"Moneyball": the impact of statistical analysis on the NFL, led by the Pats

[mayoclinic]

An interesting read from Pro Football Weekly:

Fans and media have asked the question for years, wondering when the sabermetric world of baseball — brought to the mainstream via the bestseller list and Hollywood, through “Moneyball” — would migrate to the NFL. Something certainly has been lost in translation with advanced numbers crunching between the statistically integral nature of baseball, a sport comprised of one-on-one matchups, and football, a team sport where 11 players (in theory) work together toward a common goal. Another problem: Numbers often lie, or at least mislead, in football as adjunct metrics. It’s not always as clean and quantifiable, we’ve found, as it is in baseball.

However, there clearly is something to be gained from advanced statistical analysis. It’s being done around the National Football League, perhaps slightly on the margins, but the connection is being made with more teams than you would think. Its integration into football culture has been slow but progressive. Numbers-based systems are now leaking their way into mainstream football thought. More interestingly, NFL teams are paying more attention to statistics and how to use them better. Some teams — not all, certainly — have developed a utility for these numbers in how they craft their rosters and make strategic decisions in games.

Full-scale revolution? Hardly. But it’s another tool in the shed, and an increasingly important one, too. As a new generation of football thinkers take over head-coaching, coordinating and front-office positions, the NFL has become more open to out-of-the-box ideas, using fascinating statistical evidence to debunk many long-held football axioms.

Clearly, not every team has bought in yet. But we’re seeing a wave of new thinking in the NFL where it’s becoming more vogue to go against the norm when the numbers call for it. And it has a chance to change the way the game is played, too, when it comes to strategic decisions such as going for it on fourth downs, onsides kicking before the fourth quarter and going for two-point conversions in less-than-traditional circumstances. The sometimes-contrarian Patriots would have to be viewed as one of the NFL’s leaders in this regard.

At the heart of it, that’s what "Moneyball" was: a statistical approach to take advantage of market trends and add “value” players who add more bang for the buck. National Football Post president Andrew Brandt, a former NFL executive, recently estimated that 6-8 NFL teams have hired full-time statisticians to analyze data as supplemental tools for the franchise.

ProFootballWeekly.com - Numbers crunching starting to change NFL thinking

The gist of the article is that statistical analysis combined with detailed game or scouting tape can be used on multiple levels: for personnel evaluation and roster building, for pre-game planning and opponent-specific strategy, and for game-day adjustments and play calling.

Former econ major BB has been into statistical analysis for years. Not surprising, the Pats figure prominently in the article:

Bill Belichick might have been inspired by a Cal professor’s academic paper when he decided to go for it on 4th-and-2 from his own 28-yard line, up six points late against the Colts back on Nov. 15, 2009. Forget for a minute that the attempt came up short by a foot or so — we are seeing more teams going for it on fourth down in untraditional circumstances in recent seasons.

The Patriots have been going against the grain for years, likely a combination of playing to their own personnel strengths but also likely cognizant of the numbers that show that there are some old ways of thinking that need to be retired for the time being. Other teams appear to have the same sort of numbers-based bent.

Let’s see how often the Patriots used ’12 personnel’ (one running back, two tight ends) last year. Within an instant we have our answer: a whopping 65.5 percent of the time, far greater than any other NFL club last season, no doubt because of talented TEs Rob Gronkowski and Aaron Hernandez. From there, we are given more numbers: the percentages of time the Patriots ran the ball and threw it from that personnel grouping, along with how successful they were doing each.

Want to watch tape of Patriots seventh-round CB Alfonzo Dennard covering Saints fourth-round WR Nick Toon when Nebraska played Wisconsin? Boom — all of Toon’s targets and catches are time-coded and noted for when Dennard was covering him and when it was another Huskers DB. Cumulative statistics show that Dennard held Toon in check and that it was others who were responsible for four of his five catches, 94 of his 98 yards and his one TD catch in that game last fall.

Besides the Pats, the article notes the Packers, 49ers and Ravens among teams with an increasing focus on this kind of analysis. It's an interesting and provocative read, well worth reading the full article.

 

[captain stone]

What was wrong about 4th/2 was the play called on 3rd/2: a pass instead of a run, and which was almost intercepted. A Pick-6 would actually have been preferable to what ended up happening, anyway.

2 runs right down the Clot's pencil necks would've sealed the deal.
Sorry to open an old wound, because I am otherwise in general agreement with the OP.

 

[cmasspatsfan]

What was wrong about 4th/2 was the play called on 3rd/2: a pass instead of a run, and which was almost intercepted. A Pick-6 would actually have been preferable to what ended up happening, anyway.

