Running In Style: Vikings Game-Tying Drive

In a passing league, consistently running the ball is a mistake. Consistently running the ball well is both difficult and rare.  Down 21-14 in the 3rd quarter, the Vikings used a tandem backfield as a crutch, supporting and propelling them over Carolina.  The Vikings, who lead the league in Run EPA, relied heavily on both Toby Gerhart and Adrian Peterson.  The two RBs were featured on 9 of the 13 plays, 6 of which were successful (2 registered a neutral EPA of -0.06).  Here is a look at the evolution of the game-tying drive - which increased the Vikings' chance of winning by 27% - using our Markov model.

"Clutch" Defenses in 2011

Tonight the two most 'clutch' defenses of 2011 face off. One way to measure which defenses have been most 'clutch' is to take the difference between their actual Win Probability Added (WPA) and their expected WPA given their Expected Points Added (EPA). WPA accounts for the leverage of score and time while EPA ignores that leverage. A plot of each defense's Win Probability Added (WPA) against his Expected Points Added (EPA) creates a baseline expected WPA for each defense.

Next, we can measure the difference between each defense's expected WPA and his actual WPA. This difference could be considered clutchitude. The graph below illustrates this concept. The vertical red line is the difference between the WPA and the expected WPA of a very un-clutch defense this season--the Carolina Panthers.

Note that this is not a claim that 'clutch' is a permanent or reliable quality of any team or player. It is only a recognition that some squads have performed better or worse in high leverage situations than they have in other situations so far this season. We might expect the squads at the top of the list to regress toward the mean over the remainder of the season. In other words, they might be considered as 'over-performing.' Likewise, the squads at the bottom of the list might be expected to regress upward.  For example, the Jets' #27 ranking does not imply they do not play well under pressure. It only means that their best play has tended to be in low-leverage situations to date.

Sunday's Numbers Have Been Crunched

Sunday's numbers are now available, including advanced stat box scores, top players of the week, team stats, and season leader boards.

Week Eight Game Preview: Detroit at Tim Tebow

The Weekly League diverts from its normal format to look at the Detroit Lions-Denver Broncos contest in depth.

Detroit at Denver | Sunday, October 30 | 4:05pm ET
Four Factors

The Obviously Interesting Thing About This Game
The obviously interesting thing about this game is how Tim Tebow will be making his second start for the Broncos in as many weeks -- after playing a large part in Denver's come-from-behind, overtime victory last week against Miami.

Roundup 10/28/11

Is the modern spread option really just the single wing?

Nine obscure rules in sports.

A review of the Wolfe ratings, one of the inputs to the BCS rankings.

" that's a thing anymore."

Being proven wrong is good.

I generally like the Freakonomics guys, but there are so many things wrong with this little video essay, I don't even know where to begin. I was thinking of having a contest to see who could list the most false assumptions, fallacies, and non-sequiturs. That said, I agree with the conclusion.
Here's another Freakonomics video on icing the kicker.

Playoff Probabilities Week 8

Courtesy of Chris Cox at, the latest playoff probabilities for each team.

These are generated using the NFL-Forecast software app, which uses the win probabilities generated by the team efficiency model to simulate the NFL season 5,000 times. And if you don't buy the game probabilities from Advanced NFL Stats, you can tweak them as much as you like to generate your own playoff projection. I encourage everyone to download the app and test out your own scenarios.

Well, the simulation machine said it was the High Leverage game last week, and it sure wasn't kidding. After Houston's pasting of Tennessee in Nashville, the Texans increased their overall playoff probability by 29 points, up to 91%, while the Titans plummeted in the tables, dropping 34 points to 24%.

Much of Houston's rise (and Tennessee's fall) is attributable to the way in which the Texans won—they didn't just beat the Titans, they did so with such brutal efficiency as to rise to the top of the team rankings and send Tennessee tumbling down to #18. Overall, the model now gives the Texans a 78% chance to finish with ten or more wins, an outcome almost certain to result in a playoff berth.

WP: The Record Keeps Skipping

This week's post at the Washington Post's Insider looks how badly the Redskins tend to fade as the season progresses and injuries accumulate. Plus:

  • How much WP did kicking a FG on 4th and 2 from the Panther's 13 cost?
  • Can Torain and Helu make up for the loss of Hightower?
  • How does Fred Davis compare to Chris Cooley?
  • Can Santana Moss be replaced in the lineup?

Weekly Game Probabilities

Weekly game probabilities are available now at the Fifth Down. The lead-in is a re-purposed version of my discussion of intangibles. The only upset the model likes this week is PIT over NE.

Stat Pages Now Feature Ranks

The player stat pages for both offensive skill players and defenders now have a Rank column. When you sort the table by the various stats, the ranks remain in place so you can see where each player stacks up with the rest of the league in terms of EPA, SR, total passing yards, deep percentage, etc., without having to count the rows.

The team page, offensive line page, and the adjusted SR page also now have the feature.
I had meant to add this for a while, but the script that sorts the tables wouldn't play along. I finally got around to doing the homework needed to make it happen. Counting rows for ranks was something that always bothered me, and I assume bothered you guys too. For now there are still a couple flaws, such as when a stat is good when it is low (defensive WPA for example), teams are ranked upside down. Like everything here, it's a work in progress.

Look Out Felix, Here Comes DeMarco

Selected 71st overall by the Dallas Cowboys during the 2011 National League Football draft, DeMarco Murray entered his first pro season with the tall task of distinguishing himself from the solid offensive core already in place. He definitely made his presence known last weekend in the Dallas Cowboys' 34-7 win over the St. Louis Rams.

