Slipping Through the Cracks - How NFL Teams Solve an Information Problem

Reed Albergotti at the Wall Street Journal writes about an information problem in the NFL and how it is imperfectly solved.

Nevertheless, at a time when other sports like baseball are paying more attention to how players have played rather than how they look, many analysts say the NFL is going in the other direction—focusing more on a player's raw build and athletic ability as measured by his performance in activities like the 40-yard dash, shuttle drill and bench press.

Jeffrey Nalley, an agent who represents both football and baseball players, says the problem is simple: As the NFL draft becomes a bigger event, NFL general managers who waste an all-important draft pick on a player who doesn't look like a comic book superhero can summon the wrath of millions.

"If you're going to take a guy in the first round, he'd better fit the height, weight and speed that they're looking for," Mr. Nalley says. "Honestly, they're covering their asses."

Reed gives a list of players who fit the "doesn't look like a comic book superhero" who have gone on and been important pieces on NFL teams.

The problem is this: who's going to be the best available player at position x and what do I have to give up to get him? It's a basic optimization-under-uncertainty problem - simple in formulating in theory, not necessarily solving in practice. You don't know who's going to be the best player over the next x seasons, but there are tangible things positively correlated with future productivity, namely height, weight, 40 yard dash times, and the look of the player's body.

Mr. Nalley, quoted in the article, says that teams are just "covering their asses" when they pick the specimens. That may be true at the margin, but my guess is that if you randomly picked 22 "comic book superheroes" and I randomly picked 22 comic book authors, your team would whup my team's ass in repeated games more often than not.

It should be noted that this way of solving an important information problem - who's going to be a productive football player over the next x years - is also employed by college teams. Chase Daniel, the former Mizzou great, Metroplex high school star, and Texas Longhorn fan, was all but ignored by Texas coach Mack Brown*. Brown, instead, had gone after targets who fit the typical mold better. Daniel went on to be one of the best if not the best QB ever at Mizzou while taking Mizzou to unfamiliar heights. Meanwhile, Texas remained a perennial top-10 football program.

*Brown ended up offering Daniel a scholarship, but only after his main QB target, Ryan Perilloux, reneged on an oral commitment.

The NFL Draft: Shotgun Marriages v. Marriages "For Love"

If Phil Miller's Saturday post on the NFL draft is Day 1, then here is a Day 2 examination. NFL writers/analysts love to speculate on draft picks, although a bit of a "draft-grading backlash" has emerged (See PFT on Meaningless Draft Grades.) A particular study of the draft that amounts to more than just hot air is cited by former NFL GM Charley Casserly in a post on Reiss's Pieces (part of Boston.com) and draws from a 10-year study evaluating drafts after four years had passed. Casserly describes the percentage of draft picks, by round, that wind up as NFL starters:
• First round – 75 percent
• Second round – 50 percent
• Third round – 30 percent
• Fourth round – 25 percent
• Fifth round – 20 percent
• Sixth round – 9 percent
• Seventh round – 9 percent

• The NFL draft is an example of a market "design mechanism" -- the particular decision rights, sequences, and incentives that determine choices and transactions. A subsection of this literature focuses on matching mechanisms -- including two-sided matching like marriage, medical residencies, or the free-agent versus one-sided like the draft (see Google Scholar search for a sampling).

One design-outcome question that jumps out from the draft data is whether the 6th and 7th rounds make sense when only 3 players per round become starters. A related question -- linked to this post's title -- is whether the starting rate is higher for undrafted free agents who play a role in the matching process. Does the object of affection having a say change anything? Instead of starting, what about making an NFL roster -- does this data look similar and how does it break down in comparing drafted rookies versus undrafted ones.

The matching issue pops up in media discussions of "fitting the needs" of a team but rarely goes much beyond positional depth. I'm struck, however, by how critical the matching problem can be. Player value can depend heavily on details of the environment. Randy Moss in MN (good), in Oakland (terrible), in New England (good). If Drew Bledsoe stays healthy, does Tom Brady get the opportunity to become, well, Tom Brady? Every Jacksonville Jags receiver in the past several years had been dissed as unproductive but is their productivity a function of poor skills or bad QB, bad O-line, bad coaching? Does having a say in the match change the likelihood of success?

ECONFL

Add another econ major to the ranks of NFL coaches along with Bill Belichick. Jim Scwhartz, new head coach in Detroit and former defensive coordinator in Tennessee, holds a degree from Georgetown. Judy Batista wrote a NY Times article on him last year. Here is Schwartz on his penchant for utilizing statistical analysis:
“If you ask me, Would you rather have a great fly-by-the-seat-of-your-pants guy on Sunday, a guy who can dial up plays and he’d be the best in league, or a guy who is best in the league from Monday to Saturday preparing, I respect the guy who prepares. You’re not always going to be rolling 7, 7, 7 and be hot every week. But if you prepare well during the week, you’ll be consistent from week to week.”
The article spends most of its time focusing on Schwartz's interest in statistical analysis, which brings me to my broader question. What is the value of "theory" in econ training? We may oversell the theory for its own sake and undersell its indirect contributions to our statistical abilities. Yes, concepts ranging from the law of demand to agency problems provide insights into decision. Yet, for most economists, including those who write on this blog, the lion's share of our research efforts are devoted to statistical analysis.

Indirectly, theory develops our sense and skill with "multivariable" modeling problems, particularly fostering the ability to consider and triage a variety of influences. Along with this, it develops skills and creativity at coming up with "identifying" variables and techniques (ways of distinguishing the effect of one variable from another). These stat skills provides a economists with the ability to provide empirical contributions (arguably) to something as seemingly out of our league like autism (see Feb 2007 WSJ article).

Many disciplines utilize statistics to varying degrees but few develop these multivariable and identifying skills as much as economics. This hit me as I struggled with a brain-based neurological problem in the 1990s. At the outset of the 90s, optimism reigned in brain research as more sophisticated scanning equipment emerged. In the end, however, rather than providing answers that biomedicine had been very successful in bivariate causal situations, microbe X causes symptom Y scenario, the new technologies highlighted a world that is much complex and multifactoral than simple modeling techniques, even with such equipment, could sort out -- a world much more like economies and organizations that economist study. It prompted me to do an MBA short course on complex decision analysis.

Oh well, this is a long way from Jim Schwartz and the Lion's new head coach and may strike some as heresy, so I'll stop here.