Bill James calls Sabermetrics "a search for objective knowledge about baseball." The same sort of idea is now being applied to football. A main contributor: Bill Barnwell of footballoutsiders.com. He told New York Newsday: "We use statistical analysis of game film to explore and ask questions about football that aren't asked. For every single play of the NFL season, we have someone watching that game and charting 15 or 20 different things that aren't included in the play-by play."
This week I caught up with Barnwell.
Q: Based on your analysis, rank the starting QBs in the NFC North with an explanation as to why you have them where you do.
A: I have to put Jay Cutler at the top; there's simply not very manyquarterbacks who have produced at his level this early in theircareers. There's no comparable quarterback perhaps in NFL history whowas traded at the same point of their career after achieving the samelevel of success. He turned the ball over a few too many times lastyear, but that's what will happen when you're forced to throw all thetime to catch up. His interception rate was flukily high and willregress back towards league average this year. Behind him would beAaron Rodgers, albeit with the qualifier that he may very well havealready peaked in his first season as a starter.
Third would be Mr. Favre, who didn't put up anyone's idea of agreat season last year in New York while playing against one of theeasiest schedule's any quarterback has gotten to face in the last 15years. (He is joined among those 10 easiest schedules of pass defensesby AFC East compadres Chad Pennington and Trent Edwards, while MattCassel's slate was only slightly harder.) Well behind him would beMatthew Stafford.
Q: Same question with the defenses in the NFC North.
A: Have to put Minnesota #1 by a fair margin -- their -17.2% DVOAwas fourth-best in the league last season. Chicago was second, with a-6.8% DVOA (seventh in the league); Green Bay was third, at 1.4%, whileDetroit had the worst defense in the 15 years we have compiled DVOAfor, at 29.2%.
I'd probably venture to guess that Chicago and Minnesota will beright near each other this year in a 1 and 1A sort of situation;Chicago had a lot of injuries last year, and injuries regress to themean. Green Bay's 3-4 is a mess and will take some time to percolate,while Detroit's still looking to figure out where their building blocksare on defense.
For the uninitiated, DVOA is our core metric; it takes every playof the NFL season and compares a team's performance to how aleague-average team would have done, and then adjusts for the down anddistance, the game situation, and the quality of the opposition. Theresult is a percentage; since defenses want to allow less than theleague average, when we say that the Vikings' defense had a -17.2%DVOA, it means that they were 17.2% better than league average.
Q: Considering the research you've compiled on the Vikings,including extensive film breakdown sessions, give us someout-of-the-box, unconventional takes? In other words, stuff the averageor even die-hard fan would not realize.
A: One thing that I always talk about with the Vikings is the yardsBernard Berrian lost on pass interference penalties. Last year, Berriandrew six pass interference penalties for 127 yards; if you factor inthose yards (and there's no reason not to, since the offense picks themup and the interference, by the letter of the law, prevented Berrianfrom making a catch), Berrian's receiving yards go from 964 to 1091. Itsure makes his 2008 season look a lot better.
Q: According to your numbers, Visanthe Shiancoe was the No. 2 TE infootball last year...explain how you came to that conclusion, and canhe get better?
A: We also use that DVOA statistic for individualplayers when they're thrown passes or carry the ball. We compare theirperformance to the average performance of a player at that positionafter adjusting for down, distance, situation, and the quality of theopponent.
In this case, Shiancoe's 50% DVOA means that he was 50% better thana league-average player on those plays in which he was thrown the ball.Of course, DVOA doesn't know who was in coverage, or whether thesafeties that would normally worry about Shiancoe were bugging outabout AD in the backfield.
That's why I wouldn't say that Shiancoe was the second-best tightend in football a year ago. Shiancoe's second-place rank in DVOA meansthat he was the second best receiving tight end in the league on thepasses he was thrown in the context of the Vikings offense. I stronglydoubt he'll improve; his huge DVOA figure and touchdown total istotally out of line with his previous performance, which almost alwaysmeans that the player will regress some in the subsequent season. Istill think he's a valuable player, but he's not going to put up acrazy DVOA like that again.
Q: A headline from your site: Trending down: QB Brett Favre -- Why?
A: Asmentioned above, Favre had a pretty mundane season despite playing aridiculously easy schedule. He also had a very good offensive line thatmade it to 80 starts around him, so the doubling of his sack rate isconcerning. His injured rotator cuff is also concerning; he's beenincredibly sturdy throughout his career, but he's 40 now. It's a loteasier to get hurt at 40, even if you're a physical ironman.