Known for his speed and quickness, Murray set a franchise record by rushing for 253 yards, which broke Emmitt Smith's 1993 record of 237 and was just two shy of fellow RB Felix Jones' total for the entire '11 season. It's the most rushing yards in the NFL since Kansas City Chiefs' Jamaal Charles ran for 259 against the Denver Broncos on Jan. 3, 2010. Minnesota Vikings' Adrian Peterson holds the rookie rushing record of 296 yards, which was set in 2007.

New Orleans Succeeds On Offense

Mark Ingram's reaction to Colts' defenders Sunday night.
If anybody rues the loss of Peyton Manning to a brutal neck injury then the Indianapolis Colts, it's NBC. Week Eight marked the second appearance of the Colts on Sunday Night Football, with the NFL unable to flex out of poor matchups until Week 11. The first Colts appearance saw Indianapolis take on the Steelers at Lucas Oil Stadium, a surprising matchup and Curtis Painter's coming out party as "surprisingly decent quarterback." The second time around was not so smooth for Painter or anybody else unfortunate enough to occupy the Colts roster Sunday night, as Indianapolis fell hard, dropping to 0-7 via a 62-7 thumping from the Saints at the Superdome.

The Saints took a 21-0 lead through the first quarter and a 34-7 thumping into the halftime break, with NBC losing many of those precious viewers in can typically count on with primetime NFL action. And it became increasingly difficult to watch for those who enjoy competitive football -- the Colts couldn't make a stop to save their lives, allowing scores to the Saints on nine of ten drives and managing only 11 first downs the entire game. But if you enjoy watching an offense utterly steamroll the opposing defense, the output from Drew Brees, Mark Ingram, Marques Colston and company was about as dominant as we can see, and it goes well beyond the 62 points they hung on Indianapolis when all was said and done.

Team Rankings: Week Eight

I know you must be curious about who gets to be the top dog this week, so here are your week eight rankings.

The New Number One
After demolishing a team that was ranked inside the top-ten last week, the Houston Texans leapfrogged into the top spot. Schaub was able to pick apart the Titans' defense despite not having Andre Johnson to throw to, and Arian Foster just went plain nuts on Tennessee's butts. After the loss, the Titans fell all the way from 8th to 18th. The Texans get the Jaguars next week while the Titans get a cushy bounce-back game against the Colts.

Another Argument for Tebow

Everybody loves Tim Tebow.  Everybody loves to talk about Tim Tebow.  He is one of the few polarizing figures in the NFL; everyone has their own opinion of him - one that does not often change.  While I was not a huge fan of him in college, I find myself loving him now that the world wants him to fail especially hard.  We've heard one statement over and over since Tebow joined the Broncos last season: Kyle Orton gives the Broncos the best chance to win.  I disagree.  While Orton is the better and more consistent quarterback, Tebow gives the Broncos the best chance to win.

There has been a great deal of research on the effects of variance in determining game strategies.  Brian wrote a post a few years back about how underdogs need to use high variance strategies and Dean Oliver discussed the underlying statistical concepts in basketball terms.  The main point is this: Below average teams must use riskier strategies to win.  If a team's performance is below league average, then consistency is a flaw.

That brings us back to Orton v. Tebow.  If the Broncos were a league average team or better, Orton would indeed give them a better chance to win.  Yet, the fact that the Broncos are dismal should be more of a reason for Tebow to play.  Picture Orton and Tebow as follows:

Stats Measure the Intangibles

It’s a common criticism of statistics: Stats don’t measure the intangibles. You know—the drive, character, discipline, leadership, teamwork, and all the other desirable qualities in athletes. Except that they do. Stats measure the effects of what we call intangibles, and always have.

Although there is no event at the Combine that can directly gauge a player's will to win, it still manifests in his performance on the field. And as long as the stats reflect what actually transpires between the sidelines, they capture the effects of the competitors’ intangible qualities. And if the intangibles don’t show up in the numbers, then they didn’t matter in the first place.

What statistics can’t do is separate the physical-material tangible qualities, such as speed, strength and skill, from the intangible ones. They are captured together. But that's no different than trying to separate the effect of speed from the effect of strength. All qualities, whether tangible or intangible, are inseparable using game statistics alone.

If a great player is also a great leader who inspires his teammates to play better, then that will be reflected in his teams’ statistics. It may not be captured in his personal stats, but that has always been true of the effects of teammates on each other’s performance, whether tangible or intangible.

Sunday's Numbers Have Been Crunched

Sunday's numbers are now available, including advanced stat box scores, top players of the week, team stats, and season leader boards.

Quick Reactions to Sunday's Games

WAS 20, CAR 33

Beck did as well as you could expect (-0.15 WPA, --5.9 EPA, 40% SR, 5.4 AYPA) with WAS's OL injuries. The OL was the cause of a good chunk of WAS's offensive problems (-0.12 WPA, -4.6 EPA), allowing CAR's DL to have their way with Beck, and he was unable to take advantage of the worst pass defense in the league. Hightower (0.08 WPA, 2.9 EPA, 55% SR, 5.2 YPC) was effective on the ground, but left the game with a knee injury. Newton ruled the day (0.29 WPA, 12.7 EPA, 51% SR, 8.6 AYPA, 0 Ints) largely thanks to Steve Smith (0.27 WPA, 11.0 EPA, 16 YPT on 9 Tgts). This game was the Cam Newton/Steve Smith Show. WAS blundered by kicking a FG on 4th and 2 from the CAR 13 down by 3 in the 2nd quarter.

