Last year, Fooch interviewed Managing Editor of Football Outsiders (FO), Bill Barnwell, as part of their media blitz in support of Football Outsiders Almanac 2009. That interview will long be remembered for Fooch making the statistical case that the Nolan-era 49ers were well-coached, and Bill (understandably) not playing along. Obviously, someone with better stat chops -- which, incidentally, I ate just yesterday for dinner -- needed to take over the FO interview responsibilities. In the interest of site credibility, Fooch recused himself from the 2010 interview, and bestowed that responsibility on yours truly. Well, Football Outsiders Almanac 2010 is now available for purchase, so it's time for me to fulfill my duty.
OK, so maybe I did the interview only because Fooch has been holed up in a bunker somewhere the past 2 weeks while studying for a test I'm told is a prerequisite for joining the Knights Templar. Fair enough. If Fooch wants to be in Dan Brown's next book, more power to him. I still like my version of history better, though.
So how's this year's interview different from last year's? Well, in true "Danny's posts are too long" style, I hyper-evolved it from a 5-question amoeba into a 20-question T-Rex, and will be posting Bill's answers in 2 parts rather than 1. Part 1 has landed in front of you right now, and Part 2 will be descending into this blogspace tomorrow. The plan here is to let the questions and answers speak for themselves, so I won't be adding any additional commentary to the posts except for one after-the-fact edit and the brief -- yeah right! -- introductory comment I'm about to give. Here goes...
After the jump, my introductory comment and Part 1 of the interview, whererin Bill (a) somehow finds a way to marry the concepts "49ers front office" and "intelligent ways to spend money;" (b) tells us about the rarefied air Dashon Goldson was breathing last season; (c) provides bait for the Smith-haters; and (d) assures us that, yes, Adrian Peterson is indeed a better RB than Pierre Thomas, regardless of what DVOA says...
A SPECIAL COMMENT
Aside from Alex Smith's bustworthiness and the future that was Kory Sheets, no topic stirs up more Niners Nation debate -- heated at times -- than the work of Football Outsiders. As I've said repeatedly, DVOA, nor anything else put out by FO, is on par with the books written by Saints Matthew, Mark, Luke, and John. Therefore, unlike those edicts from on high, FO's work is open to criticism wihout critics having to fear divine retribution or a date with the Iron Maiden. Indeed, although I'm one of their biggest supporters, I've been, at times, one of their biggest critics. And yet, somehow, I'm still alive. However, over the course of presenting, and inevitably debating, FO's work on here, I think the above sentiments have gotten lost along the way. As this is the first, and likely the last, time I'll ever be talking about these issues in post form, allow me a paragraph or 2 to get the debate back on track (drummer, please give me a pass on the whole "narcissistic blogger hijacks lede" idea just this one time).
I'll let Bill's interview answers speak for themselves with respect to the omniscience of FO's stats. Instead, I'll speak to the integrity of my advocacy. When I cite FO's stats in my posts, it's because I think they're the best publicly available measures of football performance. Given my statistical background in measurement methodology, which includes 2 peer-reviewed journal publications testing the trustworthiness of psychological measures in sport and exercise contexts, I'd say I'm adequately qualified to accurately assess DVOA as a measure of football performance; and for you to have confidence in that assessment. That's not to say I'm Alfred Binet or anything; just that I'm also not some wayward fly being drawn to the patio lights of play-by-play analysis. In short, the work I've done (and continue to do) specifically in the area of evaluating measures forms the statistical foundation for what I do here on NN. So I find it pretty ironic that many of the "Danny's a shill for FO" comments suggest some kind of inability and/or unwillingness to evaluate FO's measures. Pretty ironic indeed!
Of course, enough about me. The more important point here is that, when I do have a negative opinion of something FO's done, I focus my critiques on what they did; not who they are. That's the distinction I think we all need to make in our general debates about NFL stat applications, as well as in our specific debates about FO. Going forward, we all just need to accept the facts I've laid out here, and move on. FO's not Criss Angel, and I'm not the guy who believes Criss Angel actually just walked on water. If, rather than on the credibitlity of magic, we focus instead on the merits of FO's work and the validity of my applications, we'll all learn more from the ongoing debate.
The bottom lines is this: If you understand FO's stats, and love them; great. If you understand their stats, and hate them anyway, great. If you don't think stats have a place in football, and have decided to ignore FO (and I) altogether, great. If you don't understand their stats, and would like to learn more, double great. But what is not great is failing to make the effort to understand their stats, and critiquing them anyway. What's also not great is critiquing them in a way that suggests (a) FO's people are self-unaware, pompous, snake oil salesmen; or (b) I'm the equivalent of a guest host for The Chris Farley Show. In other words, given my attempt here to steer things back towards civility and reason, don't be surprised if future devolutions into these not-great things cause my head to explode like a tomato (h/t Andrew Davidson):
So, now that I've finished my (not-so) little diatribe, it's time to move on to the stuff that you actually clicked here to read. Many thanks to Bill for taking the time to respond to the gargantuan number of questions I asked. Given that FO is smack dab in the middle of interview season, and Bill himself is currently in the middle of writing up a series of posts for FO's website, he easily could have answered fewer; an option that I offered him in an advance. And thanks also to everyone who helped me out by responding to my call for questions. Coming up with 15 questions was difficult enough. For the other 5, I probably would have just re-asked Fooch's questions from last year, and let the hilarity ensue. Anyway, enjoy!
