A little background on me: Aside from being a big Jets fan, I'm a college student majoring in math. I've wanted to try applying some analytics to football related questions, and I was curious what the data had to say on QB prospects of various ages. So once I got a little free time (yes, I have a life) I did this analysis and wrote this article summarizing my findings. It was an interesting project. Hopefully Macc's staff does this sort of thing; I believe that it would really help us in the draft process. If not, I'm currently open for hire lol. Hope you enjoy the article! How Old is the Ideal QB Prospect? By HomeoftheJets In each of his first two drafts, Jets GM Mike Maccagnan has drafted a quarterback: Bryce Petty in 2015 and Christian Hackenberg in 2016. One significant difference between the two prospects is that on Draft Day, Petty was a month from turning 24 whereas Hackenberg had just turned 21. In 2015, analysts like Gary Davenport pegged Petty’s age as a red flag (http://bleacherreport.com/articles/...tty-isnt-worth-reaching-for-in-2015-nfl-draft). On the other hand, a year later, other analysts like Bob Sturm, called up Hackenberg’s age in his defense (http://sportsday.dallasnews.com/dal...rious-case-penn-state-qb-christian-hackenberg). But is it really true that it’s better to draft younger QBs than older ones? To find out, I pulled up 24 years of data from Pro Football Reference, specifically their Career AV (Approximate Value) stat, which measures the approximate contribution a player makes over the course of his career. Since analytics is all about using data to gain a competitive advantage over other teams, I calculated whether teams generally over or underdraft QBs of various ages. To do so, I compiled a Draft Value Chart for every pick from 1 to 224 during the period from 1992 to 2015 and compared the benchmark value for each pick with the actual value each QB prospect produced. I used Chase Stuart’s method for compiling my DVC, which can be found here. http://www.pro-football-reference.com/blog/?p=527 The first round of my DVC is shown below. If you want to find expected AVs for all picks, use the equation AV = -11.537 * ln(pick number) + 67.294, where ln is the natural logarithm. Pick: AV 1: 67 2: 59 3: 55 4: 51 5: 49 6: 47 7: 45 8: 43 9: 42 10: 41 11: 40 12: 39 13: 38 14: 37 15: 36 16: 35 17: 35 18: 34 19: 33 20: 33 21: 32 22: 32 23: 31 24: 31 25: 30 26: 30 27: 29 28: 29 29: 28 30: 28 31: 28 32: 27 Although more recently drafted QBs have lower Career AVs because they haven’t finished their careers, my DVC suits my purpose because I’m looking to evaluate QB ages in general rather than specific prospects. For example, Jameis Winston’s Career AV is an artificially low 13 (according to my DVC, a #1 overall pick should produce a Career AV of about 67) because he’s only played one season. While Winston may unfairly make 21 year old prospects look bad, Tommy Maddox for instance unfairly makes 21 year old prospects look good. That’s because Maddox got to play out his entire career unlike many of the QBs under consideration. Theoretically these error-inducing effects should cancel each other out. Ultimately, I got data for 240 QBs. Of these, there was one 29 year old, one 28 year old, no 27 year olds, one 26 year old, three 25 year olds, and many 24, 23, 22, and 21 year olds. Once again, to find whether teams over or underdraft QBs based on age, I compared each player’s Career AV to what was expected of him based on his draft slot. For example, the 29 year old, Chris Weinke, was selected with the 106th pick. According to my DVC, the typical 106th pick produces a Career AV of about 13, but Weinke only produced a Career AV of 9. Therefore, on average, 29 year old prospects fall short of their projected Career AV by about 4 points. Here are my findings: Age: Difference 29: -4 28: -21 26: -23 25: -6 24: -8 23: -1 22: 1 21: -3 All QBs: -2 The sample sizes for QBs older than 24 were too small to conclude anything meaningful, but it is worth noting that all 6 of those QBs underperformed. Meanwhile, we can see that on average, 24 year old QBs fall notably short of expectations. Interestingly, 21 year olds fall short as well though not nearly as badly. The ideal age seems to be 22; we can see that teams usually wait too long to draft 22 year old QB prospects. As a result, these prospects exceed expectations. Ultimately, these results make sense. Many 24 year old QB picks have peaked by the time they enter the draft, and so they fail to improve to the necessary level to be a quality NFL QB. Jets fans may be seeing this problem now with Petty, who so far has not impressed in OTAs. On the other hand, 21 year old picks are more likely to have played only a season at the college level, meaning that their performance may have been somewhat of a fluke. Fortunately for the Jets, Hackenberg was a three year starter in college, so his youth may not work against him. Finally, it’s worth noting that regardless of age, QB prospects on average fail to meet expectations by about 2 points of Career AV. Again, this makes sense because teams are more prone to reach on QBs due to their outsized impact on team performance. My conclusion is this: if you’re a GM and you fall in love with an older QB prospect (or a really young one, but especially an older one), realize that the data are not in your favor. Don’t discount taking the prospect entirely as age is only a small part of the evaluation process. But be sure view his age as a red flag and factor that into your decision making. One last note: The key takeaway from this analysis is not just that 22 year old QBs are better than 24 year old QBs. I think historically most GMs have realized that. Rather, it's that teams have consistently underestimated how much better 22 year olds are than 24 year olds.
