SB Nation

Bill Connelly | March 10, 2014

The state of football stats

How to build a better Sloan, and improve the conversation about analytics

The first time you attend the Sloan Sports Analytics Conference in Boston, the annual rah-rah hub for the sports analytics movement, you're intrigued by the "What's next?" talks and the stature of the folks with whom you're sharing a room. You see team executives, ESPN personalities, and a whole lot of other folks who make more money than you do.

In your second year, you hear the same panels and the same conclusions, and you watch the same executives dodge the same questions with the same generalities.

In your third year, you realize that the conversations aren't ever going to change and that the panels aren't the point.

When the conference first began, the goal seemed pretty simple: gather the small community of stat nerds (preferably including those in power positions) and celebrate the ways in which each is either changing his or her sport of choice or preparing to. Seven years later, it has gotten the sports world's attention. The big boys attend -- commissioners, owners, etc. -- but over time, the conference has basically become another trade show. You attend so you can have conversations in the halls and at bars, get some face time with powerful people, and listen in on just enough panels and interviews so you know what the talking points are and where others are focusing their time.

Last year's buzz words: signal, noise, process, outcome. And basketball.

This year's buzz words: biorhythms, psychology, optics ... signal, noise, process, outcome. And basketball.

Since I wear a college football hat most of the time, I'm going to talk about the conference, and the stats movement in general, from that perspective.

"Why don't teams go for it more on fourth downs?"

Baseball analytics is like a full-grown adult with a pretty good job. Basketball analytics is the soon-to-be high school valedictorian voted most likely to succeed. Pro football analytics is the second-grade son of the famous businessman who nobody's really sure about yet. (College football analytics: pro football's three-year-old younger brother.) And as a friend put it at a bar on Friday night, "Hockey analytics are sperm at this point."

Pro football analytics is the second-grade son of the famous businessman who nobody's really sure about yet.

This is where we were a year ago, and this is where we still are today.

In recent years, basketball stat conversation has trended in a lot of different directions, from the most efficient shot (the corner three) and offensive style in the sport (threes and free throws, basically), to the breakthroughs of visual data, SportVU, and DataBall. The basketball data community has quickly moved right on past player ratings (PER, etc.) that football will always struggle to create and has impacted the way professional basketball teams choose players and run offenses. The growth has been exciting and has allowed basketball to attract the biggest crowds at Sloan.

Baseball still finds room to build buzz, however. There's a chance that we're seeing a new breakthrough in defensive stats. Saturday's discussion of a new method of fielding tracking created a buzz and, in Rob Neyer's case, a tingle. Breakthroughs like this are why a lot of veterans still come to Sloan -- you never know if there will be anything useful, but you don't want to miss it if there is.

Let's just say we're still waiting on that breakthrough in football. Or we're at least looking for something new to talk about.

My friend Justin Moore of Tempo-Free Gridiron calls it Godwin's Law of football analytics: Eventually every football analytics discussion boils down to going for it on fourth downs. When casually mentioning football during an early panel, ESPN's Nate Silver mentions fourth downs. Rockets general manager Daryl Morey mentions that he doesn't know why football teams don't go for it on fourth downs. It is the equivalent of the "fouling up three late in the game" debate in basketball -- it matters, obviously, but it doesn't dictate a game's worth of decision-making -- but it's about as far as football gets in the general discussions.

It's well-established that NFL teams probably don't go for it on fourth down enough. It's like hitting on 16 in blackjack -- sure, there are times when the math works out, but your salary is on the line, and you'll get crushed and/or feel crazy for not remaining staid and conservative if it doesn't work. Teams leave potential points on the board because they are risk-averse, and it's a shame. This is not debatable. It's the case in the more liberal college ranks as well. Certain coaches are more prone to playing the percentages and going for it, but plenty still punt because that's simply what you're supposed to do.

Sloanchart2_medium College football fourth-down success rates in 2013 by yard line and yards to go.

So yes. Fourth downs are not typically utilized well. Very true. But imagine if every mention of basketball over two days quickly reverted to "fouling up three." It's a discussion that needs to happen; it's also a discussion that needs to step aside so other conversations can begin.

The Football Analytics panel was a complete waste of time

Here's the general layout of a [Sport of Choice] Analytics panel: Team Executive 1, Team Executive 2, Actual Analytics Person 1, and a Wildcard, with ESPN Personality 1 as moderator.

For football, the original lineup just got rid of the actual analytics person altogether. The panel initially consisted of San Francisco 49ers president Paraag Marathe (Executive 1), Atlanta Falcons assistant general manager (and former Chiefs GM) Scott Pioli (Executive 2), Miami Dolphins executive VP Dawn Aponte (Executive 3), Patriots receiver Julian Edelman (Wildcard?), and Pulaski Academy (Little Rock) head coach Kevin Kelley (the never-punts-on-fourth-dows guy and the actual Wildcard). Moderator: Suzy Kolber.

