In the lead-up to this year’s draft, I wrote a series of articles on drafting Quarterbacks (Good Teams’ QBs, Bad Teams’ QBs, Super Bowl Winners, QB Sweet Spot, Is first QB best, Trading up for QBs) and one on drafting Wide Receivers. There were a few reasons to focus on the QB position this offseason. It is the most glaring need in Rivera’s rebuild of the WFT, and I felt fairly confident that the WFT would address the position at some point in the draft. Wide receiver and linebacker were probably next on the needs list, but I chose to focus on the former because it seems to defy the general trend that the best players come off the board early in the draft, which makes it more interesting than most positions from a draft strategy perspective.
The WFT showed how much I know when they avoided QB in the draft, even despite Justin Fields falling out of the top 10, but at least they addressed WR twice. And they even did the right thing according to KyleSmithForGM, by avoiding the flashier WR prospects in the first round, who tend to have high bust rates, and making their first move for Dyami Brown in the third, where the value proposition for drafting receivers is optimal. Then they went back for a second bite, drafting Dax Milne late in the seventh round where there is really nothing to lose.
Another reasons why I chose to focus on QB and WR was the sense that I and many fans have that these seem to be the hardest two positions to get right in the draft. One theme that emerged consistently in the comment threads under these articles was that it would be great to compare different positions to see if some are harder or easier to draft. I spent a long time thinking about how to do a fair comparison, but could not figure it out, until I had a realization that I had been looking at things the wrong way.
The key to solving this problem comes from the analysis I did recently, when I revisited the question of whether there is a Sweet Spot in the draft to optimize the ratio of Opportunity to Opportunity Risk when drafting Quarterbacks. In the revised analysis, I used a variable which I named Relative Availability as the measure of “Opportunity.” Relative Availability is simply the proportion of players available at any draft pick number who have gone on to meet a certain performance criterion (e.g. QB Rating > 100, 8 or more starts in their first year, signing a second starting contract) within the total pool of players available in the draft.
In the revised Sweet Spot analysis, Relative Availability of future long term starting QBs provided a better measure of the Opportunity available to teams at a given draft pick than past hit rates, because it reflects the quality of the talent pool available at that pick rather than the (often bad) decision making of teams which have drafted players there. But once I thought about what it was, I realized it had significance well beyond that specific problem.
In this two-part series, I will use the Relative Availability metric to evaluate the talent pool of players at different offensive positions throughout the draft. The key question I will ask is whether some positions on offense are harder to pick correctly than others. Without giving too much away now, I’ll just say that the Relative Availability statistic provides the key to correcting some misleading impressions that might arise if we just looked at draft hit rates in isolation.
One of the most frequent comments on my last article was that it made readers’ heads hurt. I heard you and have taken two steps to make this one an easier read. First, I have broken what would have been a really long article with lots of data figures into two shorter, more easily digestible pieces. Second, breaking it down into two articles cuts the Methods section in about half which might help to make readers’ heads spin a bit less.
In the first of the two articles, I will examine past draft Hit Rates and Relative Availability of good players throughout the draft to get an initial impression of whether some offensive positions are harder or easier for teams to pick correctly in draft.
I had to pick a criterion for “good players” which could be applied equally to all offensive positions, and which was amenable to analysis of hundreds to thousands of players to provide robust and reliable results. I chose to use Pro Football Reference’s Approximate Value (AV) statistic, which is designed to provide a single number which rates the overall productivity of players across positions in a single season on the same scale.
I settled on AV in the third year of a player’s career, because most drafted players should be coming up to speed by that point in their careers and using the third year allows me to take the analysis as recent as 2018. I looked at players drafted in the 10 seasons from 2009 to 2018 to provide a large data sample for analysis. Use of the player’s third season as the criterion may have resulted in overlooking some players from the 2018 draft class who missed a year to injury or COVID opt out, and disadvantaged players that missed a year in other draft classes. I hope that’s not a major issue, because it would have been too time consuming to sort them out. Also, players missing time to injury is a fact of draft outcomes, so one could argue it would be incorrect to correct for injury time.
I used an AV of six or more as the criterion for “good” players. To illustrate what that threshold means, here are the AVs of Scott McLoughan’s excellent 2015 draft class in the 2017 season:
- Brandon Scherff 9
- Preston Smith 8
- Jamison Crowder 6
- Martrell Spaight 4
- Arie Kouandjio 3
- Austin Reiter 1
- Matt Jones 0
Scherff, Smith and Crowder make the “good” player cutoff while the others fall below the line. This also illustrates that no arbitrary criterion is perfect. Austin Reiter was a relatively late bloomer and didn’t take off until his fourth NFL season, posting an AV of 9 in 2019 on his way to collecting a Lombardi as the starting center for the Chiefs.
Hit Rate is fairly self-explanatory. It is simply the number of players drafted in the specified range of picks or draft round who meet criterion (third year AV > 5) divided by the total number of players drafted in the same range. This measures how well teams have done historically at drafting players at a given offensive position in a given range of picks.
Relative Availability provides a measure of the quality the talent pool of players at the specified offensive position at any particular overall pick number in the draft. Put simply, Relative Availability is the proportion of “good” players to total available players at a specified position at any given pick number in the draft.
Relative Availability is calculated as the number of players at a specified position who meet criterion that were selected at that pick number or anywhere later in the draft, divided by the total number of players who were selected at that pick number or later in the draft.
For example, 227 running backs (RBs) were selected at the eighth overall pick or later from 2009 to 2018. Of those RBs, 41 had an AV of 6 or greater in their third NFL season. Therefore, the Relative Availability of good RBs at pick number 8 is 41/227 = 0.181.