SEA 3, CLE 6

Ugg. Two FGs vs one. Whitehurst was a disaster (-0.59 WPA, -17.3 EPA, 24% SR, 0.8 AYPA). CLE's OL has had better games (-0.28 WPA, -3.3 EPA, 3 sacks, 5 QB Hits, 3 tackles for losses, 33% run SR). I thought they were supposed to be good? Technically, the big 4th down blunder of the week was CLE's FG attempt with 3:08 to play. It was 4th and 3 from the SEA 6. CLE was only up 3. In any other game this would have been a huge mistake, but with the way SEA was moving the ball, maybe it made sense. But I even doubt that. If you can help it, never put yourself up 4 through 6 points with just enough time for a single drive. Your opponent now has all four downs at his disposal all the way into your endzone. In a typical game, this was a -0.12 WPA mistake. They don't get much bigger. As Gregg Easterbrook will (correctly) write on Tuesday, 'the football gods spoke, and the kick was blocked.'

Offensive Line Stats Now in Advanced Boxscores

You might have noticed a new addition to the advanced boxscores the last couple weeks. In addition to offensive skill player and defender stats, there are now stats for team offensive lines.

There's conventional stuff like sacks allowed, sack yards, QB hits, and tackles for losses. But there's also the advanced stuff like run SR, -WPA and -EPA.

I've included run SR because offensive lines are so critical to the first few yards of a run. Once a RB is sprung into the deeper levels of the defense, longer gains are on his shoulders, but until then, it's primarily the line that makes the difference between a 4 yard gain and a 2 yard gain.

-WPA and -EPA are more abstract. The full explanation is here, but this is a brief explanation of how these stats work. It starts with the idea that it's offensive line's job to stop bad things from happening, like sacks, stuffs, short gains, and QB hits. So when bad things like that are done by the opposing defense's front seven, it's usually due to some failure by the offensive line. When negative EPA or negative WPA plays occur at the hands of front seven defenders, the loss of WPA and EPA are attributed to the line. All that WPA and EPA accumulate, and can only be negative, which is why I call the stats -WPA and -EPA.

Should the Vikings Have Punted?

The Vikings were one touchdown away from what would have been a stunning upset of the undefeated Packers. MIN began the drive at their own 2 with 5:26 to play. They managed to make it out the their own 36, but with 2:37 to play they faced a 4th and 10. With all three time outs remaining, Leslie Frasier elected to punt. Was this the best decision?

It's a very close call.

Converting a 4th and 10 is a 35% proposition. And had MIN converted, they would have had a 1st and 10 at their own 46 (at least) with about 2:30 to play. That equates to a 35% WP. On the go-for-it side of the ledger, having all 3 timeouts isn't beneficial because they had plenty of time to score, even without them. Time was not the factor.

A failed conversion attempt gives them an almost zero chance of winning. So on net, the go-for-it option is just:

Worst Single-Game WPA

Last week a coworker of mine (and long suffering Redskins fan) was curious who had the worst single-game WPA of all time. (I think he had a hunch it belongs to Rex Grossman.) I can't go back to "all time", but I can go back through the 2000 season. The worst (undisputed) game ever belongs to A.J. Feeley, who posted a -1.16 WPA as the Eagles' QB against the Seahawks in 2007. There may be a couple worse games, but both have some gray area which I'll explain.

Sure enough, with Sam Bradford out today, none other than A.J. Feeley himself takes over as the Rams' QB against the Cowboys.

I sometimes get asked whether it's possible for a player to have greater than a 1.00 WPA or less than -1.00 WPA. Sure. It's rare, but possible. When a player is having a very bad day, but the team around him is firing on all cylinders keeping the game tight, players can often fall below -0.50 WPA. And if they top off their bad day with a calamitous play with the game on the line in the 4th quarter, they can break the -1.00 barrier. Similarly, if the game is a shootout, and a player makes a clutch play to steal the win, he can break the +1.00 barrier.

FanGraphs Is Now Even Awesomer

The best baseball site on the Web has just become even better. FanGraphs has long featured game win probability graphs, but now they are interactive like the ones here. You can roll your cursor over each part of the graph to see the play-by-play.

Even better, FanGraphs has created a widget so that other sites can incorporate their graphs. You can click on the small down arrow in the top right corner to get the widget code or just a snapshot if desired. I've found this is a great way to spread the word. Don't forget about our own NFL WP widget.

Check it out the FG widget after the jump.

Roundup 10/22/11

Garrett's end-game risk aversion was costly. More on the Cowboys over-conservatism using WP. The bottom line here is that the WP for a team needing a TD with 5 min to go isn't much different than for one with 4 min to go or 3 min to go. Running 3 times into a brick wall to burn clock only makes sense if you believe it's the best way to convert first downs and leave the opponent with nearly no time to score.

Percentage of total offensive yards, an interesting way to look at RBs.

No wonder 90% of fans have no idea what I'm trying to say.

Buddy Ryan is not fond of QBs.

Outside the Hashes reviews the computer side of the BCS rankings. I'm all for objective quantitative rankings, but some of the component systems (I think there are 6) may not be the most sound.

Carson Palmer on the Back of an Envelope

Palmer is good QB, but he was what people call ‘elite’ for only one season, his second year which ended with that unfortunate knee injury in the playoffs. A lot of people say that since the injury, “he hasn’t been the same.” Perhaps the injury is the cause, but maybe he simply caught lightning in a bottle in 2005. Either way, how he performs as Jason Campbell’s replacement in Oakland is going to be fairly unpredictable. There are lots of considerations on both sides of the ledger—new team, out of practice, same system, familiar coach, etc. One thing we can be sure of, however, is that he’s going to be better than Kyle Boller.

Let’s take a quick look at Palmer’s career numbers to see where his typical level of performance slots among QBs in 2011.