2 runs right down the Clot's pencil necks would've sealed the deal.
Sorry to open an old wound, because I am otherwise in general agreement with the OP.

Thats a good point thats been and its been made before but I still think it was the right decision to go for it on 4th and 2, I know the board is probably still divided with some saying it ws definately the wrong call and the other thinking it was right.
Thanks for the article, interesting read.

 

[mayoclinic]

Aside from the in-game statistical analysis and decision making, I find the other 2 uses mentioned in the article even more fascinating: using statistical analysis combined with tape in personnel evaluation, and to project opponent tendencies. The article notes:

The system brings together the past — tape study and scouts’ observations — and marries it with the future, diagnosing complex situational numbers and making sense of them. And the beauty is in the flexibility: Teams can use and combine whatever numbers they believe are important and manipulate them in dozens of ways to help them shape their decisions.

If a general manager wants a tall receiver who does good work in the red zone, he can search for free agents (even players not under contract are included in the database), their production inside the opponents’ 20-yard line and then sort them by height. If it’s a third-down pass rusher a team seeks, it can measure sack and pressure productivity on a per-play basis — on any down or in any quarter of games, for that matter.

The numbers apply to NFL players, both under contract and unsigned going back several years, as well as a college database that STATS has been building the past few seasons, starting with the major conferences and branching out with each passing year. Pretty soon, STATS will have an entire college career’s worth of tape and advanced numbers on almost every player going up for the draft.

The statistics obviously don’t tell the whole story. Perspective is always a good balancing act. You might be surprised, for instance, to find out that on a per-play basis, Evan Royster was the most “successful” running back in the NFL last season (but then again, he could start for the Redskins this fall). Or that Josh McCown was a more efficient quarterback than Philip Rivers, Michael Vick, Matthew Stafford or Tony Romo. The numbers alone can’t live in a bubble.

But they can paint a very interesting picture of which players truly impacted the game the most, rather than by leaning on sometimes biased scouting reports or player or team reputations that influence, despite the numbers telling a different story. And with a larger sample size — say, an entire season, or two seasons’ worth of data — one can develop some telling trends about players and teams and which ones truly are great in what situations.

I could see the Pats using that kind of tool in analyzing opponents and figuring out who is more likely to be used in what way in what situation. I could also see them using it in personnel analysis. The draftniks project Tavon Wilson as a late round pick but the statistical analysis of his college game tape reveals that he had a significant impact on the field on a high percentage of plays, which may not show up on the stat sheets. Chandler Jones had only so many sacks, but the statistical analysis corrected for injury recovery suggests that he would have a much greater impact. Statistical analysis of the Alabama defense suggests that Dont'a Hightower had a much greater impact on the overall play of the defense than Courtney Upshaw. The possibilities are endless, if there's enough tape and the program is sophisticated enough.

Of course it's only a tool. BB - the ultimate "moneyball" guy in the NFL - has been open about the limitations of such analysis and the fact that it ultimately comes down to football players making plays. In an interview last fall he downplayed the use of statisticals with typical Belichick dry wit:

“In the end, it comes down to – you’re at all the games – it comes down to blocking and tackling and running and throwing and catching and kicking and solid fundamentals and all that. You could put the iPad on the super-duper wizard computer and whatever you want. You could throw all that crap on there and I’m sure it would come out great. I’m sure you could get some statistical analysis that would provide 28 theses for MIT. In the end, you have to go out there and play football.”

Then the real revelation of the day: Bill Belichick is not a big fan of ‘Moneyball’ in football. Yes, he has people like senior advisor Floyd Reese, consultant and confidant Ernie Adams and personnel chief Nick Caserio who due diligent football research. But in terms of the games themselves, Belichick hasn’t forgotten that it’s the physical talent that wins out in the end.

“Personally, I wouldn’t lose too much sight of that,” he said. “Same thing in baseball. Tony [Tony La Russa] and I, we’ve have talked about that a lot too. You have to throw the ball, you have to hit it, you have to catch it, you have to field it, you have to run the bases. You could go out there and talk about some guy’s batting average when the count is 2-1 at night. I mean, sooner or later, you have to go out there and play. I think you have to keep sight of that.

“As you know, I’m not the most technological person in this organization. I rely on some other people to try to help, like you said, streamline things or find a way where we can do things a little bit more efficiently. I understand that the people coming into the organization, that’s what they were brought up on and that’s not what I was brought up on, so I understand there’s a difference there. I think there’s a marriage but at the same time, I don’t think that’s the highest priority.”