Here's a quick note about team executives in these panels: They're not going to reveal anything.

Here's a quick note about team executives in these panels: They're not going to reveal anything. Of course they're not. How would it benefit them?

Marathe is smart and affable. Pioli's approachable Sloan presence in no way meshes with the cartoon version of the control-everything tyrant we heard about in Kansas City. But in these open-mic situations, all team executives basically use "analytics" as a verb. "Yes, we definitely analytics. We analytics as much as anybody else. Analyticsing is a big part of what we do. There are limits to what analyticsing can do, but we definitely analytics."

SB Nation's Paul Flannery wrote a piece about Sloan for Boston Magazine:

As the conference has gained in star power and cachet, attendees say that the panels have become more entertaining—Dallas Mavericks owner Mark Cuban is always a highlight—but less informative. "As it’s gotten more professional, a lot of the panels have had less to say, because there is an interest in keeping professional secrets," says ESPN analyst and former Indiana Pacers consultant Kevin Pelton.

To find any real insights, you have to sift through the handful of academic papers that are presented each year outside of the main hall—either that, or buttonhole a few stat-heads and try to pry out their secrets over a few drinks. All of which makes you wonder: Has Sloan outlived its usefulness? Has it grown too large to fulfill its initial mission as an academic conference?

"It’s weird to have us all in one place on the one hand," Zarren admits. But then, he says, with so many smart people around, there’s always the chance of picking something up.

I asked [Celtics assistant GM Mike] Zarren, who never goes on the record about anything related to his work, if there were papers or panels that caused him to adjust his thinking about certain issues. He paused for a moment and answered, "Yes, but I don’t want to talk about it."

As the conference becomes more renowned and is given more of an air of importance, panels have more opportunities to attract big-name executives ... who will never actually reveal anything. Or as Jacob Rosen at WFNY put it, "Mostly, the event is a popular gathering place for really smart people who don’t have an incentive to share their intelligence." Meanwhile, some of the most interesting panels -- the Building a Dynasty conversation between former Lakers and Bulls head coach Phil Jackson, the Patriots' Jonathan Kraft, and journalist (and ace conversationalist) Jackie MacMullan, for example -- don't really have anything to do with analytics at all.

It's frustrating. It also isn't a football phenomenon. The Soccer Analytics panel was even more disappointing in this regard, with AS Roma president James Pallotta and former Chelsea analyst Steven Houston (now with Expert Technical Scouting) providing engaging, semi-enjoyable conversation while actively saying nothing interesting about analytics. (Don't believe me? Go to the expert.)

After cancellations from Aponte and Edelman, the Football Analytics panel did get an official numbers guy added: Brian Burke of Advanced NFL Stats, also a 2013 panelist. Burke wrote a pretty strong "What is Football Analytics?" piece last week and attempted to draw the conversation into new, interesting areas. He mostly failed. Kolber, who didn't seem to have prepared more than about 15 minutes for the moderation job, set up the same tired, old strawmen for the executives to tear down (no, numbers can't solve everything ... no, numbers can't measure heart ... no, numbers haven't gotten very far in measuring intangibles ...).

There were highlights. Burke was briefly able to talk about the father of football analytics, former Cincinnati quarterback Virgil Carter, who did work on net point values in a pre-Excel universe. Kelley at one point said that, "College football should do a little better job of finding players," which speaks to the need for projection work that I mentioned in my own Day 2 presentation. Marathe said that "the gap in physical ability in the NFL is very narrow; the gap in mental ability is massive," which describes one of the many differences between the college and pro levels of football. There was eventually some useful conversation regarding injury prevention and the work that needs to be done in this regard (that was a common theme among many panels), and everyone involved declared most of the Combine worthless, which was fun.

But for about 50 of the panel's 60 minutes, this conversation was about deftly taking down strawmen and repeating the exact same "Yes, we analytics" and "There are limits" conversations of each of the past two years.

After a couple of days, it was easy to get discouraged about both the state of football analytics and the seriousness with which the Sloan Conference takes the sport. And because the conference is doing so well, there is no need for any of this to change. But if someone desired improvement, some pretty obvious changes could be made on the football side.