Sampling. As I alluded to above, this analysis strikes a balance between keeping things as recent as possible, while at the same time including a large enough sample to generate reliable results. So I set the performance criterion at AV in the third season to allow sampling draft classes as recent as 2018. This might cause us to miss some late blooming good players, like the example of Austin Reiter above. But it is applied equally to all players and position groups, so it should not bias the results.
As in previous articles, I attempted to analyze Hit Rates more finely in the first round than later in the draft. However, at some position groups (e.g. RB, WR, TE) that resulted in small sample sizes, which made the graphs noisy. I settled on sampling pick ranges 1-8, 9-16, 17-32, followed by whole rounds. The calculation of Relative Availability results in very large sample sizes early in the draft, which permitted sampling the first round in a power of two sequence (picks 1, 2, 4, 8, 16, 32) and then every 16 picks after that.
For graphical purposes, Hit Rates are plotted at the midpoint of the specified draft range (e.g. midpoint of round two is pick #48). Because draft rounds four through seven start at a different pick number every year, due to the distribution of comp picks, I calculated the average start and end points of these rounds from 2009 to 2018.
Draft Positions. Hit Rate and Relative Availability were analyzed by draft position as listed in the Pro Football Reference database. This provided good resolution of all offensive positions except offensive linemen. There is so much switching from draft position to NFL playing position within this group that I felt it was not possible to separate them. Therefore they are analyzed as a single position group. This problem will get much worse if I get to looking at defensive positions in an upcoming article.
Sample Size. This analysis was based on 1,223 offensive players drafted from 2009 to 2018 distributed as follows: 407 OL, 149 TE, 231 RB, 319 WR, 117 QBs.
Results - Hit Rates
When we think about the ease or difficulty of drafting players at a particular position, we tend to think in terms of the hit rates that teams have achieved in the draft. So I will start with that first. Hit Rates for drafting offensive position groups are shown in the first figure.
The Hit Rate plots are a bit noisy, particularly in the first round where sample sizes of some positions are small (e.g. RB, WR, TE). Nevertheless, a fairly clear picture emerges. Overall, Hit Rates for OL tend to be higher than other position groups. With the exception of the one glitch at picks 17-32, RB tends to be next highest and even overtakes OL slightly from Rounds four to five. That glitch is possibly a small sample issue, since only eight RBs were selected in this range.
As most readers might expect, after the first eight picks, Hit Rates for QBs drop sharply and become the lowest for any position group throughout most of the rest of the draft. Hit rates for TEs follow a fairly similar pattern to QBs but shifted a bit to the right and staying a bit higher later in the draft.
Readers who caught my article on drafting wide receivers might not be too surprised to see that they have the flattest Hit Rate curve of any position. Hit rates for WR start lower than QB, RB and OL early in the first round (I didn’t include TE in that list, because the 1.0 Hit Rate at picks 8-16 is due to a single player, Eric Ebron). It doesn’t appear that this is just a sampling size issue, since decent numbers of WR are selected throughout the first round (picks 1-8: 11 players; picks 9-16: 6 players; picks 17-32: 21 players). Rather, Hit Rates for WR appear to be low relative to other positions early in the first round and higher than most other positions in Rounds five and six.
To this point, it appears that the rank order from easiest to hardest to draft is approximately OL, RB, WR/TE (hard to split) and QB. Now let’s have a look at Relative Availability of players at these same positions, which provides an indication of the quality of the talent pool at any given pick number, minus any confounding influence of the decision making of the teams making selections at that pick number.
Results - Relative Availability
The next figure plots the Relative Availabilities of the different offensive position groups throughout the draft.
The first surprise is how, in contrast to the Hit Rates, the Relative Availabilities of the different player groups completely separate from one another. Viewed in this way, the rank order becomes crystal clear OL > RB > WR > TE > QB. The only two position groups that come together or overlap a bit are OL/RB and WR/TE.
I also find it fairly remarkable just how much more prevalent quality OL are early in the draft compared to any other position group.
Referring back to the first figure, the fact that the rank order of Relative Availabilities closely mirrors that of Hit Rates suggest an alternative explanation to the one I arrived at in the previous section. It might not be any easier for NFL teams to distinguish between good and bad OL than WR or QBs. Rather, the rank order of draft Hit Rates might simply reflect the differences in the Relative Availability of good players at different positions throughout the draft.
Plotting past draft Hit Rates revealed that NFL teams have achieved differing levels of success at drafting different offensive position groups. According to this criterion, Offensive Line seems to be the easiest position group to get right in the draft, followed by RB, WR/TE and QB, in that order.
However, additional analysis revealed that the rank order of Hit Rates closely matches that of the Relative Availabilities of good players throughout the draft, suggesting the former might simply be a reflection of the latter, rather than a true difference in relative difficulty of scouting players at different position groups. In other words, good Quarterbacks may be more difficult to pick than other position because they make up a smaller proportion of the total available talent pool, rather than any inherent difficulty in evaluating their particular skill sets compared to receivers, running backs or offensive linemen.
By now, statistically-minded readers may have realized something about the Relative Availability metric. It does, in fact, have a very specific relationship to Hit Rate. In the second article of the two-part series, I will reveal what that is (unless someone spills the beans in the comments, first), and use it to tease out the effects of scouting on draft Hit Rates, which should make it possible to answer the question posed in the title.
As usual, thanks to James Dorsett for expert editorial assistance. I may have managed to coax him out of semi-retirement, stay tuned.
Which WFT draft pick on offense is most likely to outperform his draft status?
This poll is closed
Sam Cosmi, OL, 2nd round - hit rate 59%
Dyami Brown, WR, 3rd round – hit rate 23%
John Bates, TE, 4th round – hit rate 13%
Dax Milne, WR, 7th round – hit rate 0%