Baseball-style analytics becoming a part of NFL decision-making – Milwaukee Journal Sentinel

Football analysts seem to agree that the reason Packers GM Ted Thompson has succeeded in building rosters is because he values draft picks - the more picks the better the odds are to make a good pick.








By Tom Silverstein of the Journal Sentinel


Imagine someone walking into the offices of Bill Belichick or Ted Thompson or Kevin Colbert and telling them that all things being equal, their chances of outperforming the other 31 teams in the 2016 NFL draft were slim to none.

“The way you’re evaluating talent doesn’t work,” they would hear.

Now imagine the guy entering the domains of three of the most successful personnel men in the NFL today is carrying reams of printed graphs, flowcharts and bar graphs thick enough to put any team’s playbook to shame.

(Think of the scene in the movie “Moneyball” in which actor Jonah Hill sits down with a roomful of trusted baseball scouts and acts as if everything they’ve been doing for the past 30 years is irrelevant.)

Chances are he’d be thrown out on his pocket pouch.

It might not be happening like that in the offices of the New England Patriots, Green Bay Packers or Pittsburgh Steelers quite yet, but the kind of analytics that has taken over baseball is definitely knocking on the door in the NFL.

And it may just be a matter of time before it busts through.

“There’s always an opportunity for it,” said Aaron Schatz, editor-in-chief of FootballOutsiders.com, one of only a handful of websites devoted to football analytics. “NFL teams are already using it, but they just aren’t letting on to it. Just because you don’t hear them talk about it doesn’t mean they’re not doing it.

“Lots of teams are criticizing it in public and using it in private.”

The analytics community is critical of the way football teams do business in the same way it challenged the deep-seated scouting practices in major-league baseball almost 15 years ago. It believes scouting in the NFL is not advancing and there is a need to modernize the approach.

Football is the last major American sport to jump aboard the analytics vessel, mostly because of the cone of silence teams have built around their process. The problem, analytics people will tell you, is that they’re all doing it the same way and no one has an edge over the other.

“With the amount of data they have and the amount they’re investing in scouting, teams should draft better,” said Michael Lopez, assistant professor of statistics at Skidmore College who analyzes football. “The first round should be better than the second and the second better than the third and so on.

“We should be at the point where all the picks in the first round are good picks.”

The analysts believe statistics matter — a lot. But football doesn’t provide the sample sizes and easily decipherable outcomes baseball does. While you can use statistics to compare any two hitters’ ability to get on base, you can’t use them to measure the difference between a quarterback and a safety because their statistics are apples and oranges.

Football analytics entails the use of different mathematical tools to create sensible conclusions.

Through regression analysis, for instance, an attempt is made to compare players of different backgrounds, hopefully turning an apple into an orange for the sake of comparing two oranges.

By using this method, statisticians can create a model that will predict whether the Los Angeles Rams should pick North Dakota State quarterback Carson Wentz or California quarterback Jared Goff with the No. 1 pick.

Comparing statistics for the two wouldn’t tell you much. Wentz played against far weaker competition in the Football Championship Subdivision (FCS) than Goff did in the Football Bowl Subdivision (FBS). But as Schatz explains, you can create values for the positions around each quarterback, both of those with whom he plays and those with whom he plays against.

The values can be plugged into an algorithm that adjusts their statistics for the variable between FCS and FBS levels of talent.

“Goff is better,” Schatz said. “He’s got more experience, he’s got better stats and he played a harder schedule. He had more offensive talent around him than Wentz but not much more.”

The same kind of computing can be done for pass rushers or nose tackles using other kinds of criteria such as times the player was blocked one-on-one, talent level of the players he beat off the line of scrimmage, times the player was held, times he created a pressure but didn’t get a sack and times he used speed vs. power. The algorithms can produce numbers that allow general managers to compare a player with others in the upcoming draft or previous drafts.

It isn’t a slam dunk that Goff will have the better career, but neither would it be if he were scouted with traditional methods. What analysts seek to do is add another dimension to the scouting process that over time will lessen the margin of error.

“The analytics coordinator needs to have equal footing in how their input is implemented, including in-game decision-making, player evaluation, draft research and game planning,” wrote Trey Causey, creator of the blog “The Spread” and a consultant to at least one NFL team. “This does not mean that the analytics coordinator trumps the football people, only that their voice is an equal one.”

The analysts believe mathematical interpretations need to be added because current scouting practices aren’t progressing. In their opinion, the NFL draft is pretty much a crapshoot.

“If you look at the performance one year to the next, there’s very little correlation between the two,” Lopez said. “A team that drafts well one year is not necessarily going to draft well the next year.”

Among the conclusions the football analysts seem to agree upon are:

■ The reason Thompson, Belichick and Colbert succeed in building rosters is because they value draft picks and understand the more picks you have the better your odds are to make a good pick.

■ First-round picks are weighted far too heavily in the draft because so few of them are sure things, and if you can trade one for more picks your odds for success are better than if you were taking one shot at a premier player.