THE 49ERS CHAPTER IN FOA 2010
Florida Danny (FD): Given the 49ers' status as a trendy sleeper team, and the real possibility that they might actually be good for the first time in a long time, a lot of the readers on Niners Nation are borderline apoplectic about FO's prediction of 6.1 wins for their beloved team. What I'm more interested in, though, is that FO estimates the 49ers to have just about an equal chance of winning 9 or more games (13%) as they do of winning 0-3 games (12%). Can you briefly explain how FO's projection model could arrive at such a counterintuitive pair of equally likely outcomes?
Bill Barnwell (BB): That has to do with the way that we generate the win projections for each team. We don't actually go through our system and generate a win projection for each team; we actually generate a DVOA projection for each team on offense, defense, and special teams, which produces an overall projection. Then we run 10,000 simulations of the NFL season using these projections while incorporating things like home-field advantage, weather for certain teams in given months, and other (proprietary) variables that allow us to answer the question of how likely a team with, say, a -10% DVOA is to beat one with a 5% DVOA at the beginning of December. I suspect the wide variance in the specific 49ers simulations has to do with the nature of the NFC West, which has no clear-cut favorite.
FD: In FOA 2010, Doug Farrar mentions that one result of the 49ers' recent front-office restructuring was Paraag Marathe becoming one of the significant power-players in the organization via being "in charge of all numbers-crunching..." Given your knowledge of NFL front offices and your contacts throughout the league, where would you say the 49ers currently rank in terms of incorporating statistical analysis into their football operations: among the most stats-reliant, among the least, or somewhere in the middle?
BB: I would say -- from what I know publicly -- that they are in the first third of front offices as far as using statistics in a coherent, effective manner. When I say publicly, I don't mean that I'm holding back information; I just mean that a team that's very quiet about their operations (like New England) could be using statistical analysis up and down their organization and just not be telling anyone. I don't think that the 49ers are attempting to reduce players to numbers by any means, but I think they do a good job of finding intelligent ways to spend their money without making dumb mistakes or overvaluing certain assets.
FD: Regarding the offensive side of the ball, it seems like the model predicts a dip in passing efficiency primarily because the 49ers will likely be using much less shotgun in 2010 than the 50% of the time that they used it with the college-spread-product Alex Smith at QB in 2009. How much, if at all, do you think the addition of Davis and Iupati via the draft, along with the continued development of 2nd-year WR Michael Crabtree, might offset these negative effects of less shotgun?
BB: I wonder whether Davis and Iupati will make a serious impact as rookies. I'm not going to be naive and suggest that the 49ers' offensive line was the stuff of legend last year, but every position has a growth curve, and I'm not sure if Davis is ready to play tackle at the pro level.
On the other hand, there are other issues that should come into play. Smith played the easiest schedule of any quarterback in football last year -- he still gets to play the NFC West, but it won't be as easy as it was a year ago. They were a dominant team in the red zone and pretty mediocre elsewhere, and we've found that teams with those profiles almost always regress in the red zone in the subsequent season. I do think a full season of Crabtree will help, of course.
FD: Doug notes the curious circumstance wherein Frank Gore had one of his best seasons in 2009 despite running behind the worst run-blocking OL in the league. Personally, I think the sunrise-esque predictability of Jimmy Raye's play-calling in the running game (e.g., a logic-defying 71% of all runs were up the middle) explains a lot of this apparent paradox. In your opinion, how much of the OL's poor performance in 2009 was due to having poor run-blockers along the OL; how much was due to having a predictable play-caller?
BB: I don't really think you can lay exact percentages in one area or the other. Certainly, if you look at the 49ers by zone for last year, they were far better running to either left end or right end than they were to anywhere in-between.
FD: On the defensive side of the ball, FO's projection of a DVOA decline seems to be primarily based on regression to the mean in turnover rate. However, the 2 players that Doug cites as being potential sources of this regression to the mean are one 26-year-old (Dashon Goldson) coming off his 1st season as starter at a turnover-creating position, and another 26-year-old (Ahmad Brooks) coming off his 1st season as a primary reserve at 3-4 OLB, a position that seems to suit his skill set more than the 4-3 ILB he played during his rookie season with the Bengals. Again, this seems counterintuitive. Rather than pitting you against Doug here in the specific context of 2010, I'll instead simply ask for your general assessment of the prospects for these 2 players going forward in their careers.