So, I asked, have you embraced technology as much as you’re going to?

“I probably embrace it a little bit more now than I did a few years ago – at least I can turn it on,” Belichick said.

http://m.weei.com/sports/feed/blog/...-isn’t-bringing-‘moneyball’-nfl-anytime-soon

 

[patsfaninpittsburgh]

One wonders if giving up meaningless yards up 3 TD's is statistically significant.

 

[MoLewisrocks]

Steve Belichick was a pioneer in the statistical analysis of in game probability scenarios. Ernie was fascinated with his analysis and sought out his son at Wesleyn for that reason. It's a little different than the moneyball approach of analyzing individual players (as opposed to player usage and play success situationally). And while I do think he considers statistical data about individual players, he mostly relies on film because at the end of the day it matters not how fast you time on speed and agility tests in shorts and tee shirts against air after months of focused training, it's how fast and consistently you play in pads against better competition.

 

[patfanken]

Thanks for such a provocative and interesting thread, Mayo. A very enjoyable read so far. I think the article makes an important point that highlights the critical difference between Baseball and Football. It the relative effect of the "team" vs the individual for each sport.

Football, IMHO, is the ultimate team sport. I can't think of any other team sport where the where the sum is more often greater than the individual parts. Nor is there another sport where coaching and individual game planning can even the playing field . A sport where the most talented players don't always win. The most talented TEAM does.

All that makes trying to quantify and crunch numbers much more difficult than Baseball, which is more a game about a series of INDIVIDUAL contests. Plus you run into the PFF problem of how you determine what is success or failure. Its often either too subjective, and/or too hard to determine the underlining factors in assigning praise or blame. For example was it the WR's fault for the pick or was it the QB's. Only the coaching staff would know for sure, because they would know all the particulars necessary to make that determination. Any outside source would be just guessing.

Breaking down film certainly tells you a lot. But it can't be broken down to the point that it will tell you how the individual will project into the team dynamic. and how that team dynamic (or lack thereof) will effect the ultimate outcome of the game.

This "Moneyball" stuff can be useful, but only within certain boundaries and contexts. I think BB seems to striking the balance. These are numbers to "consult" on,factors to be aware of, but not necessarily to rely on.

 

[mayoclinic]

Thanks for such a provocative and interesting thread, Mayo. A very enjoyable read so far. I think the article makes an important point that highlights the critical difference between Baseball and Football. It the relative effect of the "team" vs the individual for each sport.

Football, IMHO, is the ultimate team sport. I can't think of any other team sport where the where the sum is more often greater than the individual parts. Nor is there another sport where coaching and individual game planning can even the playing field . A sport where the most talented players don't always win. The most talented TEAM does.

All that makes trying to quantify and crunch numbers much more difficult than Baseball, which is more a game about a series of INDIVIDUAL contests. Plus you run into the PFF problem of how you determine what is success or failure. Its often either too subjective, and/or hard to determine the underlining factors in assigning praise or blame. For example was it the WR's fault for the pick or was it the QB's. Only the coaching staff would know for sure, because they would know all the particulars necessary to make that determination. Any outside source would be just guessing.

Breaking down film certainly tells you a lot. But it can't be broken down to the point that it will tell you how the individual will project into the team dynamic. and how that team dynamic (or lack thereof) effect the ultimate outcome of the game.

This "Moneyball" stuff can be useful, but only within certain boundaries and contexts. I think BB seems to striking the balance. These are numbers to "consult" on,factors to be aware of, but not necessarily to rely on.

I completely agree. That's part of why I find this kind of thing more interesting in terms of personnel selection and development than in actual game day planning and adjustments. The players have to make the plays, and they have to do so as a team. But this kind of stuff may help identify guys who are more likely to have the kind of skills needed. Add in interviews and personal workouts and discussions with college coaches, and you can get a fair amount of information. It's just a tool, so like any tool it can be used inappropriately or relied on too much. But it's interesting, and possibly useful if appropriately balanced.

 

[PatsWickedPissah]

Data mining has the potential of finding heretofore invisible Vrabels. But once you have the #s, it's imperative to go take a long look at the film to discern whether you've stumbled on a statistical fluke or a pony under the barn pile.

I wonder how many morons will stumble on this phrase and cite it as proof of BB "cheating"?

Let’s see how often the Patriots used ’12 personnel’ (one running back, two tight ends) last year. Within an instant we have our answer: a whopping 65.5 percent of the time, far greater than any other NFL club last season,