D4c01-0245_mediumThe Football Analytics panel. (Photo by Greg Cooper, courtesy of 42 Analytics)

1. Real moderators. The draw of pulling in a Suzy Kolber or Brian Kenny (for baseball) is obvious. Kolber has obviously worked in the field of NFL journalism for a while now, and she's good at her job. Having her run the panel like an interview makes perfect sense ... in theory. In practice, however, the revolving door of moderators means we start every year's discussion in the same place as last year's. There was a novelty in Kolber's introduction and general conversation -- Numbers, huh? Fun! What should we make of this? -- that ensured that no new ground would be broken even if the participants were willing (and they mostly weren't).

It's obvious that the team executive aspect of these panels isn't going to change, but there is still the opportunity to bring in an analytics person to run the discussion, someone who has maybe attended or watched a few previous panels, knows where the conversation has gone, and knows where the conversation could go. Not going to put Aaron Schatz or Brian Burke (originally) on the panel? Let them run the panel instead. Suzy Kolber is, again, good at her day job, but nobody came to the panel to listen to her. Might as well give that job to someone more qualified to run an actual conversation about analytics.

Of course, of the major sports, the basketball analytics panel came the closest to this by putting Grantland's Zach Lowe in charge of the conversation; he was pretty good at his job, but he couldn't prevent the two Celtics representatives (head coach Brad Stevens and assistant GM Mike Zarren) from saying relatively interesting nothings or Stan Van Gundy from delivering his annual "I like stats, but you nerds don't know shit" message.

2. New blood. Marathe and Pioli had been on Sloan panels before. The only thing more frustrating than an executive trying hard to say nothing beyond, "We analytics!" is seeing the same person say it the next year, too. Only Kelley was new to this discussion, and his bona fides are basically being the guy who goes for it on fourth most of the time. He made some fun points in each of his two panels, but he was the only new one.

Football is still the precocious youngster in the analytics conversation. But while plenty of writers and analysts (and nerds hired by teams) work to break new ground, the Football Analytics panel itself did not.

3. A new approach altogether. If team executives aren't going to say very much in this format, change the format. Maybe create a Health Analytics panel, since that was a recurring topic. Maybe focus the conversation on something more specific within the football analytics universe -- if you want to talk about fourth downs, create a Football Tactics panel; if you want to talk about talent evaluation and the Combine, create a Making the Combine More Useful panel -- so we at least don't have to kill 20 minutes with the same "Numbers, huh?" small talk.

And these complaints are based around pro football discussion. College football talk is still mostly non-existent. There was a panel about judging teams for the impending College Football Playoff, and Kelley did say the word "Auburn" a couple of times, but when it comes to panels, that's about as good as it got.

There is good news, however.

ESPN's on the case

That's not an ironic statement. With just about everything in sports, it has become common to say that ESPN is the best and worst thing going. The Sloan Conference is no exception. The presence of "the Worldwide Leader" at the conference is overbearing and unmistakable, but it has also championed analytics in a major way, for both its own good and the good of the analytics community as a whole.

The fact that ESPN has dived into college football stats can only be considered a good thing overall.

ESPN sets the narrative and does so in a lot of ways that drive you absolutely nuts. But it gets a lot of things right, and the fact that ESPN has dived into college football stats can only be considered a good thing overall.

During the College Football Playoff panel, ESPN's stats guru Dean Oliver said, "Sports analytics were designed for college football." And he was right. You like projection? College football uses projection for two different paths -- high school to college and college to pro! You like tactical discussion and fourth-down debates? Some college coaches actually go for it on fourth like you're supposed to! College football has large differentials when it comes to physical ability and mental acuity, and the range of tactics used to win a game is dramatically larger than at the pro level. College football is desperate for new voices, and the simple fact that ESPN is paying attention, even if only at a rankings-and-win-probabilities level for now, is a wonderful sign.

So how was the Playoff panel? It was ... fine. It was certainly successful as a way to show off some of ESPN's new college football toys -- its FPI rankings, its Win Probability averages, et cetera. And it reaffirmed what we already knew: that the Playoff committee will have some difficult decisions to make at times. But while certain members of the panel seemed interested in delving further into numbers, the conversation was sidetracked at times by ESPN's Brad Edwards, who at one point suggested (with encouragement in his voice) the "non-football" members of the Playoff committee might defer to the "football people" in some regard. Between that and the suggestion that committee members have day jobs but will be encouraged to watch as much football as possible, the misses in this panel were certainly noticeable. But as a Step 1 toward more and more substantive conversation, it could have been a lot worse.

They gave me 30 minutes

Granted, it was at the end of Day 2, when a lot of conference-goers had already departed for either home or the hotel bar. Granted, it was just an Evolution of Sport presentation in front of tens of people. But I did get to give a "Hey, we do exist!" presentation about college football analytics this year. Between that, the Playoff panel, Kelley's presence (indirectly), and some other Evolution of Sport presentations -- one about an automated playbook that tracks player paths on All-22 film, and another from the folks who wrote the interesting Newton's Football -- this is clear and inarguable growth, just as having $25 to your name is technically better than being broke.