BB: I think there's more to our expected regression than simply turnover rate, but I'll address that. Our suggestion here is that those players aren't likely to produce turnovers at that rate because no player is likely to produce at those rates over several years, not because of anything innately to do with Brooks or Goldson uniquely. How often do reserve outside linebackers get six sacks and force four fumbles? The answer is not very regularly; only seven players in the past 20 years have five or more sacks as a linebacker without starting a single game. Only two players in the past 20 years have forced four fumbles as a linebacker without starting a game. Brooks did both those things last year. Maybe he'll do it again. Maybe his role will increase and he'll continue to rack up the turnovers. I suspect it's not likely.
As for Goldson, he had two sacks, three interceptions, and three forced fumbles; he's one of only 11 instances of a safety doing that in 20 years, and no one has done it in three seasons. If we just focus on the turnovers, Goldson was one of two safeties with three INTs and three forced fumbles last year; Ed Reed was the other. Reed's done that twice during his career. Brian Dawkins has done it twice in his career. Troy Polamalu's never done it. Bob Sanders has never done it. I mean, even if Goldson's a great safety, the odds of him producing at that rate again are remarkably slim.
OTHER FO-RELATED TOPICS
FD: As I mentioned earlier, a lot of NN's readers have taken FO's perennial panning of the team's playoff prospects as a sign of anti-49er bias. What's your general response to people who play the bias card against FO?
BB: It's the reaction we hear from about 28 fanbases every year. The fans of the team we predict to post the league's best record see us as the geniuses who finally confirmed their beliefs, the fans of the teams we predict to take a huge leap forward back us for seeing the young talent percolating in their organization, and the awful team we predict to stay terrible has fans that wallow in self-pity. Pretty much everyone else thinks that we're underestimating them. Every year, a handful of those fans are right. We've said mean things about the 49ers because we've written six books, and they were based on six seasons in which the Niners went a combined 33-63.
FD: The ideal in statistics is 100% objectivity. However, subjectivity inevitably creeps into the process via things like deciding what variables to include and omit from one's models, and what levels of statistical significance are considered meaningful. Roughly, what percentage of the development and continued refinement of DVOA is/was subjective; what percentage is/was subjective?
BB: In part, I can't answer that question, and with regards to the part that I can, I think we're confusing two separate concepts -- DVOA, the play-by-play analysis metric, and our projection system, which is based on DVOA. The goal with each is to improve its predictive ability, and our goal is to find objective variables that do so. There are very few times when we even have the opportunity to introduce subjectivity into the equation, let alone situations where we actually do.
A good example of a subjective variable might be one we have in the projection system that says, essentially, "Peyton Manning-led offenses always exceed our expectations. Bump up this offense if it involves Peyton Manning." Even then, though, the subjective variable objectively improves the equation.
FD: Despite your admissions to the contrary, people seem to think that you guys have the same level of confidence in a team stat like DVOA as you do in an individual stat like DYAR, for example. For the record, in which of your stats do you have the most confidence; in which do you have the least confidence?
BB: Oh, come on! That's like asking us which one of our children is the prettiest and which is the ugliest!
That's sort of a hard question to answer; I don't think I can really say "This is the stat I have the most confidence in". The issue is more about how there's a difference between a statistic being part of the equation and serving as a total value metric of player performance.
Let's take DVOA for running backs, for example. Last year, Pierre Thomas led the league in DVOA for qualifying running backs. Do I think he was the best running back in football? Of course not. No one should. Pierre Thomas plays in a great passing offense, which opens up opportunities for him in the running game. With those opportunities, he was very efficient and didn't turn the ball over, which makes him a valuable back; it doesn't mean that, say, Adrian Peterson wouldn't do better. You have to take the individual statistic and consider the context.
Now, consider receiving DVOA for running backs, which is a stat I don't like to use very frequently. A running back who is thrown the ball on a dumpoff on third-and-20 and gets four yards will end up with an ugly DVOA for that play. (He's only compared to other running backs in that same situation, but it's still not going to produce anything close to a positive number.) That's a product of the player's usage pattern, and it doesn't mean that the player is a bad receiver. It's another situation where you have to consider context.
Of course, compare DVOA to receiving yards. Those four yards he gained on the dumpoff are just as relevant to his yardage total as four receiving yards he gained on third-and-3. Does that seem logical?
FD: Over the lifetime of FO, you guys have received plenty of criticism, some justified and some not. In your opinion, what's the most valid criticism FO has received; the least valid?
BB: Oh wow. I don't think I want to answer either end of that one! I know that we are really sensitive to criticism and do try and listen to critiques that are testable. Ideas like, say, "You underrate the 49ers!" aren't really taken seriously; if someone comes to us and says "You're underestimating teams that play in the 3-4 and produce higher than average turnover rates", well, that's something we can -- and do -- test.
FD: When consumers of your work provide what you perceive to be constructive criticism or good ideas, how often do you incorporate these feedback-driven criticisms/ideas into your work? Can you provide an example of one situation -- without naming names -- in which you did just that?
BB: I think we tend to incorporate them into articles or features, whether they be on the site or in the book. There's a lot of examples of this -- a recent one would be SackSEER, the edge rusher projection system from reader Nathan Forster, which gets a long feature in FOA10.
TO BE CONTINUED TOMORROW...