The best of the best, the cream of the crop QBs usually leave for the NFL draft a year early as juniors......That would be a big reason why 22 years old, or somewhere in the middle is why it works out that way Just a guess
Approximate value. It measures how much a player contributes to his team. It isn't perfect for individual players but it's very useful for large sample sizes. Thanks for noting the omission, OP now edited.
The interesting part is not just that 22 year olds are the best in terms of ability. It's that even knowing that, teams still undervalue 22 year olds. 22 year olds who deserve to go higher sit there while teams reach for less effective 24 year olds.
Going back to 1992 might be a mistake in light of the 2011 collective bargaining rule that drastically altered the cap hip for rookies.
If this were my job, I would repeat the procedure for every conceivable factor. I would then prepare a report for the GM telling him which traits are commonly undervalued and which are overvalued. Since it isn't, I only did age because it's relevant to the Jets right now.
It may or may not be relevant now, it's just the one you picked. Maybe the secret is in shoe size or arm length. If you're going to get analytical you have to choose the right variable or combination thereof.
There is no secret variable. You look at all the variables and see if you can get information about them that you factor into your drafting process. If I had all the time in the world, I would do just that, but I don't. So I picked age because I'm posting on a Jets forum, and we just picked QBs in back to back drafts on opposite ends of the age spectrum.
It might be, but I need a sample size of more than five recent drafts. I picked 1992 as the first year because that's when free agency started. Also, even now teams usually draft QBs in the hopes that they'll find a franchise guy rather than looking for a few years of cheap production. Other positions are different, which is why I believe Chase Stuart came up with a new value chart based on only the first five years of production.
I'm just going to throw something out here for you to think about - maybe the years of organized football is more important than whether a guy is 22 or 24. One guy who played peewee could have 15 years of coaching by then, another might have only started in his junior year in high school and only have five. Who's going to be the better player? Hard to tell. The kid with 15 years might be just about as good as he can get the other one might benefit from a lot more work. The two years of actual age difference may mean nothing. That's why I'm glad you're the one with these exercises, not me.
You're absolutely right about the coaching. Football is really tricky to analyze because there are so many factors that go into evaluating players. In any case, your point gives me hope about Hack. He's only 21, which ordinarily seems to be a minus, but he's had so much experience for his age.
I would think that's because they think the 24 year olds are more mature emotionally/mentally, but it is still surprising, because even if they pan out, with everything else being equal, they probably will have two less years in their careers.
22 year olds are likely to have more starting experience in college than younger prospects. It makes sense that they work out more often. There are two factors that are important to sort out in figuring all of this out. The first is the effect of free agency on all of this. Your study does that by starting in 1992. The other factor that will become more important is the hot Junior QB coming out factor, which is much more recent. If all the QB's who are likely to be great in the NFL come out in their Junior years moving forward, well that's likely to reduce the best age by a year.
