Football Freakonomics: What Can Linsanity Teach Us About the Upcoming NFL Draft?

(Photo: Nicholas La)

The following is a cross-post from our Football Freakonomics project at NFL.com. Check out the interactive graphic and, at the end of this post, the video.

In his first six NBA starts, Jeremy Lin averaged 24.3 points and 9.5 assists while leading the Knicks to six straight wins.

If those numbers were attached to someone like Kobe Bryant or LeBron James, you wouldn’t bat an eye. But until a couple weeks ago, Lin was little more than roster fodder, an undrafted player already cut by two teams and about to be cut by his third. That’s when a desperate coach who had run out of able-bodied point guards threw him into the fire. The rest – for the moment, at least – is history.

Let’s be honest: the reason we’re hearing so much about Lin is because he was overlooked. This might lead you to think he’s a true anomaly, a great game-time athlete who somehow slipped through a pro sports league’s finely-tuned talent-scouting machine.

But if you look closely at the NFL, you’ll find Jeremy Lins all over the place. And with the NFL Draft coming up in April, you have to wonder just how scientific the science of drafting football players really is. Is Andrew Luck really the golden goose that Indianapolis is banking on, or might he turn out to be yet another top-tier bust?

Our latest Football Freakonomics episode — the last one this season — argues that the draft is much more of a crapshoot than most of its practitioners would have us think. The evidence is everywhere. Consider the research of Cade Massey and Richard Thaler, who find top draft picks to be seriously overvalued. Consider the data presented in the interactive graphic here, which reveals the average draft position for the top five players this season in key categories. For instance:

2011 Passing Yards  (Avg. Draft Position for Top 5 Performers = 51.4)

 

Player

Overall Pick Number

 

 

 

Drew Brees

32

 

 

Tom Brady

199

 

 

Matthew Stafford

1

 

 

Eli Manning

1

 

 

Aaron Rodgers

24

2011 Rushing Yards (Avg. Draft Position for Top 5 Performers = 115.8)

 

Player

Overall Pick Number

 

 

 

Maurice Jones Drew

60

 

 

Ray Rice*

55

 

 

Michael Turner

154

 

 

LeSean McCoy

53

 

 

Arian Foster**

Undrafted (257)

* Rice was the MVB (Most Valuable Bargain) in our inaugural Dough Bowl.

** We generously counted each undrafted player as if he was the first player chosen after “Mr. Irrelevant,” the last player chosen in that year’s draft. In Foster’s case, that would make him pick No. 257.

2011 Tackles (Avg. Draft Position for Top 5 Performers = 78.6)

 

Player

Overall Pick Number

 

 

 

London Fletcher 

Undrafted (242)

 

 

D’Qwell Jackson

34

 

 

Chad Greenway

17

 

 

Pat Angerer 

63

 

 

Curtis Lofton

37

2011 Receiving Yards Leaders (Avg. Draft Position for Top 5 Performers = 118.2)

 

Player

Overall Pick Number

 

 

 

Calvin Johnson

2

 

 

Wes Welker

Undrafted (256)

 

 

Victor Cruz

Undrafted (256)

 

 

Larry Fitzgerald

3

 

 

Steve Smith (CAR) 

74

This year’s sack leaders represented the highest average overall draft position: 

2011 Sacks (Avg. Draft Position for Top 5 Performers = 32.7) 

 

Player

Overall Pick Number

 

 

 

Jared Allen

126

 

 

DeMarcus Ware

11

 

 

Jason Babin

27

 

 

Jason Pierre-Paul

15

 

 

(T5) Aldon Smith

7

 

(T5) Terrell Suggs 

10

Keep in mind that an average draft position of 100 is eq
uivalent to a top pick in the fourth round. What’s most interesting is that in several major categories, the top five performers included at least one undrafted free agent. Or, put another way: One of the top five performers in these categories includes a player who wasn’t even thought to be among the top 250 players of his rookie class!

Granted, these numbers aren’t exactly encyclopedic. But they do a good job of showing just how much luck is involved in the draft — to say nothing of how much Luck — and how hard it is to forecast the future.

That’s why more and more teams, in all sports, are doing a deep statistical dive to try to identify undervalued players before spending too much money on the overvalued ones. So if you’re an NFL team looking to optimize your draft picks this year, maybe you’ll consider hiring a young guy to crunch the numbers for you. Maybe some economics major from Harvard who happens to know his way around the gym? Maybe someone like … Jeremy Lin?