The Weekly League: Notes and Ideas for Week Seven

It's a metaphor.

This week's edition of The Weekly League features

1. Largely superficial previews of the Chicago-Tampa Bay, Green Bay-Minnesota, and Indianapolis-New Orleans pro football games.

2. A table of pythagorean records using Expected Points Added, by which measure the Tampa Bay Buccaneers are exposed!!!1!


3. Armchair Psychology, Armchair Statistics, and Armchair Other Sciences.

Chicago at Tampa Bay | Sunday, October 23 | 1:00pm ET
Four Factors

• This seems like an appropriate time to do a Jay Cutler Sack Attack update.
• Doing so, we find that Cutler has been sacked 19 times in six games this season -- but that only five of those sacks have come over the last three games.
• Question: Is that an (a) trend or (b) product of arbitrary endpoints?
• Answer: We don't know.
• Or, rather: I, personally, don't know. There's maybe someone who does.

Playoff Probabilities Week 7

Courtesy of Chris Cox at, the latest playoff probabilities for each team.

These are generated using the NFL-Forecast software app, which uses the win probabilities generated by the team efficiency model to simulate the NFL season 5,000 times. And if you don't buy the game probabilities from Advanced NFL Stats, you can tweak them as much as you like to generate your own playoff projection. I encourage everyone to download the app and test out your own scenarios.

Moving Up
Their win over Houston jolted Baltimore into first place in the team rankings and gave a 21% boost to the Ravens' overall playoff probability, putting them at the top of the tables in the AFC. In addition, the Ravens are now forecast to get a first-round bye in over half of the simulations.

Moving Down
Things tightened in the NFC East last week, as the Redskins lost a division matchup and the Cowboys' failure to capitalize on Patriot turnovers dropped Dallas to 2-3. As a result, both teams saw their playoff probabilities decline 19%. However, the Cowboys' strong position in the team rankings (#2 overall) continues to propel them to a playoff berth in 64% of simulations.

Weekly Game Probabilities

Weekly game probabilities are available now at the Fifth Down. The lead-in is a re-purposed version of my critique of power rankings.

The Cowboys' Squandered Opportunities

Finally, Dallas's reign of terror atop the Advanced NFL Stats team rankings is over.

The 5-1 Ravens moved into first place this week with a 29-14 decimation of the Houston Texans. The Cowboys now sit at second place, partially thanks to a third loss on the season. Like the previous two losses, Dallas played well enough to win.

Unlike the Cowboys' other two losses, however, it wasn't about their own turnovers, but what they did with the Patriots' turnovers that resulted in their demise. The Patriots have been uncharacteristically sloppy in terms of ball control this season. Their 10 turnovers through their first six games is four more than the 2010 season, and Sunday marked New England's second four-turnover game of the season. In as tightly contested a game as last Sunday's -- one which finished 20-16 in the Patriots' favor -- it isn't difficult to imagine multiple scenarios in which those four turnovers turned the tide in favor of the Cowboys.

Prediction Accuracy

Doing this website for the past few years has turned me into a student of human nature. In addition to the irrational biases coaches and other decision-makers in football seem to exhibit, I’ve learned a lot about football fans in general and about my own readers in particular. Yes, you. I’ve learned what gets clicks, what gets attention, and how to get a reaction. I resist such tactics to the best of my ability, but sometimes it’s impossible to avoid.

If a human brain does anything, it’s trying to make sense of the world around it. It sees the shadows on the cave wall, and it makes a guess about what’s going on outside the cave. The human brain yearns for coherent order in the world and certainty about the future. One way we create order in the world is to literally put things in order. We make lists: to-do lists, Christmas wish lists, F/A-18 landing checklists, etc. And to gain certainty about the future, we make predictions: weather, stocks, sports, etc.

I once wrote to our own Carson Cistulli in an email that “People love lists and they love predictions. What I give them is lists of predictions.” Frankly, it wouldn’t matter how accurate the predictions are. People just crave them. The future creates anxiety, and whether accurate or not, predictions, help relieve it. At least, that’s what my astrologer tells me.

Anyway, over the past couple years there have been a growing number of misleading comments about the accuracy of the game probability model. I’ll rectify this at the end of this post, but bear with me as there are a few thoughts I’d like to share along the way.

WP: Injuries to the O-Line

Today's post at the Washington Post's Insider looks how badly the injuries to the Redksins' offensive line will hurt. Plus:

  • As poorly as Grossman has played this year, it's surprising where this season stacks up against his previous seasons.
  • What do John Beck's previous stats look like?
  • Attempting a FG down 17 in the second half on 4th and 3 on the opponent's 8 is not how to win games.
  • The defense is the bright spot, but where does it rank against other league defenses?

Team Rankings: Week 7

If you're wondering if your favorite team the worst in the league, it probably is. But you should read on, anyway!

The New Top Dawg
The Ravens beat the Texans this past weekend and were able to jump up from outside the top-5 onto the top of the pile. While the numbers gave Houston a ton of respect, the Texans were without Mario Williams and Andre Johnson, so the Ravens' win probably shouldn't have allowed them to leap so far in the rankings. Even though they probably don't deserve the top spot, Baltimore is still a very good team on both sides of the ball. With Lee Evans and Torrey Smith keeping safeties honest, Anquan Boldin and the Ravens' tight end duo should be able to take advantage on intermediate routes. Oh, and don't forget about Ray Rice, the little engine that flew up the hill so fast it made your head spin.