If you consider Thompson’s drafts, you can see what the analysts are saying. The same guy who fell in love with Clay Matthews in 2009 drafted Justin Harrell in 2007. The same guy who drafted Aaron Rodgers 24th in the ’05 draft selected Brian Brohm 56th in the ’07 draft.

You can go on and on.

Analysts are not saying that Thompson isn’t good at his job. They are saying that traditional scouting methods make it difficult to outperform expectations year after year. All the teams have equal access to information. Personnel people move around the league all the time and so ideas get widely circulated. Draft philosophies become apparent over time and can become the norm.

In a presentation to the 2012 MIT Sloan Sports Analytics Conference, Cade Massey, practice professor in the Wharton School’s Operations, Information and Decisions Department, co-authored a draft study that concluded that randomness dominates the process because it’s so hard to predict human performance years into the future.

He concluded that statistically just about everybody is equal in drafting.

Thus, his recommendation was that teams increase their number of picks in order to improve their odds and don’t pay for the right to move up in the draft because you think you’re smarter than everyone else.

To this end, Thompson is a success.

He hoards picks — through trades and the compensatory system he has had 104 in 11 seasons, or an average of two extra per year — and thus gets more shots at being right. (In the same time span, Belichick has made 100 picks and Colbert 93). And so even in those down years, he has managed to find at least one really good player.

“It keeps Green Bay at the top of the league,” Schatz said. “That and having a great quarterback for the past 25 years.”

In recent years, NHL and NBA teams have started using analytics both to scout players and to improve their in-game decisions. The NFL has been much slower to adopt it. But more and more teams — including the Packers — are hiring statistical analysts to help them gather and interpret data.

Most of them are helping coaches understand trends and re-evaluate potential outcomes to help them make better decisions — the analytics world thinks coaches should go for it more on fourth down. Some teams use them to evaluate practice and game stress levels to help reduce player injury. Most of them work on the football side and not the scouting side of the organization.

But Thompson revealed in his pre-draft news conference that he was branching out into analytics in player evaluation.

“There’s a lot more information now,” Thompson said. “Sometimes that’s good, sometimes that’s probably not so good. There’s all kinds of analytical stuff that you need to have someone paying attention to. Maybe not necessarily me, but somebody else paying attention to it. It’s more work. It’s more comparisons.”

Getting someone in the NFL to talk about how they’re using analytics is like trying to convince a stranger to give you their PIN. Even with the promise of anonymity, several executives known to be using analytics declined to speak on the matter.

The biggest breakthrough in the NFL in analytics was the Cleveland Browns’ hiring of Paul DePodesta, who was brought in to be the chief strategy officer. Hill’s character in “Moneyball” was based on DePodesta during his years as Oakland A’s general manager Billy Beane’s right-hand man.

DePodesta’s hiring was met with raised eyebrows in the NFL, but he has assured Browns fans that he is not applying the same rules to football that he did to baseball. His job is to use analytics to improve decision-making.

“It’s not really about numbers or algorithms,” DePodesta told the Browns website. “For me, it’s really just about a mind-set and the mind-set about trying to use information to make better decisions, especially in the face of uncertainty, which is what all of these professionals sports are really about.”

Former Packers general manager Ron Wolf sees little value in analytics as a rule. In retirement, Wolf spends his winters in Florida and often frequents the St. Louis Cardinals’ spring training facility in Jupiter.

There he spends many hours commiserating with former Cardinals manager Tony LaRussa and general manager Walt Jocketty. One can only imagine the discussions the three have about scouting talent, making draft decisions and running a franchise.

None subscribes to the “Moneyball” philosophy.

Wolf said the closest he came to such a thing was an experiment Oakland used incorporating physical testing numbers.

“When I was with the Raiders in the ’80s, we tried it with height-weight-and-speed guys, guys that tested very well at each position,” Wolf said. “We tried to draft those guys, the high-testers. And not one of them made it. Not one. Seriously.

“I don’t know enough about analytics in football. I don’t know how you can incorporate that. As you well know, you can take a statistic and use it any way you want. You can twist that anytime you want to do that.”

Many teams use a form of analytics in the form of Pro Scout Inc., a scouting service run by Mike Giddings, a former University of Utah head coach. The service provides independent evaluations that teams can use to cross-reference their own evaluations.

But it’s not the same as having someone in-house.

Another former GM, who is not authorized by his organization to speak publicly, said, “We all use some kind of analysis. Baseball has taken it to a new level. I don’t know how much of it is going on in the NFL. But it may be a way to get the data into your hands quicker.”

Many teams already go back and study their drafts to assess mistakes in judgment. The analysts want to be able to help avoid those mistakes before they happen. Right now, no one other than the Browns seems to be all in on analytics.

Causey proposes other teams dive in.

“Organizational diversity is good and leads to better decision-making,” he wrote. “Hiring exclusively people with close personal connections to the game leads to groupthink and conventional wisdom.”

It’s going to be hard for analysts to convince the NFL they’re missing the boat. Traditional scouting is a time-honored profession and teams aren’t likely to let go easily.

Which may explain one NFL naysayer’s question about analytics: “How many World Series has Billy Beane won?”


Tom Silverstein thumbnail