College football analytics in 2014

"We are now getting plays off every 12 or 13 seconds," says Ohio State's Woody Hayes. "We are moving so fast I frequently can't get a play in from the sidelines. We'll hit 100 plays a game soon." This, coming from one of football's bastions of the conservative, makes it plain that something big has happened. [...]

"What's happened is obvious," says [Alabama head coach Bear] Bryant, the master of defense. "First of all, due to the pro influence, there are more good pitchers and catchers coming out of high school. They all want one of those Joe Namath contracts. Then, of course, most colleges use their best athletes on offense, as backs and receivers. That's not necessarily true in the pros. They've got some of their best athletes on defense, especially corner-back. When the defense is forced to spread out, it must go to man-to-man coverage. But if the offensive boy—the pass receiver—is a better athlete than the defensive boy, he'll beat him. So you have to go to double coverage, and that weakens you against the run."

That's from a piece the inestimable Dan Jenkins wrote for Sports Illustrated in 1968, a year in which "the total offensive yardage per game (629) [and] the total points per game (39.3) ... all proceeded at a record-breaking pace," and "fast, deceptive Arkansas scored more points (29) against stubborn Texas than Frank Broyles ever has—and lost by 10," and "the top college teams got off about 40% more offensive plays than the leading pro teams." The numbers have changed -- among other things, Baylor averaged 52 points and 618 yards by itself last year -- but the debates that dominate college football haven't really changed. Ever. We were arguing about amateurism in the mid-1950s. We were debating if it was a good thing that passing offenses were setting records in the late-1960s. We were arguing the merits of a playoff in the 1970s.

This is the way college football operates, and as a college football person it is difficult to stay frustrated about Sloan. Three years of the same conversation? A mere drop in the bucket.

In my happy place, however, what would I like to see at a Future Sloan? How would organizers go about affecting a true impact in the football universe?

D4b16-0270_medium The College Football Analytics panel. (Photo by Greg Cooper, courtesy of 42 Analytics)

College Football Analytics panel. I obviously have my own selfish reasons for this, but college football is a sport ripe for analytical discussion, and not simply about the Playoff committee.

Craft a panel around a coach with personality and uniqueness to burn -- a Mike Leach, a Gus Malzahn -- and discuss the topics du jour (pace, the spread offense, and the "continuous game," as Nick Saban has called it). Bring on TCU's Gary Patterson, Virginia Tech's Frank Beamer, or Missouri's Gary Pinkel to talk about unique methods of talent projection. (And bring Kevin Kelley back to once again talk about how he thinks college football misjudges talent.) Supplement the discussion with the requisite ESPN personalities, bring on an analytics person or two, and you've got an interesting hour. Spring football hasn't started for many schools at that point, so you could probably find some interesting coaching candidates.

Building a Better Combine panel. If team executives aren't willing to share state secrets but can agree that having a kicker run a 40-yard dash is pointless (and fun), invite them to talk about what changes could be made to make the NFL Combine a more useful, innovative spectacle. The Athlete Analytics panel certainly took steps in this direction. Involve a current or former college coach (or current pro player, for that matter) to talk about the preparation that goes into the Combine from the player's perspective.

Injury research paper. The college level needs an injuries database. Someone looking into doing such research could find any number of relevant topics to present at Sloan, including a comparison of the hurry-up/no-huddle offenses to slower, more grind-it-out styles. But from one of two perspectives -- either "What styles lead to the most injuries?" (the CFB Matrix recently attempted to tackle this one) or "What impact does the loss of a quarterback/running back/whatever have on a team's output?" -- this is a relevant area ripe for discussion.

High school projections research paper. If only because I want someone to incorporate all the shaky pieces of data in existence -- high school camp numbers, stats, etc. -- and begin the process of projecting how high schoolers will perform at the college level. Are there areas of the country, or certain positions in certain areas of the country, that are a bit underrated or overrated from a recruiting perspective? If you adjust for levels of competition (big-school districts, small-school districts, etc.), is there a feasible way to make something of this data?

Game-charting research paper. We are still plugging away, attempting to clean up the data from the 2013 Charting Project, but when we are through, we'll have 300 games and between 40,000 and 50,000 games charted and available for analysis by request. It's free, it's available, and it's ripe for analysis on game theory, quarterback/passer analysis (not unlike this or this), heat mapping, or tendency tracking.

Further video work. The automated playbook presentation was interesting in its ability to efficiently do what teams already do in the scouting process, i.e. tracking tendencies. This is an exciting development from a speed-of-analysis perspective. The next step is to show what coaches could do with this data that they don't already do.