Games I Watched (A Perhaps Never Recurring Column)

Jim Harbaugh bested Jim Schwartz. Not with a handshake but trap blocks that baited Detroit’s elite defensive line and exposed its not so elite linebacker corps. Not with an emasculating slap on the back but a coverage scheme that cooled Calvin Johnson and forced Matthew Stafford into targeting his tight end, Brandon Pettigrew, to little avail. There is something undeniably enticing about two typically inscrutable head coaches almost coming to blows on the gridiron. A tiny aspect of the vast and essential but largely hidden world of head coaching in the NFL revealed: the need to be the alpha among some of the fiercest, most powerful men on Earth. Schwarz—a letterman linebacker at Georgetown with no professional playing experience—slapped aside, excused, big-brother like, by Harbaugh—a former first-round pick by the Chicago Bears— on national television. But it was the game, the significance of a 4-1 49ers team traveling to Detroit to face the 5-0 Lions, mere days after the Tigers had been eliminated from the postseason, winning, slapping a little mortal fear into the team and its fanbase, that was relevant. Not the scandal that followed.

The Problem with Power Rankings

Power rankings, everybody's got 'em. ESPN, Sports Illustrated, and every sports page of every major paper in the country. There's not much football going on during the week, so fans need something to chew on.

There's one big difference between other rankings and the kind you'll find here. The rankings here are tuned to be predictive rather than explanatory. There's a lot of randomness in football game outcomes, and the model here is designed to ignore the noise and focus on the signal. This approach can sometimes produce curious results, as teams with relatively poor records are often ranked well ahead of teams with better records.

At the time this was written, a 2-2 team was ranked #1, while the two 5-0 teams were ranked #5 and #8. There was even a 4-1 team ranked #15, behind a 2-3 team ranked #13. Certainly, there must be something wrong here, at least if we're to believe the dozens of comments in the weekly rankings posts.

Maybe not. Consider a world in which all NFL teams were perfectly evenly matched. Every game would be little different than the flip of a coin. In this kind of world, there would likely still be at least one 5-0 team and one 0-5 team, despite being completely equal in team strength. In this hypothetical world, a power ranking based on team record would be an exercise in self-delusion.

The real NFL isn't too far off from the National Coin Flip League. Certainly there are better and worse teams, but there is also randomness. So although team records are not complete illusions, they are partially random and often misleading, especially early in the season.

Oakland's Opening Onslaught

Who needs Jason Campbell when you have Shane Lechler?  The Raiders all but won their contest against Cleveland on their opening drive.  After their first offensive snap, the Raiders were favored at every point the rest of the way.  In fact, the lowest win probability for Oakland following their first drive was 53% which quickly rocketed to 71% after a Jacoby Ford 101-yard kickoff return.  Jason Campbell led a 15-play, 7:48 minute drive, capped off by a 4-yard TD run by Darren McFadden.  Campbell and crew had 11 successful plays out of the 15, 8 of which were runs (a run Success Rate of 80% on the drive).  The drive, which resulted in a Win Probability Added of +0.22, can be seen modeled here:

Quick Reactions to Sunday's Games

CAR 17, ATL 31

I thought CAR was going to win this one, and they came pretty close despite Newton’s 3 interceptions. That’s what did in CAR. On net, Newton had both positive WPA and EPA. It’s just that Ryan had a great game with +0.55 WPA and +14.4 EPA. Turner had a good day too, with +0.12 WPA and +3.2 EPA on 139 rushing yards at a 5.1 YPC clip. CAR’s defense is atrocious—last in EPA, WPA, and bottom 10 in SR. I’m still not very high on Atlanta. They got a bonus TD at the end of the game, but it was really a 7-point win over a weak team at home.

IND 17, CIN 27

Dalton looks to be on the right track--+0.41 WPA, +14.5 EPA, 78% comp pct, an 8.3 AYPA, no sacks and no ints. A lot of the credit goes to the Bengal’s o-line: 1 QB hit, no sacks, no tackles for losses for +0.13 WPA and +5.0 EPA. So far this season, he’s helped CIN get to a 4-2 record with +5.8 EPA and a 62.4% comp pct. IND was actually in the game until Garcon’s fumble. Carlos Dunlap had the play of the game returning the fumble for a TD. He totaled +0.29 WPA. The CIN defense looks more than credible, totaling -5.2 EPA, -0.88 WPA, and 57.4% SR--ranking 8th, 1st, and 4th respectively. Dalton will probably improve as the season goes on. Keep your eye on the Bengals.

SF 25, DET 19

All the attention is on the post-game handshake, but it should be on the two defenses. Both Smith and Stafford had poor games. Smith was -0.26 WPA, -14.4 EPA with only a 27.8% SR! Stafford’s day was only slightly better, with a -0.20 WPA, -5.1 EPA, and 41.9%. Frank Gore had a big day with 141 rushing yds for 9.4 YPC. Still, his SR was poor—33%. The reason SF pulled off the upset was Ginn’s punt return for 40 yds and Smith’s go-ahead 4th down TD pass to Walker for the win, which was the biggest play of the day. DET had a 19% chance of winning with almost 2 minutes to go after that play, but couldn’t put together a drive. SF should have gone for it on 4th and 4 from the 19 in the 4th qtr. Instead they attempted a FG, which turned out be good, making it a 6-point game. A conversion ends the game then and there. Even a successful FG still gave DET a chance at a game-winning TD drive. SF also made a mistake by punting on 4th and 1 on their own 49. They were down by 4 with 11 min to play.

GB 24 STL 3

I’m surprised GB didn’t win by more.

Sunday's Numbers Have Been Crunched

Sunday's numbers are now available, including advanced stat box scores, top players of the week, team stats, and season leader boards.

The Weekly League: Notes and Ideas for Week Six

Indy's Curtis Painter: probably a better quarterback than babysitter.

