Over a decade ago, well before I was on staff there, Football Outsiders published David Lewin’s fascinating research, that the eventual fate of NFL quarterbacks could be predicted with reasonable accuracy knowing simply that (a) they were selected in the first two rounds of the draft, (b) how many games they started in college, and (b) their collegiate completion percentage.
To the extent this relationship existed, it completely broke down probably before Dave’s work was even published. I don’t want to rehash his work or the ideas, but simply note one of the maybe not so obvious implications of his work: fewer quarterbacks should be drafted in the first two rounds. The ones that started many games and completed a high percentage of their passes, go on and keep drafting those. But the Akili Smiths and Jim Druckenmillers of the world who didn’t start many games or complete many passes, don’t take those.
The potential further implication of that research would be that players with many starts and a good completion percentage were good bets. Assuming teams would draft quarterbacks at roughly similar rates, fill the void created by not taking the Smiths and Druckenmillers with those guys. Even at the time Lewin’s work existed, though, those same qualifiers didn’t predict the success of quarterbacks drafted after the first two rounds, thus my addition of point (a) to research results that were largely described as consisting of (b) and (c). This research, then, depended further on NFL teams not realizing the relationship and responding accordingly.
This may seem like a rather abstruse, potentially recondite way of starting off an explanation, but as I think back on over a decade of thinking about the NFL draft, there are essentially two types of heuristics for finding good players, or at least not taking bad ones:
(1) Ones that, like Lewin’s research, inevitably break down over time, whether naturally in response to other changes, by NFL teams becoming aware of it, or both; and
(2) Ones that do not necessarily produce solutions, essentially telling even teams picking high in the draft they don’t have good options.
From that perspective, that “analytics” is considered a dirty word in NFL circles doesn’t seem too surprising. NFL teams have to draft players with the picks they have now, and waiting until process type (2) tells you there’s a player you should pick there isn’t always an option (see, e.g., Sashi Brown’s fate in Cleveland). The answer, then, may be to use flavor du jour (1) to help identify which prospects are higher risk in general and which risk factors you have to use.
Putting this in context of an actual player discussion, any evaluation of Carson Wentz coming out of North Dakota State would have said that there are few enough potential high draft pick QBs coming out of I-AA/FCS there will be an element of irreducible risk to the basic analysis. A team that stuck to type (2) because making a mistake at QB gets you fired would not have drafted Wentz. As a decision-maker, you just had to get past that to select Wentz. Philadelphia did and reaped the rewards, but plenty of other teams have ignored other risk factors, whether analytical or more irreducible, and gotten burned by it.
But the draft is just one of the various methods of building a football team. It has particular characteristics. Since I’m talking quarterbacks here, the most important one is that all contracts are extremely cost-controlled relative to the free agent price. We haven’t seen any “true” quarterback contracts in an extremely long time; barring something shocking, Kirk Cousins will be the first quarterback this millennium hitting free agency after a couple good seasons as a starter at a reasonable age not coming off a major injury. How much he makes will not be a product of limited cap room, his rookie deal (still important for Matt Stafford, the last QB still directly benefiting from being a high pick under the old CBA), or negotiating in the shadow of the franchise tag, which normally dominate discussion.
But even with those limitations, we know what quarterbacks get paid, and it’s a lot. Just to enter into the game, you’re looking at a minimum of $15 million for a Mike Glennon, while Carson Wentz’s cap hit in 2017, as the #2 overall pick, was just over $6 million. That difference is basically equal to a starting pass rusher, and it becomes a top 10-type pass rusher if you look at the split between even a high pick rookie like Wentz and the good part of the starting quarterback market. If the quarterback you take ends up as even an average starter, you’re much better off picking him in the draft even before you account for the fact that Cousins aside, good starting QBs without red flags don’t hit the market.
It therefore makes sense, especially in the current environment, to draft a quarterback even if the expected hit rate is much higher for a different position. Which, I know, is basically the opposite conclusion any sort of type (2) process like Lewin’s suggests. If you’re thinking this sounds like a great recipe for muddled thinking, you’re not alone; conceptually, “don’t draft bad quarterback prospects early” and “draft a quarterback prospect early if you think there’s a chance he’s good” are both extremely justifiable maxims.
And let’s look at the quarterbacks drafted in first round in the past 10 drafts a holdover head coach:
|Year||QB||Head Coach Fate|
|2010||Bradford||fired after Year 2|
|2010||Tebow||fired in Year 1|
|2011||Gabbert||fired in Year 1|
|2012||RG3||fired after Year 2|
|2012||Weeden||fired after Year 1|
|2014||Bortles||fired after Year 3|
|2015||Winston||fired after Year 1|
|2015||Mariota||fired in Year 1|
|2016||Goff||fired in Year 1|
|2016||Lynch||left after Year 1|
|2017||Trubisky||fired after Year 1|
We’ll see what happens with Bill O’Brien, who seems to have just won a power struggle, and Andy Reid, but those are pretty dismal results. More coaches were fired in the first season after drafting a quarterback than got to coach him in a second season, and the only one to make it to a third season was Gus Bradley.
There are a few obvious responses to this, like these teams were bad, so it’s no surprise the teams were fired. This suggestion only goes so far; a number of the quarterbacks went to teams that weren’t that bad. The Broncos (Tebow), the Jaguars (Gabbert), and the Rams (Goff), to name three, had natural picks in the middle of the first round and chose their new signal-caller only after making a trade. The dynamic highlighted by an old Bill Walsh quote probably played a bigger role; the owner sees the pick, the GM sees he chose good players, and the team isn’t winning games so it must be the coach’s fault. A new coach (I didn’t track these the same way, but many of them survived much longer than holdover head coaches) doesn’t carry the same baggage in terms of judging team performance before v. after picking the shiny new QB who is supposed to solve all these problems.
That may feel like, and have been, a bit of a digression from what felt like the main point, but I think it goes to stress what I see as the underlying dynamic, that quarterback success requires (1) that there be a team around him, because one position can only do so much in a team game, and (2) organizational clarity is a must. Somebody, including Doug Pederson, had to push for Carson Wentz, to get past the irreducible risk, and then Pederson had to be the one to put him in the best position to succeed. The path of development is rarely absolutely straight and up, even for ultimately successful players (see, e.g., Wentz’s 2016 season, with a passing DVOA of 26.6% through three weeks and -19.3% thereafter), and a united organization may be some of the difference between bumps that are straightened out (Wentz’s 2017) and players plateauing and ultimately failing. In a world of deeply conflicting maxims, without the size needed to create impersonal processes, that’s the way it has to be.
But, please, for us stat guys, stop drafting guys early thinking you can convert erratic general accuracy into specific accuracy without a specific, easy to fix flaw. Know what the risks are, and have a plan to fix or mitigate them. Or, as I like to put it, tell me what a player can’t do and then tell me why I should draft him anyway.
This post was inspired by the post Thoughts About Drafting Quarterbacks by Chris B. Brown at Smart Football. I wrote a brief draft of this post before the 2017 NFL season began, got stuck, and abandoned it for a while. As I wrote the bulk of this post it mostly became an excuse for me to write what I wanted to write about rather than a post serving as commentary or meta-commentary on Chris’s post, which has thoughts on base rates and the value of the Mitchell Trubisky trade that are largely orthogonal to what I wrote above, which may speak more to his point #5 at the end.