The NFL draft starts this Thursday and ends sometime in 2015 I think, so I finally cranked out some graphics that I have been wanting to look at for awhile. This work depends on the player valuation that Doug Drinen came up with: Career Approximate Value/CAV. I should also mention that Kevin Meers did a pick valuation chart I found really illuminating here.
Data are from the 1980-2006 drafts. Average CAV starts falling off after 2006 because players haven’t been in the league long enough to assess their career (though you could fudge in a couple/few more recent years by scaling up the CAV of players who are still in the league if you had to). For most of these I dumped everything after pick# 250, so we’re looking at 7 rounds + a little into round 8 pre-1993. Data are from Pro football reference.
Finally, for some reason the vast majority of CBs and FS/SSs were all called “DBs” in the dataset. I’ll do a lookup and post an updated version of this for part 2.
Let’s start with what kind of Career Approximate Value you see by draft pick:
That QB at pick 199 is, of course, Tom Brady. The distribution is very bottom-heavy no matter where you slice it, but the first pick average CAV is ~67, your average pick CAV~=16, and your median pick CAV~=6.
Breaking that down by position:
Drinen wisely didn’t let any K or P or FB have a stratospheric CAV. TE probably looks different in recent years/will look different going forward. First thing that jumps out at me is that historically your best bet, late in the draft, are Guards and Centers, and as they’re similar positions the higher valley might not be a fluke.
So when have teams tended to pick which position?
This isn’t a great graphic, and a lot of it is “no duh” stuff. Most of the action is in the first line segment (difference between picks 1-25 and 26-50), and this would be my summary:
-Positions that were disproportionately selected in early/mid round 1: QB, T, RB, DE
-Positions that were disproportionately NOT selected in early/mid round 1: LB, DB, G, C, TE
-Positions that were rarely selected early and slightly less rarely selected later in the draft: P, K, FB
-Positions that were selected steadily through draft: WR, DT
-I’m not sure if that plunge in late-round-7 flyers on LBs and DBs is due to a pre/post-1993 difference or what.
Let’s take a look at picks that boom (overachieve) or bust (underachieve). Here is a plot with the players at CAV “deciles” (e.g., 20% = better CAV than 20% of players at that pick and worse CAV than the other 80% at that pick), with lines fitted to each decile.
One could make a case for different definition of a bust pick or boom pick, but I went with:
Boom: Above the “70%” curve AND CAV AT LEAST 15 (Late in the draft less than 30% of picks are CAV>=15)
Bust: Below the “30%” curve OR CAV <=1. (Late in the draft more than 30% of picks are CAV <=1)
Middling: Everything else
Note that this means an average player picked in round 7 is a boom pick, while an average player picked in round 1 is a bust. The overall picture is the following:
Finally, let’s look at the proportion of boom/busts by draft pick window and position:
-QBs were the most prone to boom/bust (have the thinnest green ribbon), unsurprisingly.
-Early picks on TEs and RBs have been prone to bust.
-OLs drafted early (relatively) rarely bust.
Here’s a quick look at the quality of draft classes:
I knew 1983 was considered a great class. I did not know the year before was so bad! 2014-05-11 edit: it occurs to me that this class was disrupted by the strike, but for another take see the 538 piece pingpack in the comments below
For evaluating team draft performance it makes sense to consider how good their picks were relative to how good an average pick at those positions are to control for their draft position, whether they had traded picks away, etc.:
This is mostly a descriptive post but: there was no statistically significant effect of team on draft performance. (If I had asked ONLY “did Pittsburgh draft reliably better than other teams 1980-2006?”, the answer would be p~=0.01 “yes”, but with alpha inflation you can’t say this wasn’t a fluke because we’re looking at 32 teams). That said, if you are a Bengals fan you are probably not surprised that they are anchoring the scale, here.
The draft looks more like a crap shoot when you break it down by year:
Finally, let’s use that actual-expected CAV measure to see the biggest booms and busts drafted 1980-2006. LIST MODE (at least I’m just putting them in a table and not in a slideshow):
|4||Jerry Rice HOF||SFO||1985||16||116.53|
|5||Dan Marino HOF||MIA||1983||27||107.63|
|7||Derrick Brooks HOF||TAM||1995||28||103.07|
|11||Rod Woodson HOF||PIT||1987||10||94.57|
|14||Mike Singletary HOF||CHI||1981||38||90.28|
|15||Michael Strahan HOF||NYG||1993||40||89.08|
|16||Richard Dent HOF||CHI||1983||203||88.96|
|20||Emmitt Smith HOF||DAL||1990||17||87.01|
|21||Aeneas Williams HOF||CRD||1991||59||86.88|
|22||Bruce Matthews HOF||HOU||1983||9||86.06|
|23||Rickey Jackson HOF||NOR||1981||51||85.18|
|34||Andre Reed HOF||BUF||1985||86||80.44|
|35||Curtis Martin HOF||NWE||1995||74||80.35|
|37||Thurman Thomas HOF||BUF||1988||40||80.08|
|38||Bruce Smith HOF||BUF||1985||1||79.67|
|39||Lawrence Taylor HOF||NYG||1981||2||78.43|
|42||Randall McDaniel HOF||MIN||1988||19||77.96|
|47||Anthony Munoz HOF||CIN||1980||3||75.96|
|52||Marshall Faulk HOF||IND||1994||2||74.43|
|62||Warren Sapp HOF||TAM||1995||12||72.57|
|63||Shannon Sharpe HOF||DEN||1990||192||72.57|
|70||John Elway HOF||BAL||1983||1||70.67|
(It’s admittedly a little weird to have 1st overall picks be considered “biggest booms”)
(JaMarcus Russell was drafted in 2007, but unless he has an improbable comeback he will edge out Leaf by a few points, here, just because he was 1st pick overall)
Part 2 will look at the relationship between draft success, which teams bet on which positions, and team success (in terms of winning actual NFL games).