This week's edition of The Weekly League features

1. Handcrafted previews of the Indianapolis-Cincinnati, Carolina-Atlanta, and Dallas-New England football games.

2. A table of pythagorean records using Expected Points Added, by which measure Minnesota and Carolina show some promise.


3. Lies and other lies.

Indianapolis at Cincinnati | Sunday, October 16 | 1:00pm ET
Four Factors

• If you're curious about which quarterback recorded the highest Adjusted Yards per Attempt (AYPA) last week -- or even if you're really not curious, turns out -- either way, the answer is "Curtis Painter," at 10.3.
• "Curtis Painter" is also the answer to the questions "Which quarterback is third overall (in a limited sample) in terms of Net Yards per Attempt (NY/A, which includes sack yardage but not interceptions)?" and "Which quarterback has been considerably more efficient than Kerry Collins this season, despite playing in the same offense?"
• In terms of questions to which "Curtis Painter" is not the answer, one of them is "Who's the author of the book Cop Kisser?"
• Because the author of that book is Steve Zultanksi, is why.
• Steve Zultanksi, who writes a poem (in that book, Cop Kisser) called "A Poem for Dick Butkus," a stanza from which is "I'm Dick Butkus. / I'm a little / butt."

2011 Play-By-Play Data Now Available

I remember working so hard to get access to NFL play-by-play data, at least in a format I could use for analysis. I always knew if I could only get the data, I could do some cool stuff with it. 

PBP data through week 4 is now available here. As in previous years, I plan on updating it throughout the season, around each quarter-mark. Previous seasons can be found here. The first download of the year is always the most troublesome, so future updates should be quicker.

Roundup 10/15/11

Football as a state machine.

ESPN's Total QBR ranked Tebow ahead of Rodgers for their week 5 games. Some people thought that was ridiculous. Here is ESPN's response. ESPN is correct. There are four important considerations. First, TQBR is a rate stat, not a counting (cumulative total) stat. Second, TQBR excludes YAC, something the GB receivers generate by the bushel. Additionally, TQBR over-weights clutch play. And lastly, TQBR includes QB rushing, something many people overlook because conventional stats rarely report it. Here's one way to look at it: When Rodgers entered the game, the score was tied and he helped create a +11 net point advantage over two halves. When Tebow entered the game, Denver trailed 23-10 at halftime, and in one half of play he helped generate +8 net points.

With teams coming off of byes for the first time this season, it's a good idea to review this interesting post by Denis O'Regan at the Community site. It appears that there is a considerable advantage for road teams coming off byes, particularly those who are already favored to win.

Modeling turnovers in football and basketball.

If there is anyone's thinking that has provided the theoretical bedrock to understand the curiously poor decision making in football, it's this guy.

I've been harping on this for years. In baseball, the league scorer can decide if a play is a hit or an error. NFL scorers should be able to do the same thing with interceptions, or at least classify them as clean, tipped at line, tipped by receiver, etc.

Playoff Probabilities Week 6

Greetings, good people of Advanced NFL Stats. This week we'll take our first look at the state of the NFL playoff races, in a post complete with numbered lists, bullet points, and made-up numbers. Note: The data below was all calculated using the NFL Forecast app developed by Chris Cox over at, which uses the win probabilities generated by the team efficiency model to simulate the NFL season 5,000 times. And if you don't buy the game probabilities from Advanced NFL Stats, you can tweak them as much as you like to generate your own playoff projection. I encourage everyone to download the app and test out your own scenarios. Since this is the first week, it'd probably be good to review some things to keep in mind when interpreting the numbers:

Weekly Game Probabilities

Weekly game probabilities are available now at the Fifth Down. This week's (triple guarantee lockdown put-it-in-the-bank) upset special is the Panthers-Falcons matchup.

The Battle of the Rookie QBs

In a fun matchup last Sunday, two rookie quarterbacks faced off during the Cincinnati Bengals-Jacksonville Jaguars contest. Thanks to a strong fourth quarter in which Cincinnati rallied for 17 points, the club won the game 30-20.

Freshman quarterback Andy Dalton led his club over fellow first-year chucker Blaine Gabbert. To be honest, neither QB was overly sharp, although Dalton clearly had the edge. He posted a 3.3 EPA (0.09 EPA/P), compared to Gabbert's ugly -11.4 (-0.33). The Jaguar QB also fumbled twice, resulting in 26 lost yards. Dalton's completion rate was also a full 10% better at 63.6%. The WPAs for the game were 0.02 to -0.12 in favor of the Bengal, showing that neither player had a truly strong game.

It wasn't all bad for Gabbert, a graduate of the University of Missouri. He bested his opponent - and Texas Christian U alum - with 221 yards (vs 179) and avoided the interception. Gabbert also tried to make up for his two fumbles with five carries of his own, which resulted in 11 yards gained.

Deadspin/Slate Roundtable: Passing in 2011

Here's a second post in the roundtable series at Slate and Deadspin. This one encapsulates and updates my recent look at the causes of the passing explosion in 2011. It's a little more readable and entertaining than the original graph-filled post.

Packers 25, Falcons 14: A Quarterback's League, Illustrated

In Sunday night's tilt between the Atlanta Falcons and the Green Bay Packers, one thing was clear from the get go: the Packers were losing the ground game. First they watched the Falcons use rushes of 15, 11, and 17 yards to advance into Packers' territory en route to an opening touchdown. Then, Michael Turner punched in the Falcons' second touchdown to give them a 14-0 lead early in the second quarter. Meanwhile, the Packers struggled to get the running game going, picking up only four successful rushes on 10 attempts in the first half. The Packers would end up with only 58 rushing yards, succeeding on only 37% of rushes (7-for-19). The Falcons, somewhat inexplicably, only rushed 18 times, but they succeeded on a 63% of them (12/19) in picking up 86 yards.