* * *

Your 2014 College Football Analytics panel

In the absence of an official College Football Analytics panel this year at Sloan, I created my own. Serving as a moderator of sorts, I interviewed three influential football personalities from three different aspects of the sport: Washington State head coach Mike Leach, CBS college football analyst Gary Danielson, and the senior director of ESPN's Stats and Info group, Jeff Bennett.

I interviewed them separately, and I spliced together their quotes below to function as a roundtable discussion. I took care to avoid using quotes out of context. My own thoughts are in italics.

The need for communication

One of the overriding issues every year at Sloan is simply the need for strong communication. It doesn't matter if you have the most perfect set of numbers in the world if you cannot communicate those numbers in a way that allows for a given person -- analyst, coach, player, etc. -- to understand. It is no different for college football stats.

Late in Alabama's win over LSU last November, Gary Danielson referenced a Football Outsiders stat of mine, Adjusted Sack Rate, to both describe Alabama's iffy pass rush and talk about how the Tide put together a good defense without much pressure.

Gary Danielson: We go through our production meetings on Thursday and Friday, and these great kids, these production assistants, throw stats at me that I rarely use. I really have a tough time trying to make it work. I try to grab one for entertainment purposes. "Here's one that might work." To me, stats best tell the story of what has happened, not what will happen. I find it interesting, but I just don't use it a lot.

"Stats best tell the story of what has happened, not what will happen."

I tend to do a game from a strategy standpoint. They've been running the ball successfully; now will they take that running success and take advantage from a passing standpoint?

Because I do one conference [Danielson and Verne Lundquist call the SEC's game of the week for CBS], and the status of that conference doesn't change, I use stats that are more germane. I do not use what I consider outlier stats -- I eliminate Coastal Carolina and those teams. I might throw a wrench in what you're trying to do, but at the risk of being Harold Reynolds, I feel college football stats are the most worthless stats in sports. The stats become meaningless because of the disparity in talent.

Danielson's point is understandable. To get anywhere with advanced stats, one must account for the talent differential, which is immense. Even teams within the same conference can end up with strengths of schedule drastically different. It's sometimes difficult to find meaning in run-of-the-mill raw stats when they don't provide for any context.

Jeff Bennett: The hardest thing for those looking at the numbers is to adjust for the competition they played. The human mind can't do that very well. If I want to look at Ohio State's offense, I want to look at the defenses they played. It's like Dean [Oliver of ESPN] always says: the numbers see every game.

Danielson: If I could get more of the stats that I consider good-against-good and eliminate stats that are meaningless -- inferior-talent stats -- it would be great. Alabama crushes inferior teams, crushes 'em. You don't learn anything from that. You're talking to a guy who's very frustrated by college stats. I don't let it get me down, but I don't put a lot of weight into them.

At Football Outsiders, Brian Fremeau and I have slightly different takes on strength of schedule, but they tend to work for our own respective systems of ratings. I've long espoused a general "It's not who you play, it's how you play" motto, in which you learn a lot about a team even if it's killing a weak team. (If a team dominates that weak opponents more than others do, it still means something.) Fremeau tends to put heavier weight in the formulas on games against better teams. One can understand the logic in either case. Regardless, the major problem with disregarding "inferior-talent stats" is that you're left with an especially small sample, in a sport where the sample is already tiny.

Bennett: From my experience with ESPN, what we're finding in terms of added value is principles we've pulled from other sports -- win probability, etc. A lot of analysts draw on their experiences, but sometimes they only see through observation and lose the context of the situation. We're trying to look at the context, the aggregate of the plays.

We don't want to overwhelm our analysts. We want to listen to what they're saying and try to help them use facts to back that up. Or tell them, "Okay, you see this, but the numbers say this." The way we build relationships with them, and the way we tailor this at ESPN, we're serving both the analytics team and the entire company.


The hurry-up, no-huddle approach to offense has been around in different variations for decades. The two-minute drill has long been a part of the sport, and the HUNH movement brings a lot of those principles to the table for more than just the last two minutes of each half. The hurry-up has been a major topic of conversation this offseason, with one group of coaches pushing to stop the game for substitutions for 10 seconds after each play. That rule was tabled, but the conversation isn't simply going to stop.

Mike Leach: We're a little more limited in using the hurry-up than some. I can't say we've ever utilized it as much as some teams do, but we've never not utilized it. We want to be quick but not hurried.

"Each play is a potential opportunity to score, so we want as many plays as possible." 184261579_medium (Getty Images)

First of all, each play is a potential opportunity to score, so we want as many plays as possible. Second, I think sometimes players can overanalyze plays, and there's a rhythm to it that can be beneficial. There's a point where everything breaks down, and you can't execute that fast if you're not careful. But you can get the defense off-balance; in other words, you can execute your offense quicker than they can respond. They can do the same thing with their line movement and blitzes. There's a cat-and-mouse quality to that.

