Drafting Quarterbacks

by Kevin Meers

Using my previous analysis of career approximate value (CAV) on how to value draft picks, this study analyzes how the results apply to the quarterback position. As the five-month debate on what the Colts should do viz. Peyton Manning and Andrew Luck is just beginning, these findings can help inform the Colts’ decision. As with the entire draft, quarterbacks selected first among quarterbacks and earlier in the draft have more expected upside and more expected value. The following analysis looks at the quarterback position in terms of both order drafted and overall pick.

Let’s first look at quarterbacks by order drafted. Both the upside and expected value of the first overall pick stand out here. The upside of drafting the first overall quarterback taken is two and a half times greater than the downside; these both fall very rapidly as more quarterbacks are drafted. Here’s the graph:

The median is a much better metric to use here because there are so many outliers that skew the mean up. For the first overall quarterback taken, however, the mean and median are both 59, just a little worse than David Garrard at 61. The upside, however, is unparalleled. The upper bound is 152, just three below Brett Favre at 155. This upside drops immediately, falling to 116 and 96 at the second and third overall quarterbacks selected respectively. Those are still great picks: Tom Brady has a CAV of 115 and Mark Brunell has a CAV of 96; however, this is the absolute best-case scenario for these picks. They are much more likely to be closer to the median quarterbacks selected second and third overall: second quarterback should be right below Tim Couch at 31 CAV; the third should be right above Patrick Ramsey at 14 CAV.

These findings also imply that, in general, NFL teams are good at identifying players who will become good quarterbacks. If they did not, the above graphs would have trendlines sloping up.  Given the strong negative slope of the median curve above, it is clear that CAV falls continuously as the draft continues. There are obvious exceptions to this trend: Alex Smith going over Aaron Rodgers; Tom Brady drafted after Chad Pennington, Giovanni Carmazzi, Chris Redman, Tee Martin, Marc Bulger, and Spergon Wynn. However, perspective is important. These are the exceptions; the rule is that teams are good at identifying talent at the quarterback position.

Order drafted is not the only important way to analyze the quarterback position. Overall draft pick is also important. Looking at the data from draft pick instead of order selected requires grouping picks into 16-pick (half round) buckets; without this grouping the sample sizes are way too small. Even with these groups, the models are very noisy. The graph and analysis follow below.

The large groupings hide the drop in value from early to late round picks in the mean and median graphs. That said, there is still a clear downward trend from early rounds to later ones. The upper bound line is the most important line here. The extreme drop after the 33-48 bucket is the really significant point to take from this chart. Before the 48th overall pick, any quarterback has the potential to be the next Jim Kelly or Drew Bledsoe. After that pick, the upside vanishes. For the rest of the draft, the maximum upside of every quarterback is right around the David Garrard range. Mark Brunell solely causes the peak at the 113-128 bucket, which demonstrates that it is possible to find a talented quarterback late; it’s simply highly unlikely.

Tom Brady

The conclusion here is simple: if you need a quarterback, draft a one as early as possible. If you draft a quarterback after the 48th overall pick, temper expectations. There’s a reason Tom Brady’s story stands out: he’s the only quarterback of 262 drafted after pick 100 since 1980 to have a CAV over 100. Only Mark Brunell and Trent Green are close, but Brady’s CAV/year is a full 4 points higher than theirs, putting him in a class apart. If teams want a good quarterback, they must draft one early.

About the author


View all posts


  • Terrific work, Kevin. Some interesting stuff to think about as we approach what could be a QB rich draft.

    I think you may be giving teams a little too much credit for being good talent evaluators–you forgot to mention the fact that top picks get more playing time/opportunities as a result of their high draft status. As someone who watched every moment of the Tim Couch era in Cleveland, 31 CAV is probably generous for his talent level. If Tee Martin were drafted first overall, he would likely have a much higher CAV since his team would have given him at least a couple years as a starter (even JaMarcus [CAV 7] got nearly that much time).

    It’s difficult to adjust for this issue. Subtracting the lowest historical CAV for each pick may help reveal true talent level/contribution. For example, for all first overall picks, you would remove 7 because that’s the absolute minimum someone at that spot should get for CAV. For later picks, you would subtract 0 as no team feels pressure to play a late pick.

  • This is all wrong. Indeed, “the conclusion here is simple: if you need a quarterback…”, draft a team first.
    Possibly Rodgers?
    Stabler (then Plunkett)
    Van Brocklin

    And then there’s:
    Collins (especially him)

    What’s the difference in that professionally judged talent? The former (not early picks) were drafted by quality teams and the latter (top picks) by league fodder.

    The fewer pieces such as this that fail to recognize football as a team game, the better. Thanks for your work, but you fail.

    • Hi Not the Objectionable Type:

      My post analyzes how the value of quarterbacks drafted first in their class or using earlier picks in the draft falls quantifiably as the draft continues. I am not arguing that quarterbacks make or break a team. Of course football is a team sport, and the quality of the team around a quarterback affects his performance. However, all football teams need a quarterback, and analyzing that piece of the puzzle is important.

      The names you list are all included in the data I analyzed (except Kurt Warner, who was not drafted), along with every other quarterback who was drafted from 1980-2005. That data set includes not only late round picks by quality teams that worked out well and early round busts drafted by “league fodder”. You can pick out a subset of my data to argue almost anything: a sample size of JaMarcus Russell, Ryan Leaf, Tom Brady, and Matt Hasselbeck would lead you to believe that teams should never draft a quarterback in the first round, but rather in the 6th round only. Instead of looking at special cases, I prefer looking at as large a sample size as possible.

      On your point about “quality teams” and “league fodder”: I am currently working on a team by team breakdown of the draft to see if, for example, the Patriots actually do draft better than, again for example, the Redskins. Please check back soon.

      • I think your undertaking is an impossibility given how reliant the QB position is on other team units. Now, you could run your same research on a subset of QB’s, (how did these drafted QB’s perform for GB, DAL, etc., and how did those perform for CIN, NYJ, etc.) but combining them will only lead to faulty results because, look around the league and tell me what great QB is playing for a crappy team. You can’t because they don’t exist.

        Actually, they do exist but we wouldn’t know who they are because Bradford (for example) will never become known as a great QB playing for that franchise.

  • I hope to see similar analyses at other positions. I suspect that dadler3 is correct that quarterback playing opportunities are skewed by draft position. A high draft pick gets plenty of chances while a low draft pick might not get much playing time. A DB drafted in later rounds probably gets to play on special teams or 3rd and long, eventually working into the starting lineup with good performance.

    I’m curious about why you chose the particular functions you did (exponential and natural log) for your trend lines.

Leave a Reply

Your email address will not be published. Required fields are marked *