Ben Martin

Comparing Jeremy Lin, an un-drafted player in the NBA to un-drafted NFL players isn't a fair comparison. The NBA draft is a more accurate predictor of future success than the NFL draft. There have been zero un-drafted NBA players make the NBA hall of fame. Ben Wallace will probably be the first but outside of Wallace and John Starks you will be hard-pressed to find any true stars in the NBA that went un-drafted.

Adrian B

Steroids only give you 2 -3 years of elite performance. Hence, the arc of pro careers is dependent on the amount steroid use. SEE: Adrian Peterson v. Maurice Jones Drew

Jake

Outside of passing yards for Quarterbacks the metrics you chose are only tangentially relevant to determining who is a great football player. Unlike baseball and to a lesser extent basketball, so much teamwork and orchestration is required, not to mention so much more variation in team strategy and tactics, to make a football team operate that judging any player's value based on his standard counting numbers is not terribly useful (exceptions are basically quarterbacks and kickers/punters). Many of the most successful top 5 draft picks have been left tackles, for instance, a position where pedigree seems to be incredibly predictive of success.

I've been deeply involved in baseball sabermetrics, less so in football, so I can't quote with as much precision, but it seems that the most relevant work with football statistics has been around team strategy rather than player evaluation, and the most relevant player stats have been determined by crunching numbers based on team play (they don't necessarily translate year to year or team to team). In other words, even advanced football stats tend to be more descriptive than predictive in the long run. In case you all are wondering the only NFL team to ever make a point of using advanced stats to find players was the Mike Nolan/Scott McLoughan era 49ers, and those teams went 16-32. Yikes.

Read more...

Adam Roberts

Team strategy is definitely the most important use of statistics, but player evaluation is fine. The problem is that even the most die-hard stat-junkies are often biased toward certain statistics. Also, some useful statistics are only measured by private firms, placing a very high barrier-to-entry on public analysis.

It seems that you can parse out some reasonably good measurements of a player's value given the right data, but nobody has figured out how to do it fully just yet. The trick is that the average player spends about 3 years in the league, meaning there is a limited value to forecasting.

The real goal should be identifying undervalued free agents and trade opportunities.

Adam Roberts

I intend on reading the paper listed above, but want to point out that this kind of approach is ridiculous.

Here are just a few problems with draft-pick analysis:

1) Team needs: it is possible there is an amazing OLB who is most valuable, but if the worst 10 teams want QBs, DTs and WRs, he's not getting picked and who's going to trade up to get an OLB?

2) Sometimes you have your eye on a player but can't afford to move up. The Falcons traded up for Julio Jones - hopefully he proves worthwhile but it was expensive. The Redskins just handed away 3 first round picks to move up 4 spots and take a QB (RG3 if the analysts are right). I can all-but guarantee that it won't be the right play in the long-run, even if RG3 isn't the next Ryan Leaf. 2 years of a supporting cast that won't have first-round caliber improvements means that he will have to work exceptionally-well with what he has to be competitive.

3) Sometimes there's no need to move up - Aaron Rodgers went in the first round, probably the 3rd or 4th QB - still very distinguished. Often it is going to be impossible to tell where a player will stand among players with similar pedigrees. Who's to say any QB who went earlier wouldn't have flourished similarly under Favre's tutelage?

4) NFLPA draft contract structure basically prescribes value. While the NFL is a monopsony for players, you are guaranteed money based on draft position; relatively weak or strong draft classes will shift the ROI of drafts.

5) I'm not sure, but I don't think draft-position rookie contracts are position-specific so player "values" would be skewed relative to the market.

6) Football is a team sport, but most metrics don't effectively isolate many players' contributions, making it hard to value many players. For example, any QB is going to look good with a solid O-line, and Fitzgerald/Marshall as WRs, but that doesn't mean the QB did the work. If he spends his career with great protection and receiving, he could rack up HoF stats without being amazing. (Think of Shanahan's approach to RBs.)

7) Until we have strong metrics for all positions, we can never have accurate relative valuations, thus the draft will never be "optimal" and will tend to make many mistakes. Even then we have to rely on the ability to forecast with these metrics.

8) College data is poor because there is no way to tell which pieces matter. "Cupcake" games against easy schools pad stats as much as a phenomenally tough schedule may hinder them. Unfortunately it is impossible to get good quality stats on all players so there is some luck to it no matter what.

One thing I haven't seen is work on physiques/physical characteristics or fitness regimens and how they correlate with future performance. These would have inherent biases that would require controls, but I think you'll find some interesting results.

Read more...