Danielson: [Hurry-up teams] keep things very simple in their play-calling -- no intricate pass protections -- and very easy to repeat.

Leach: I think defenses are already catching up. I think there's a cyclical quality to football -- whatever goes out of style will come back in style. I think the hurry-up stuff has kind of hit the midpoint. I think there will still be plenty of hurry-up offenses this year, but it's still on the decline. You do it, and if the defense responds well, you get out of it. And the defenses are responding well to it.

I can't say we had a great defense this year, but we didn't have much trouble with the hurry-up. We had trouble with other things, and we need to get older, but the hurry-up wasn't a problem.

Danielson: I do think Nick [Saban, one of the proponents of the 10-second rule] has a point: football was not originally meant to be a continuous sport.

But I think it's stupid coming from Nick, and I don't understand why he's doing it. He has the best players! He theoretically should have the most well-rounded players, and he could look more like the Seattle Seahawks than anybody else in college! But he fights it. Just swallow hard and put guys in there who can do a lot of things. You've got the better players!

Leach: This certainly isn't the first time in history that guys are using the hurry-up. This is just something that has been dusted off, really. There's less teams doing it than were doing it a couple of years ago, but if by moving quicker and executing quicker, you can get an advantage, you'll do it.

Danielson: You could almost bet on this: for every 10 plays, there's roughly three first downs. About 30 percent of plays in college football, then, are run when the clock is stopped for a first down. When they were trying to adjust these rules, they were going about it from the wrong end. The hurry-up is a tactic, just like screen passes and traps and pick plays. Everything is a tactic, but the HUNH is given way too much credit. It's just producing more plays.

Leach: One of the biggest fallacies that exists outside of coaching is the notion that games and game plans are mainly focused on fooling the other guy. "Oh, these guys tricked 'em here and tricked 'em there," and it's a series of trick plays. Like Roadrunner and Wile E. Coyote. On a simple level, it's back to Vince Lombardi -- a game of execution. Teams that are any good are going to be predominantly based on execution, whether you're offense and defense.

Call it "the Boise State Fallacy."

For 10-second rule proponents like Saban and Bret Bielema, the battle has frequently centered around health concerns for the players. The language and analogies used by both were regrettable at times and won't be rehashed here, but the general idea was that if a defense is not allowed to substitute tired players off of the field, health issues could get magnified.

Danielson: I think there's logic behind the health issue. I think the 10-second rule is asinine. The offense should have the option of snapping the ball. But in all of sports, football is the only sport where only one side of the field can control substitutions.

When the offense does substitute, first of all, the receivers are 40 yards downfield, and they can just step out of bounds there and bring in two guys near the line of scrimmage. Make those guys come back to the line of scrimmage, then exit.

Leach: Erk Russell, one of the greatest coaches in the history of college football, he was the defensive coordinator at Georgia. He created the Junkyard Dogs and of course had a great legacy at Georgia Southern. He said [paraphrased], "This business of too many plays, I don't buy into that at all. The defenses are responsible to keep themselves off the field. If you're too tired, let 'em score." There's a simple solution: stop 'em.

"TV was made for football with a huddle. People like to grab some chips, watch highlights. Now it's more like soccer."

Danielson: Here's a pretty easy idea: Stop stopping the clock after first downs! We've discounted the value of timeouts because the two-minute offense is a joke ... to me, anyway. Other people like it. Before, we had a 25-second clock, and coaches complained that officials were not uniformly spotting the ball, and teams were getting penalized for delay.

I said this somewhat jokingly to my team: TV was made for football with a huddle. People like to grab some chips, watch highlights. Now it's more like soccer. If I were king of the world, I'd ask my TV guys, "You hate this, don't you?" And they'd say yes.

On advanced stats in general

The purpose of my Evolution of Sport presentation at Sloan was pretty simple: Talk about how college and pro football are different in terms of acquisition, development, and deployment of talent, and talk about what's been done on the college analytics boat thus far. There aren't many hands on deck just yet.

Leach: You have to keep exploring as much as possible. You're going to keep discovering things. In the end, the value is in what can effectively be communicated to your players. It doesn't matter what you know -- if they can't understand it, it's probably going to be pretty obsolete information. Once they understand it, they have to trigger, execute, and respond quickly. There's always going to be a learning curve that exists with that, but that's the information that has value.