Q: The Vikings' most overrated player is? And underrated?
A: Geez.That's a tough one. Maybe Bobby Wade for most overrated? I'm not reallysure why he has a job and not 15 or 20 other receivers who can doexactly what he does for the league minimum.
I would have said Antoine Winfield for underrated, but he finally(deservedly) made it to the Pro Bowl last year, so I won't pick him.Maybe Kevin Williams -- I think people realize that he's very good, butmost people don't realize what an absolute force of nature he can be inthe middle, even as a pass rusher. He's still young, obviously, buthe's several years into a Hall of Fame career at this point.
Q: From the naked eye, the Bears defense was terrible last year,but your numbers indicate otherwise...please explain and what do youexpect from maybe the Vikings' No. 1 nemesis in the North?
A: Iwould suggest that the naked eye probably isn't the best tool foranalyzing how 11 players perform over a thousand snaps or so.
I don't even think you could say they were terrible from anyperspective. They gave up 350 points last year, which was 16th in theleague, so their scoring defense was right around league average.
Then,of course, you have to consider the context. Namely, it's the fact thatthey faced 201 drives last year (not include "drives" that werekneeldowns at the end of a half); only one other team (Philadelphia)faced as many as 195. The league average was 176, so Chicago faced 25more drives than the average team.
Divide the points they gave up by the number of drives they faced,and Chicago only gave up 1.68 points per drive; that was ninth in theleague, which is far better than 16th and closer to our seventh-overallranking. They didn't get much help from their offense or special teamscoverage, as the average drive Chicago faced started from the 31-yardline, which was eighth-worst in the league. (Minnesota was actuallyseventh-worst, and you can probably figure out why.)
Their 2009 performance depends on health and the pass rush. Theycouldn't rush the passer with their front four last year, so LovieSmith had to blitz more frequently and mix up his looks defensively;the result was a decline in their performance, and when the secondarystarted going down with injuries, they were basically hanging on fordear life. If the secondary stays healthy and they can actuallygenerate a pass rush up front, the defense could be very, very good --right up there with Pittsburgh and Baltimore.
Q: Green Bay is switching to a 3-4 defense, although word is, itwon't be 100 percent 3-4 dominated, but will its defense improveenough?...And do you buy on the idea that QB Aaron Rodgers is ready toexplode for a super year after his phenomenal preseason?
A: Easy question first: Throw out preseason stats for players. Almostentirely useless, and absolutely useless when you're consideringveteran players.
The short answer about Green Bay's 3-4 is thatwe don't know yet. It depends on how well the players they're movingaround adapt to their new roles. Past scheme switches in the leaguehave shown that teams switching to a 3-4 from a 4-3 tend to strugglewith the switch over the first four games before settling in andimproving on their previous level of performance, but the variance withan individual team as opposed to the "average" switch can be vicious.Again, it's a health thing; that secondary's going to need to stayhealthy for the Packers to be able to go through growing pains upfront. If Al Harris and Charles Woodson age quickly … look out.
Q: Give us a team or two that we expect very little from (like lastyear with Atlanta and to a degree Baltimore) that has a very goodchance to be in the mix for a playoff spot?
A: Our pick for thisseason was St. Louis, although that might be up in the air now thatAdam Carriker is out for the year. Our projection system compares themoves that they made to how other teams have built their teams in thepast, and compares the splits from their 2008 season to previous teams'campaigns, and sees them improving by a fair amount in a weak division.We're also projecting Jacksonville to return to prominence, thanks to amuch-improved (and much-healthier) offensive line.
Q: What do you say to the individuals who disagree that you canreally assign win probabilities in a game like football where so muchis due to chemistry, injuries, and luck?
A: I don't think it's abinary decision; it's not "Statistics can't be applied to football" or"Statistics can be applied perfectly to football". What we do is focuson the process, understanding that the nature of a 16-game schedule orthe bounces of a ball on a single play can have a huge impact. Our goalis to build models that work over 15 seasons of team performance andserve as better predictors of team performance. DVOA is a betterpredictor of a team's win-loss record in the subsequent season than anyother statistic, including the team's record in the previous year.
Our statistics will never be perfect; no statistical model ever is.There's plenty of research we do into the game and why teams win,though, that's fascinating and counterintuitive while remainingdemonstrably true. The work we do is certainly good enough to deservethe attention it's received, even though there's still much more to bedone.