Just Win, Maybe!

I just did a post as part of a Slate-Deadspin roundtable on the past weekend in the NFL. In it, I assess the state of strategy in the league by looking at Sunday's fake punts, surprise onside kicks, and 4th down decisions. The NFL isn't really a Just win, baby! kind of league. It's more like a Just try not to lose too fast, baby! kind of league.

You can read the post at either site: Slate or Deadspin. And don't miss the other entries from the roundtable project.

Team Rankings: Week 6

Wonder who the top teams in football really are? Well wonder no more, because the week six rankings are here to set your mind at ease.

The Top Trio
The Cowboys were off this week and they managed to hold on to the top spot. The Saints barely edged out the Panthers, but they still maintained the number two ranking, likely thanks to their week three win against the Texans, who stayed the course despite a loss to come in third. The Saints will likely hold on next week unless Tampa Bay pulls an upset, but the Texans are in danger of falling at the hands of the Ravens. It's all fluid, so stay tuned, America.

The Janikowski Factor

Raiders kicker Sebastian Janikowski just banged home three 50+ yarders in his team's win over the Texans. A few weeks ago he tied the NFL record with a 63-yarder on a warm night in Denver. I was curious about just how much more successful he was at longer distances compared to the average kicker, so I dug up the data.

The graph below is his FG success by line of scrimmage (not kick distance) compared to the league average. His advantage, particularly at longer distances is evident. It looks like a about a 5 to 8 percentage point advantage at extreme ranges.

Saints' 8-Minute March

Up 10-6 early in the 2nd quarter, Drew Brees and the Saints offense took the field on their own 14.  At the time, the Saints had a 63% chance of winning the game.  What ensued was a 16-play, 8+ minute drive that ended in a Mark Ingram TD and an 83% chance of winning the game.  With the use of our Markov model, we take a look at the evolution of this marathon drive.

Sunday's Numbers Have Been Crunched

Sunday's numbers are now available, including advanced stat box scores, top players of the week, team stats, and season leader boards.

Roundup 10/8/11

Reader and Community contributor Jim Glass points us to this NFL Films cut on the 1978 passing rule changes.

Hue Jackson wants the NFL to become a running league again. Good luck with that. Darren McFadden is breaking out this season, and Jackson's play calling may be part of the cause according to Brian Goff. This is something economists call complementary inputs. I think of it as 'interaction' effects. As I discussed in my podcast with Carson, football is all about interactions between player performance.

Say we could rate players on a 1 to 10 scale. In a sport like baseball, if you have a player who is a 4 and a player who is a 9, you get a total of 13 units of production. But in football, if you have an offensive line that's a 4 and a RB that's a 9, you get 4 units of production, maybe 5. This is because the RB isn't going to be finding many holes to showcase his abilities. If you have an o-line that's a 9 and a RB that's a 9, you don't get 18. You get 81.

A Q&A I did with Robert Vega of Battle Red Blog (SB Nation Texans blog) on how to interpret advanced player statistics. We discuss Arian Foster, Matt Schaub, and Shaun Cody.

What's Going On With Passing In 2011?

Passing numbers have been on the rise since 1978, but they’ve exploded this season. In 2010, offenses averaged 222 yds per game, while so far this year they’re averaging 246. Compare that to the 2000 season in which offenses averaged 206 yds per game.

What’s going on with passing? Is it higher completion percentage? Is it deeper passes? Is it simply more pass attempts? Let’s take a closer look at the numbers to get a better idea of what might be causing the dramatic increase.

Before I get to my theories, I’m going to throw a bunch of graphs at you. Each graph looks at one or more dimension of the passing game by year. Pay close attention to the axis numbers on the left side of each graph so you’re not deceived by varying scales. First, let’s look at Expected Points Added per Play (EPA/P). This combines all the other stats into one understandable number representing the net point differential gained or lost by each pass play. This includes sacks, penalties, turnovers, and everything else. This will give us an overall idea of passing’s relative potency over the past decade.

The Weekly League: Notes and Ideas for Week Five

Titan quarterback Matt Hasselbeck invites you to talk to the hand

This week's edition of The Weekly League features

1. Bright and shining previews for this week's Tennessee-Pittsburgh, Green Bay-Atlanta, and Chicago-Detroit football games.

2. A table including each team's pythagorean record using Expected Points Added.


3. Reason.

Tennessee at Pittsburgh | Sunday, October 09 | 1:00pm ET
Four Factors

• The Titans currently have the third-ranked offense per GWP and yet it likely has absolutely nothing to do with running back Chris Johnson.
• To wit: Tennessee currently ranks last in run EPA (-18.1), run EPA/P (-0.19), and second to last in run success rate (31.2%).
• Matt Hasselbeck, meanwhile, is averaging 7.3 adjusted yards per pass -- third only to Aaron Rodgers (8.0) and Tom Brady (7.9).
• Unfortunately for the Titans, much of the passing success has been due to the presence of Kenny Britt (16.5 EPA, 0.63 EPA/P, 11.1 YPT), now on the IR.
• Still, Hasselbeck managed an 8.8 YPA last week, targeting four difference receivers at least four times.

WP: Redskins Defense Is Carrying the Team

Today's post at the Washington Post's Insider looks how special teams helped them squeak by the Rams last week. Plus:

  • Ryan Torain returns with a vengeance.
  • The defense has been carrying the team so far.
  • How likely are the 'Skins to make the playoffs?

Weekly Game Probabilities

Weekly game probabilities are available now at the Fifth Down. This week I look at the Jets-Patriots match up and discuss why great offenses tend to beat great defenses.

