russell Wilson, Marcus Mariota, Robert Griffin III, Andy Dalton, Colin Kaepernick, Brett Hundley, Andrew Luck, Kevin Hogan, and Dak Prescott. Seven of those ten have career ANY/As above Jay Cutler’s, and six of those seven also have a median indexed passer rating above average. Five are pro bowlers, including Robert Griffin III (who isn’t one of the six, and whose career was unquestionably affected by injury). In fact, Kevin Hogan and Brett Hundley are the only clear duds. Obviously, a number of fantastic players are missed. Still, as a conservative approach to narrow things down, this isn’t a bad check.

Of the 2021 prospects, this list would include Trevor Lawrence, Kellen Mond, and Sam Ehlinger, with Ian Book 52 yards shy of the mark. Meanwhile, Zach Wilson, Justin Fields, and Mac Jones don’t have enough passing yards, Kyle Trask is short on passing yards and rushing yards, Davis Mills essentially doesn’t have the stats at at all, and Shane Buechele has thrown too many interceptions without rushing for enough yards.

Does this mean that Mond will play better than Jones? Not at all. However, it is worth noting that the best quarterbacks are often not the ones taken the earliest, that decent to good prospects can be found later on, and that there are players who might be available later on who (on paper) look promising.

Be sure you check out Josh’s previous entries in his Scrambling for a Quarterback series.

---

Responses

EJ Snyder: Nice work. I appreciate the grind.

I especially like the correlation to (negative) rushing yards and INT’s. Nice middle ground for taking shots with the best hit rate in the “best of the rest” pool of QB’s where value can be found. This puts some parameters around that and while it’s not perfect, nothing is. So it’s better than that.

Patti Curl: When I look at Josh’s results, I can’t argue that his criteria seem to identify the type of quarterback that’s successful in today’s NFL better than Parcells’ do. I do have some pause when I see them applied to this year’s class, though. Ruling out Fields (my number 2) Wilson and Mack due to too few passing yards is rough. I get that a cumulative production criteria is valuable, but as Josh argued himself, it can penalize people for having enough success to declare early.

Trey Lance would have been excluded anyway because his interception percentage is too low. A controversial criteria but one that I agree with on principle. This could also, in a way, be a product of Lance’s low sample size, since there was a large amount of luck involved in his zero interception season including several balls that could/woulda/shoulda been intercepted on a different day with a different butterfly flapping her wings.

My favorite takeaway from these results is that they finalize my decision that I prefer Kellen Mond over Davis Mills as a day two QB. Thanks for the nudge I needed!