When I watch the NFL draft, I sometimes wonder why teams feel the need to draft certain positions over others. If teams choose the best player available, how do they compare players of different positions? Now that we’ve analyzed each team’s success in the NFL draft, I want to touch on the success of each position in the NFL draft and the inherent variation in the draft by position.

While the NFL draft has seven rounds with all 32 teams receiving one pick per round (7 rounds x 32 teams = 224 picks), the additional 32 (sometimes more) compensatory picks effectively create an additional round in the draft (224 + 32 = 256). Therefore, instead of analyzing the seven rounds as defined by the NFL (with some rounds having more than 32 picks), I separated the picks into 8 rounds in order to evenly distribute the number of picks in each round (and because I wanted to be a rule breaker and make things more complicated). Since compensatory picks are not awarded until the third round, this will not affect any conclusions based on the first and second rounds. Below is the range of picks that I include in each round. Any references to draft rounds later in this post reflect the distribution of picks by round as shown here.

**Position Breakdown**

From 2002 to 2012, there were 2,813 players selected in the NFL draft. Of those choices, 48 were mistakes by teams who chose either a kicker or a punter (more on this in a later post). After removing those choices, 50.3% of players chosen were offensive players and 49.7% were defensive players. It makes perfect sense that almost exactly half of players chosen are offensive players and almost half are defensive players, given that there are 11 starters on each side of the ball. When we think about this from a total roster composition perspective, teams are fairly balanced in terms of the distribution of offensive players compared to defensive players. Using the roster composition laid out by Marc Lillibridge of Bleacher Report (slightly adjusted to remove one defensive back and add a third quarterback), as shown below, teams typically include 25 offensive players and 25 defensive players on the 53 man roster (in addition to a kicker, a punter, and a long snapper).

This equal distribution between defense and offense also shows up if we isolate the first two rounds of the draft. 48.4% of players chosen in the first or second rounds between 2002 and 2012 were offensive players, meaning 51.6% were defensive players (if we exclude the one kicker, Mike Nugent, on whom the Jets wasted their first pick of the 2005 draft, the 47^{th} overall pick).

Now that we’ve established that, on average, teams don’t favor one side of the ball over the other early in the draft, let’s take a look at how different positions compare by draft round. The tables below display the number of players chosen by position and by round and the number of players chosen by position as a percentage of total payers selected in that round.

Not surprisingly, tight ends and interior offensive linemen are among the lowest percentage of players chosen in the first round and in later rounds. Teams don’t value these positions – another way to interpret this is that the marginal benefit provided by a guard with first round ability (i.e., potential first round AV production) compared to a guard selected in a later round is lower than the marginal benefit of another position with first round ability compared to a player of the same position selected in the same later round.

So, what positions do teams value? Teams like wide receivers. The most drafted position is wide receiver (Note that the data I used do not differentiate between cornerbacks and safeties. If you look at ESPN’s draft data, some players are listed as corners, some as safeties, and some as defensive backs. It’s easier and more accurate to not separate these out, but I digress.). If we go back to Marc Lillibridge’s roster composition chart, wide receivers are the second-most populous position on a football team, behind linebackers, which are only 16 selections behind from being the most drafted position. Both of these positions typically require 2.5 to 4 starters, given the rise of spread offenses (more wide receivers, third linebacker is a 0.5 player), the increased play of nickel cornerbacks (more wide receivers, third linebacker is a 0.5 player) and the prevalence of the 3-4 defense (requiring 3.5 to 4 linebackers).

Wide receivers, unlike linebackers, are also one of the most drafted positions in the first round. 11.6 percent of players chosen in the first round were wide receivers. We see more of them, relative to other positions, in the second through fourth rounds, however. Relative to other positions, teams draft wide receivers whenever they can find them, whether that be in the first round or later. Linebackers, on the other hand, are not drafted as much in the first round, but teams think they can get good value from linebackers as the draft wears on.

In the first round, teams prefer to select QBs and defensive lineman, relative to later rounds. Look at the drop off in the percentage of first round picks who are QBs compared to the second round. 9.7 percent of players chosen in the first round are QBs, compared to 3.4 percent in the second round, 4.0 percent in the third round, and on. Teams like to take their quarterbacks early, don’t see much value in using a mid-round pick on a quarterback, and take a flyer on quarterbacks later. This makes sense to me, given the importance of the quarterback position. Quarterback-needy teams try to find their franchise quarterbacks early. Teams that don’t need a quarterback take a quarterback late for depth but won’t waste a valuable mid-round pick on a player who won’t compete to start in the short term.

Defensive linemen, especially edge rushers, have always been valuable commodities. Teams have shown that in their high selection rate in the first round, especially relative to other positions, though teams still draft defensive linemen in later rounds in droves. This helps me understand why the selection of Mario Williams over Reggie Bush in 2006, who was seen as a once in a lifetime running back at the time, was not a complete shocker.