Podcast on FanGraphs

Carson recently invited me on his weekly podcast on FanGraphs. Jump on over to the coolest baseball site on the Web to hear Carson invent the adjective footbally, hear me forget Reggie Bush's name, and learn other stuff, including:

-similarities and differences in baseball and football analytics
-where my office is
-what the big inefficiencies in the NFL are
-why I was single for most of the lifespan of this site
-a thorough discussion of the NFL's running game
-and why so many Redskins fans hate me.

Ryan Kerrigan Backs Redskins Defense

The Washington Redskins are first in the East Division with a 3-1 record and the club has a rookie to thank for some of its success. The 16th overall pick of the 'skins out of Purdue, Ryan Kerrigan has adjusted well to his linebacker role after spending much of his time in college as a defensive end.

Kerrigan, a native of Muncie, Indiana, currently has the third highest Positive Win Probability Added (+WPA) at 0.83 amongst linebackers in the National Football League. Only Michael Boley of the New York Giants (0.87) and Sean Lee of the Dallas Cowboys (0.86) have performed better. The trio is the cream of crop in the NFL with the next highest +WPA coming in at 0.66 from Terrell Suggs of the Baltimore Ravens. Both Suggs and Boley are veterans of the league – a combined 16 years between them – while Lee is in his sophomore campaign.

Team Rankings: Week 5

Curious readers, here are your team rankings heading into the fifth week of NFL action.

The Top Dog: Dallas Cowboys
I know what you're thinking: "how can a 2-2 team with a negative point differential be the top team in the league?" While both of those facts are true, the Cowboys have also had an unfavorable schedule, as they have faced three top-12 clubs. Early in the season, this model will heavily weigh "strength of schedule," so there you go. Also, Dallas feels like one of those teams that can put up good numbers on a regular basis without winning many games. A paper team, if you will. The 'Boys get a break this week to prepare for the Patriots offensive attack.

Selected Recap: Ravens 34, Jets 17: All The Turnovers

Sunday Night Football, ostensibly, gives a primetime view to what should be a great matchup between two of the league's top teams. Sometimes, this ideal is foiled by circumstances outside of the team's control -- take, for example, Peyton Manning's injury lowering the quality of Week 3's Steelers-Colts matchup. But sometimes, a good matchup is there, and both teams play ugly, ugly football.

This was certainly the case in Week 4's Sunday Night matchup, as both the Baltimore Ravens and the New York Jets combined to fail to produce a 100 yard runner or a completion percentage over 33%. Both teams did manage to produce seven turnovers, however. This was the source of Baltimore's big victory: three of the four turnovers forced by the Ravens' defense turned into six points on the other end.

Cutler's Sack Struggles

Jay Cutler is a sack-magnet.  Cutler led the league in sacks last year with 52 and is currently the 2nd most sacked quarterback in the NFL with Sam Bradford taking that crown after a 7-sack performance Sunday.  Most of the blame usually falls either on his offensive line or on himself for holding the ball too long.  Both of these contribute to his sack-a-licious resume, but the biggest contributing factor may be the system in which he plays.

First, let’s look at the evolution of the shotgun offense.  In 2000, just under 20% of all pass plays were taken from shotgun.  As of 2010, almost 55% of all passing plays were taken from shotgun (57% so far this year!).

Should the Seahawks Have Kicked or Thrown?

At halftime, SEA trailed ATL 24-7. They managed an improbably 2nd-half comeback to within 2 points, trailing 30-28. SEA moved the ball into ATL territory in the closing seconds, but faced a 4th and 8 at the ATL 43 with 13 seconds left on a stopped game clock. SEA still had one timeout. They had enough time to risk a pass that could have dramatically improved their FG chances if successful. Instead, they chose to attempt a 61-yard FG. Was this the right call?

FG accuracy really falls off a cliff when the LOS is outside the 35. At very best, SEA could hope for a 20% shot from 61 yds, and that's very generous, allowing for solid footing and good weather.

We can typically expect 4th and 8 conversion attempts to be successful 37% of the time in that part of the field. SEA had a TO remaining, which means they could use the middle of the field. ATL would have to guard deep against a game-fatal TD, so I would put their chances at 37% at a minimum. A successful conversion means the LOS would be at the 35 at least, making the kick a 50/50 proposition at least.

Sunday's Numbers Have Been Crunched

Sunday's numbers are now available, including advanced stat box scores, top players of the week, team stats, and season leader boards.

The Weekly League: Notes and Ideas for Week Four

Dallas Cowboy linebacker Sean Lee: good at tackling.

This week's edition of The Weekly League features

1. Award-winning previews for this week's Detroit-Dallas, New York Jet-Baltimore, and Indianapolis-Tampa Bay games.

2. A table that compares each team's actual points versus their Expected Points Added.


3. GWPs, PROBs, and other crowd-pleasing acronyms.

Detroit at Dallas | Sunday, October 02 | 1:00pm ET
Four Factors

• As the table towards the bottom of this post indicates, the Cowboy defense has been better so far than the team's 69 points allowed would suggest.
• Consider, for example, how Dallas has allowed the sixth-fewest net yards per pass attempt (5.8) in the league, the fifth-fewest yeards per play (4.8), the sixth-lowest EPA per passing play (-0.01), and the sixth-lowest overall EPA per play (-0.05).
• If you have time, also consider how the defense is ranked seventh overall per defensive GWP.
• Now, consider how the team has allowed those 69 points, the 17th fewest (or 16th most -- however you like it) in the league.
• It's something that, much like a C+C Music Factory song, makes a person go "Hmmm."