Using data to predict arrest rates of NFL draft picks

The final lap has begun. The NFL draft is less than three weeks away, and for the most part, teams have solid evaluations and a good idea of when players will be drafted. Decisions are made and priorities are settled.

These last harried days are consumed with double- and triple-checking some of the most difficult assessments: How to judge players who have encountered off-the-field trouble or otherwise shown a concerning character trait.

This spring, teams must judge Eastern Kentucky pass rusher Noah Spence, who was banned from the Big Ten because of extensive drug use. They must decide if they can trust Ole Miss defensive lineman Robert Nkemdiche, who has a marijuana possession charge and a drunken fall from a hotel balcony on his record.

Absent an alternative, teams are doing what they always do. They're interviewing the players, speaking to past coaches, combing through public records and making subjective judgments. But what if there were another way to project future behavior? Could a data set serve as a predictive tool if filtered properly and analysed appropriately?

A group of college professors and researchers has studied that question as part of a paper on off-duty deviance in professional settings. Their peer-reviewed work was published this month in the Journal of Applied Psychology.

There were two NFL draft-related results. First, that between 2001 and 2012, players with publicly-documented pre-draft arrests were nearly twice as likely to be arrested after reaching the NFL than those who had not been arrested. The second, which is perhaps less obvious and more valuable, was that there was a small but clear correlation between arrests and Wonderlic tests scores. Players who scored below the mean in the researchers' sample were also about twice as likely to be arrested in the NFL as those who scored above it.

"The effects are relatively small," said author Brian Hoffman, an associate professor and chair of the industrial-organizational program at the University of Georgia. "But it's important here because when making multimillion-dollar decisions, a small effect can be very meaningful. A player's getting a four-game suspension can be a big deal, competitively and financially."

Not all NFL decision-makers embrace this type of data, in part because it provides probabilities rather than absolutes. But shrewd general managers would be wise to incorporate these findings into their evaluations. It won't give them any more certainty than traditional methods of homework and due diligence, but it could help confirm or give them reason to question their subjective conclusions.

"If I were a decision-maker, I wouldn't view getting into trouble as a zero-sum game," Hoffman said. "You check off that they've been in trouble and know what that has meant for others on a percentage basis. And then there's a factor that would make the likelihood a little worse: If they score lower on the Wonderlic. Really, that tells you there's even more work to do there."

Louis Riddick, an ESPN analyst who spent 12 years evaluating personnel for the Washington Redskins and Philadelphia Eagles, accepted the premise of the study while offering some important caveats.

"It does reinforce some of the things that common sense would lead you to believe," Riddick said. "That someone who has not made good choices in the past and doesn't understand the connection between actions, decisions and consequences is someone you want to be worried about."

But, Riddick said, what the NFL draft pool really needs is a shift in the way teams handle draft picks. Rather than attempting to project troublemakers, NFL teams would be better served by emphasizing post-draft development.

"Everyone is looking for ways to predict future performance," Riddick said, "whether it is through three-cone tests, shuttle drills or anything else. Then they assume that these guys have it figured out once you get them, and they don't. That's where the focus should be, helping them be better people and players, rather than hiring a psychologist to tell you which players are more or less likely to get arrested.

"Because the reality is, you just don't know, and you can't know when it comes to human nature. It's so hard. You can talk to everyone, from the friends to the coaches to the gas-station attendants, and so many times you're literally just holding on and hoping you've done your best evaluation. Human development off the field is something that is lagging in this league and has to be ramped up and taken more seriously."

There are some cases, of course, when the data simply confirms the obvious. By most accounts, NFL teams have scaled back their expectations for Nkemdiche. Although he is a top-five talent, "it's hard to get somebody to say something really positive about Nkemdiche right now," ESPN draft analyst Todd McShay said earlier this spring. Ultimately, Nkemdiche will join a team that is fully aware of the risks he brings.

But what about the more complicated cases? The numbers in this study are not overwhelming, and they account only for documented arrests and not other character issues such as missing meetings, insubordination or failed NCAA drug tests. What can the data tell us?

Consider Spence as an example. The Big Ten banned him for failed drug tests while he was at Ohio State, and after he transferred to Eastern Kentucky, he was also arrested last May on the relatively minor charge of alcohol intoxication in a public place.

Spence has said he "got caught up in the college lifestyle" and, like every prospect in his position, insists his problems are behind him. The research tells us that players with his history during the sample period had a 21.9 percent chance of getting arrested in the NFL. Players who weren't arrested in college had a 13.1 percent chance of legal trouble in the NFL, according to the calculations of co-author Brian Lyons, an associate professor at Elon University.

Wonderlic results remain confidential in most cases, but the researchers secured scores from the 2002 and 2003 drafts after they were inadvertently posted on a public website. For that period, they found a mean grade of 21.7 on a scale of 1 to 50. Approximately 18 percent of the players who scored below the mean were arrested in the NFL, while 9.5 percent of those who scored above it were arrested.

It's fair to debate the extent of the numbers' relevance, but Hoffman said that the results were not out of line with those of other industries.

"More generally," Hoffman said, "we find a relationship outside the NFL between intelligence and arrests as well. Intelligence is associated with impulsivity and foresight to extrapolate the implications. Those who have more intelligence have more self-control, are more cognizant of the consequence of actions and more likely to abstain. I'm not sure that's what is going with NFL players, but many of the crimes we see are crimes of impulse rather than premeditation -- you get mad and punch someone. In that context, this has some relevance."

If you're looking for a foolproof way to identify future NFL troublemakers, you won't find it here -- or anywhere. But as the chart compiled by ESPN research specialist Vince Masi shows, the NFL is skewing younger for reasons that are both obvious (the rookie pay scale) and ominous (concerns about future health).

The impact, both on the field and through any absence, that draft choices can have on NFL teams has never been greater.