**Premier Positions**

This emphasis on quarterbacks and defensive linemen as the premier positions is further shown in the distribution of players selected by position in each round. The table below shows the number of players selected for a certain positions in each round as a percentage of the total number of players selected for that position.

23.6 percent of quarterbacks selected in the NFL draft were drafted in the first round. This drops off to 8.3 percent in the second round and 9.7 percent in third round before spiking back up to 16.7 percent in the seventh round. This further confirms my assertion that teams tend to stay away from quarterbacks in the middle rounds. Is this rational? For the most part, I think so. Recency bias would tell us that I’m crazy for saying that, given the success of Dak Prescott, Russell Wilson, and Kirk Cousins (click on the Russell Wilson link, it’ll show you how worthless it is to grade the draft the day after). However, of the 45 quarterbacks drafted in the third to fifth rounds from 2002 to 2012, 37 generated 4 points of approximate value (“AV”) or less over the first five years of their career. Except for a few outliers, most of these guys don’t contribute much, if anything.

Late round quarterbacks haven’t done any better than mid-round picks. Only 3 of the 58 quarterbacks (Matt Cassel, Tyler Thigpen[1], and Derek Anderson) produced more than 4 points of AV over the first five years of the career. However, one of the flaws of AV is that it fails to measure value of players who are unable to see the field. Since there is only one quarterback on the field at a time, many mid-round picks will have a tough time displanting an entrenched starter from the starting lineup, especially early in their career. Some of these guys may have been able to produce on the field, but never got the opportunity.

It’s smarter to use a later pick on a player who won’t see the field, given that the earlier the pick, the more valuable it is. The expected value of a later round pick is lower than that of a mid-round pick. From the perspective of opportunity cost, there seems to be no reason to select a quarterback in one of the middle rounds, given the relatively low amount of output. Take a quarterback early if you need one (but not in 2011! – Gil’s a little off on this one). Build your quarterback depth later in the draft, if needed, and don’t waste a selection that could be used to find a potential starter at one of the other 21 positions. If the Seahawks had expected Russell Wilson to be as good as he is, they would have selected him in the first round.

Behind quarterbacks, defensive ends and defensive tackles are the next two positions that are highly coveted in the first rounds. If teams are drafting for need, it suggests that teams consistently need high end quarterback and defensive linemen play, which cannot wait for later rounds. If teams are drafting to take the best player available (as defined as producing the most amount of AV over their career), it suggests that teams consider defensive linemen to be extremely important positions in terms of team success (a major input in AV production ‘cause, you know, teams want to win games).

**So which positions are the most successful in the draft?**

In order to analyze the variation in AV production by position, we have to first understand how the AV data are distributed. AV production is not normally distributed (a lot of 0s and 1s, not a lot of high numbers), with a lot of low values. There are a lot of players who barely make it into the league or have a marginal impact. Not many are Hall of Famers or even consistent starters. Statistically speaking, this means that the average AV is higher than the median (i.e., the average is higher than the middle value in the data) and the data is skewed right. Therefore, the average and standard deviation do not accurately represent the data. As shown in the table below, the average for each position is always higher than the median.

Instead, we can better understand the data by analyzing the median and the upper and lower quartiles. The median is the middle number in the set – 50 percent of values are higher and 50 percent of values are lower than the median. The lower quartile is the 25^{th} percentile – it’s the median of the lower half of the data. Likewise, the upper quartile is the 75^{th} percentile and the median of the upper half of the data. For example, 75 percent of quarterbacks drafted from 2002 to 2012 generated less than 7.8 (the upper quartile) points of AV and 25 percent of quarterbacks generated more than 7.8 points of AV. While this is high, it doesn’t compare to other positions. Tackles have the highest ceiling at 13.4 points of AV Part of this is due to the inherent issues with AV, which I will discuss in a later post.

For all positions, the lower quartile is 0, which makes sense given the large numbers who don’t generate any AV. The median is the highest for offensive linemen and defensive ends, suggesting that these positions are the least risky to select. However, as discussed before, tackles have the highest ceiling. Assuming I evaluated two players of different positions as equal on all fronts, I would draft the tackle first given the higher ceiling.

However, none of these positions are selected the earliest, on average. That belongs to defensive tackles, who are drafted with the 119^{th} pick, on average. Quarterbacks and defensive ends are right behind defensive tackles at 121^{st} overall, a very small difference. As I discussed before, teams like their quarterbacks and defensive linemen. Maybe they should like their tackles just as much.

[1] When I was a freshman in high school, I was the admin of the Tyler Thigpen online fan club. I’m serious. I really had no life at the time, not that much has changed.