It matters what questions you're asking, it matters how you most effectively present the questions, and it matters who you pose the questions to. What was that Facebook movie? Social Network? One college kid wants to meet girls. The other kid wants to get stoned and watch TV, but he's really good with a computer. The first one says, "Wouldn't it be great if you had a computer deal, and it had profiles, and you could make contact with all this variety of people you have things in common with?" The other guy says, "That's easy." "What do you mean? What about ..." All the first kid sees are limitations. The second guy says, "No, it's easy." He goes over and starts typing. "Shut up for a second, I got this." Twenty minutes later, he's got this thing all typed out, and he's got a financing plan. He could have answered that question a long time ago if anybody had asked.

We discuss a Football Study Hall post I wrote in January, called "The Five Factors." The goal of that piece was to isolate the five pieces of data most relevant to wins and losses: efficiency, explosiveness, field position, finishing drives, and turnovers. They are all interrelated, and work still remains in terms of separating individual, uncorrelated factors from each other. But it was a popular piece, mainly because it makes sense as a concept: Where you start, how you move, how you finish, and whether you get bounces determine your football fate.

Leach: How do you blend it together, and how do you use the information to get the results you hope for? That's going to be interesting to find out. We know turnovers are important, but how do you get 'em? They're not just going to give you the ball.

Sometimes there's a timing to these things that can be important. The variables that exist early in the first quarter might be different midway through the fourth quarter. It's flowing.

But I'm fascinated by it. I see things like this as a resource, as a trend that's going to get better. But if I'm too aggressive with it, it becomes like becoming a religious fanatic. It distracts you from being able to respond quickly to the task at hand.

Information's only valuable if it can be applied practically. And with 22 different players, and with each team's ability to substitute to whatever situation they want to ... how do they apply all this information quickly enough to be effective?

Stats are for losers

It sometimes seems like every coach in the history of football has at some point said, "Stats are for losers." That's mainly because the stats presented to them in media gatherings and whatnot -- "Hey coach, you're 100th in passing yards allowed per game" -- are of no use whatsoever to them. But in reality, they are potentially more analytical and well-versed in observing trends than coaches of any other sport.

Leach: We've got a ton of film and the ability to generate a ton of stats and whatnot.

Bennett: When Pat Fitzgerald came by ESPN last summer, his PR guy was worried about him being a stats-are-for-losers guy. But he was a great conversation. He was very well-informed on concepts like efficiency.

Leach: Sometimes coaches'll say, "In this zone, six percent of the time they run this trick play." But what are you going to do with six percent? They're going to beat you with all the stuff they do 94 percent of the time. There's a point to where you have to respond to their best pitch.

And as a coach, you have to be careful of clouding your mind. Doesn't matter if you know more than your players -- you like to think you do -- but what you communicate to your players, you have to keep their minds clear. It has to be value or else it's just mud.

Bennett: What three plays are they most likely to run on third-and-goal ... you're sorting through the data in seconds, and you can inform coaches in this regard. Heat maps of formations, player personnel.

There are teams dabbling with optical tracking in the NBA, and they're scaling their product for other spots, including football. That's probably mostly on the pro side, but I know Alabama is interested in getting more analytical and using this information more wisely. We can give them all the data they want; we need to do it in the most relevant fashion.

Leach: Of course, in the involvement of all this information, there will be guys in the short run that are ahead.

On the speed of data compilation

As optical data tracking and similar concepts reach further into the college football world, it could basically start to do what coaches already do -- formation tracking, etc. -- only faster. In theory, that could free up time for coaches to work on other things. What do they do with that time?

Bennett: They're still going to rely on their grad assistants and their own charting methods. They're still going to know things about assignments that numbers aren't going to know. The insiders will still know the individuals better. There will be optical tracking, radio frequency tags, etc. Biorhythmic information. We're just at the tip of the iceberg when it comes to football data.

Leach: The biggest change in my career is going from film to video to digital. You can mass-produce the video and educate players in a much faster way. Your players can get tape on what they did well or didn't do well.

In response to a joking question about the free time that buys him, Leach points out that when it comes to practice, teams only spend so much time breaking down an opponent.

Leach: Let's say you have a pass play. The quarterback has a whole bunch of footwork and where he goes pre-snap. The receiver has all sorts of considerations of the coverage, his footwork, his ability to use his hands on a release, where he might find a zone. The running back: does he protect? Is he in the route? How quick does he make the decision to transition into the route? The line pass-protecting.

"It's a constant series of corrections. Scheme isn't particularly difficult."

It's a constant series of corrections. Scheme isn't particularly difficult. I don't think it is, anyway. If it is, you need to adjust your scheme. Everything is a series of corrections. So it's a constant battle of reinforcement and skill development, and it goes beyond recognizing the skills you have to develop. It's how you create muscle memory within the time that you have without over-conditioning or under-conditioning your athletes. No matter how sophisticated it gets, there will always be that.

