By BBD Guest Writer: Paul Seebald
Here’s a link to the post from last year, if you would like to read it along with the predictions I made (in January, it should be noted). Some of them didn’t turn out (like my prediction that we wouldn’t draft a RB), but in general my observations turned out pretty well: 1) I said that DE and DB were two very highly regarded positions, and our first two picks were a DE and a DB, 2) I predicted a couple of LB’s in the mid rounds, at least one in rounds 3-5, and we drafted Sheppard and White, 3) I predicted at least one OL in the later rounds, we drafted Hairston in the 4th and Jasper in the 7th (I’m going to count him as OL).
This time around, I’ve included both of the Buffalo drafts that have already occurred. If I had the time, one thing I’d do is take the numbers of players drafted at each position for Buffalo the last two years and compare to the numbers of players drafted for San Diego. This way we can see what positions Nix might be more inclined to draft in the coming years.
I’ll say that the same disclaimers apply now as they did in my post last year.
Anyway, here’s the updated summary of players drafted (supplementary draft and punters/kickers not included):
QB – 5 total, average drafted position of 107.8
RB – 8 total, average of 111.9
WR – 7 total, average of 104.1
TE – 2 total, average of 116
OG/C – 4 total, average of 131
OT – 14 total, average of 171
DT – 4 total, average of 178.8
DE – 6 total, average of 61.5
LB – 12 total, average of 128.3
DB – 14 total, average of 89.8
The conclusions that can be reached from this data are essentially the same as last year. Mostly because Nix’s draft last year followed these averages very closely.
I’m going to briefly insert a mathematical discussion below. For those who aren’t interested, feel free to skip it, but it’s something that I’m looking at to help extend this analysis.
For a typical Gaussian profile, two values can essentially define the profile: mean and standard deviation. When discussing trends, just looking at the mean does not provide all of the data, even if the trends are entirely Gaussian.
Similarly, the mean value of the draft slot of the positions do not tell the whole story, especially when there is relatively small amount of data to work with. Is the average draft position of the quarterbacks really representative of when they’re drafted? The standard deviation of the five QB’s slot that Nix has drafted is 99.8. So what does that tell us? Essentially that the majority of the time (about 68%) of the QB’s are drafted between slots 8 and 207, which accounts for all but about 40 draft slots. This doesn’t tell us much, except that knowing what we do, we can assume that very few of the QB’s are actually drafted near that average value, which is true. The exact slots that each of those QB’s were drafted are: 1, 32, 216, 81, and 209. The apparent observation from this data is that the selection doesn’t follow a Gaussian curve. Nonetheless, we can find a few more numbers that help us make some conclusions.
One of the numbers that I used this time was skewness, which essentially tells us how asymmetric the distribution is. As an example, let’s say we have five test grades for students: 100, 99, 98, 97, and 0. Obviously an extreme example, but the average value for this “distribution” is 78.8. The skewness is -2.23, though, indicating that the distribution is actually weighted towards one side, in this case towards the higher numbers. Thus, we can say that for our purpose, if a positional group has a negative skewness, then it is weighted towards the lower rounds (higher number draft slot), and a positive skew is weighted towards higher rounds than the average value. I consider a skew value of absolute value of ~0.5 or higher to be significant.
For those who skipped, the important part is bolded above. There are a few important skew values, listed here:
One final note I’ll make before providing some summary conclusions is that I like to differentiate between two types of priority that Nix puts on certain positions. One is quantity, the other is quality. Drafting a lot of a position (no matter where it is) indicates a quantity of priority, whereas drafting a position very early indicates a quality priority.
Here’s the summary observations:
1) Defensive backs are a high priority in both ways – quantity and quality. Nix drafts them highly and often, as indicated by the high numbers (14 in 10 drafts, average slot of 89.8 and skew of 0.89)
2) Defensive ends are a quality priority – only 6 in 10 drafts, but average slot of 61.5 with a positive skew of 0.42.
3) Offensive tackles are a quantity priority, and the opposite of a quality priority. Only one has been drafted before the 3rd round (Marcus McNeill in the 2nd). 14 total in 10 years, average draft position of 171 with a negative skew of -0.65.
4) Defensive tackles are a low priority in both ways – only 4 over 10 years (one of those in this analysis is Jasper, even though in my head I count him as an OL now), average slot of 179 and a high negative skew of -1.81. Troup was the only DT not selected in the 7th round, though 4 of them is not a large sample size.
5) Linebackers are a high quantity priority and a medium quality priority. The skew is extremely close to zero, which indicates symmetry around the average draft slot of 128.3. From observation, linebackers tend to be drafted in the mid rounds (3-5).
The other positions generally don’t have noteworthy numbers, except that TE’s are rarely drafted. When they are, it’s in the middle rounds.
I’m finding this year much harder to predict than last year, but here’s my predictions anyway for 2012:
1) We will draft at least one WR, probably 2+. Nix has only drafted one in two years at Buffalo, while he drafted 6 in 10 years in San Diego. In 2011, Nix did not draft a single WR for Buffalo. So what happened in San Diego in years following drafts without a WR? There were four drafts without WR’s in SD. One of them he retired immediately after. One of them a single WR was selected (round 6), and two of them had two WR’s selected (rounds 2 and 5 during one year, rounds 1 and 5 in another).
2) We will draft at least one DB, most likely 2+. One will be in the top 3-4 rounds.
3) If Nix is going to mimic his outlier OT pick of Marcus McNeill, it will be this year given our need. Expect at least two OT’s this year.
4) If a DE is selected, it should be in the early rounds (1-3),
5) Expect multiple LB’s in the mid to late rounds. I’ll go as far to say that I’m very confident we won’t draft a LB in the top 2 rounds.
6) We probably won’t draft a RB this year (crosses fingers). Let’s hope this one works out better than my same prediction did last year.
Some of these predictions are obvious and have been made by many people. To take it one step further, I’ll predict the numbers we take for each position given our number of picks (obviously this won’t be exact, but I expect to be within about 1 for each position): QB (1), RB (0), WR (2), OT (2), OG/C (0), DE (1), DT (0), LB (2), DB (2).
To finish out this post (finally, I know), I’d like to mention some ways I hope to use and develop this analysis in the future:
1) Expand this to other established GM’s (e.g. Ozzie Newsome). I’m interested to see if other GM’s have obvious trends like Nix does.
2) Create two factors to quantify the priority GM’s (Nix in this case) place on various positions based on quantity and quality, respectively.
3) Use these priority factors along with team offseason needs to more accurately predict how each team will draft. Possibly even calculate probabilities that each position will be calculated in each round.
4) Find a way to adjust this analysis for the fact that each team doesn’t choose where it selects. For example, a player selected at pick 100 may also have been selected at pick 80 by the same team if they had a pick at 80. A significant amount of variability is almost certainly introduced by the nature of the draft.
Tags: 2012 NFL Draft