Football is studied by people of all different shapes and sizes. Their focus and emotional level is constantly changing. No matter how precisely you can eventually evaluate a team, there's always going to be the aspect of constant improvement of your own skill, how to specifically teach.

It's a team sport

One of the trickiest aspects of football is pretty obvious: There are a lot of guys moving with a lot of purposes in a lot of directions. In sports like baseball and basketball, one stat goal is typically to create player ratings. But without knowing everyone's assignments on a given play, it's difficult to create a rating for, say, a left guard. A lot of work has been done on the offensive skill positions, obviously, but even then it's a challenge to figure out how to divvy out credit to each party involved. On a 50-yard pass, how much credit is due to the receiver? The quarterback? Perhaps the line (or running back) that protected the quarterback? ESPN uses charting data available to it -- among others, Stats LLC sells detailed charting data to schools and media -- to get at the answer to that question.

Bennett: With our Total QBR stat, we look at the expected points added or lost on every play. Using Dean's analysis, there's a division of credit.

Danielson: When I was with the Lions, Bubba Baker went to the Pro Bowl for us with the Lions. He had a lot of sacks one year (8.5 in 1982), but I swear seven of them came against [Bears offensive lineman] Dan Jiggetts. I told him he needed to take Jiggetts with him to the Pro Bowl!

The message: Evaluating individuals is hard.

On All-22

Speaking of 22 players, the NFL made All-22 film, the coaches' film, available to viewers for a fee a couple of years ago. Is there any hope of that coming to the college game?

Bennett: Maybe it's eventually made available in major conferences.

Leach: It appears there's a market for it, and I'd be stunned if there's not.

Wouldn't college coaches raise a stink about it? They're still accusing each other of doctoring film (Ohio State 2012, Auburn 2013), and we're well into the second decade of the 21st century.

Leach: I don't think the NFL went to its coaches and said, "Hey, we've got a great idea!" No coaches would have said that's a brilliant idea. So if it's going to happen, and there's a market for it, it's probably going to happen.

In my first three years in coaching, we used 16-millimeter film. There were certain teams that, you'd get a copy of their film, but they'd never let you see them score. One-yard run or 80-yard pass, you never got to see them score. Even now, a couple of teams don't let you see the scoreboard. So you have to chart whether it's first down, second-and-three, whatever. When video first came out, you'd get video from somebody, and they'd screw with the tracking. "Yeah, we sent 'em the film." Or they'd send it a day late, and you're dealing with a tight turnaround. It could have been accidental. Maybe the tracking was an accident, or the flight really did get canceled.

On efficiency

Back to stats. One of the initial conclusions I'm drawing from fleshing out my Five Factors idea is that perhaps success rates and efficiency are worth far more than originally thought. The idea of Success Rate -- deeming every play a success or failure based on a given definition (gaining 50 percent of necessary yardage on first down, 70 percent on second down, and 100 percent on third or fourth down) -- is both easy to understand and pretty easy to calculate on your own.

"You want to have explosives. They're like home runs, and routine plays are like on-base percentage."

Danielson: Mike Leach believes in routine plays being the major factor. Success rates are really something you might want to highlight.

Leach: Explosives are important. You want to have explosives. They're like home runs, and routine plays are like on-base percentage. At the end of the day, on-base percentage is going to come out on top, but the one thing about explosives is that you can put it out of reach really quickly. But it seems to me that the power hitting is quite a bit streakier than the on-base percentage.

From a stat standpoint, the better adjective might be less reliable.

Leach: I think the reason baseball is used as often as it is, is that it's a great stat game, and everybody's doing things from more of a set position. Football's hard because 22 players are moving in 22 different directions simultaneously. Finding a way to capture that is a little tough.

But I think all of this ... some of it's fun, some of it's academic, and some of it's very practical. I saw an article years ago, from when we landed on the moon. In the quest to put a man on the moon, this article listed all of the innovations and technical progress that took place as a result of chasing that goal. It wasn't going to achieve perfection or perfect predictability. There were just a lot of innovations to make.

You keep asking questions, and I think it's surprising what you'll find.

Producer: Chris Mottram | Editor: Jason Kirk | Title Photo: Greg Cooper, 42 Analytics

About the Author

Bill Connelly grew up a fan of the Miami Dolphins (post-1970s glory), Pittsburgh Pirates (ditto), Portland Trailblazers (ditto again) and Missouri Tigers. That he still enjoys sports at all shows both severe loyalty and a potential personality disorder. He spends his evening playing with excel sheets and watching DVR'd football games from ESPN Classic. See more of his work at Rock M Nation, Football Outsiders and Football Study Hall.