I don’t know if any of us have ever seen a game quite like the end of Baltimore-Minnesota. With 2:05 left in the game, Baltimore faced 4th-and-goal from the Vikings 1-yard line. The Ravens trailed 12-7.

In what looked to be the game-winner, Joe Flacco hit Dennis Pitta for a touchdown pass. A two-point conversion to Torrey Smith put the Ravens up, 15-12.

On the next play from scrimmage, Matt Cassel hit Jerome Simpson for 27 yards. With 1:27 left, Toby Gerhart rushed up the middle for a 41-yard touchdown, which looked to be the game winner. The Vikings now led 19-15.

But Jacoby Jones returned the ensuing kickoff for what looked to be a game-winning, 77-yard touchdown, to put Baltimore back on top, 22-19.

Matt Cassel then threw a couple of incompletions, before throwing a screen pass to Cordarrelle Patterson…. that the rookie to the house for a 79-yard touchdown. That looked to be the game-winner, as Minnesota now lead 26-22.

But then Joe Flacco drove the Ravens down the field, and hit Marlon Brown for a nine-yard touchdown with four seconds left, in what was actually the game-winner. Baltimore left with a very unlikely 29-26 victory.

Add in the Cassel-to-Simpson touchdown on the second play of the fourth quarter, and that means there were six touchdowns in the final quarter that were lead-changing scores. That’s an NFL record.

Prior to this game, only four games saw five lead-changing touchdowns in the fourth quarter. A Bills-Raiders Ryan Fitzpatrick/Jason Campbell shootout from 2011, a Bruce Gradkowski-fueled comeback win over Ben Roethlisberger and the Steelers in 2009, a Monday Night thriller between the great Mark Sanchez and Chad Henne earlier that same season, and a Giants-Cardinals shootout from 1962. Hat/tip to the great Scott Hanson for alerting me to the record.

Just above these words, it says “posted by Chase.” And it was literally posted by Chase, but the words below the line belong to Steve Buzzard, who has agreed to write this guest post for us. And I thank him for it. Steve is a lifelong Colts fan and long time fantasy football aficionado. He spends most of his free time applying advanced statistical techniques to football to better understand the game he loves and improve his prediction models.

---

The way to win fantasy football games is to have players that score a lot of points.  Players tend to score more points when they get more touches.  One of the most important factors in determining how many touches each player is going to have is to determine the Game Script ahead of time.  As we all know positive game scripts result in more passing attempts and negative Game Scripts result in more rushing attempts.  But I am going to try to project the pass ratio using two key stats, Pass Identity rating and the Vegas spreads. We can use these projected pass ratios to build our own projections or at least look for outliers and figure out how to adjust players from their year to date averages.

Regular readers know what Pass Identity means. For newer readers, you can read here to see how Pass Identities are calculated.  But the quick summary is that Pass Identity grades allow us to predict the pass ratio of any game where the point spread is zero. This is because Pass Identity tries to eliminate the Game Script from the pass ratios.  For example since the Bears/Cowboys game is a pick’em this week, we can predict the pass ratio of the Bears by using the following formula.  League average pass ratio + (A + B) *C, where

(A) = number of standard deviations above/below average the Bears are in Game Script (-0.49);

 

(B) = number of standard deviations above/below average the Bears are in Pass Ratio (+0.53); and

(C) = the standard deviation among the thirty-two teams with respect to Pass Ratio (5.3%)

Of course, the product of (A) and (B) is the Pass Identity grade for each team; once we determine that, we multiply that number by the standard deviation of the pass ratios of all teams to get us a prediction for the pass ratio in a game with a Game Script of 0.0. Since the Bears have a Pass Identity of basically 100, the projected Pass Ratio for Chicago against Dallas is 58.9%.

We can then compare this projection to Chicago’s year-to-date pass ratio of 61.5% and predict that all else equal Jay Cutler and the passing game should score about 4% ^[1]Since 58.9% is 96% of 61.5%. less this week than their average week where as Matt Forte and the run game would score about 4% more.

[continue reading…]

I contributed to this New York Times graphic regarding the Jets struggles.

Let’s just assume that Auburn defeats Missouri this afternoon and Ohio State defeats Michigan State tonight. Which team would have pulled off the more impressive feat: Ohio State, going undefeated against a relatively easy schedule, or Auburn going 12-1 against a harder schedule? That’s a tricky question to answer, but here is one way to think about it.

To make the math easier for everyone — and the answer won’t be practically different otherwise — let’s eliminate the eight easiest games on each team’s schedule. For Ohio State, that means elminating wins over Florida A&M, Purdue, San Diego State, California, Buffalo, Illinois, Penn State, and Indiana. For Auburn, we remove wins over Western Carolina, Arkansas State, Florida Atlantic, Arkansas, Mississippi State, Washington State, Tennessee, and Mississippi. A team arguing that it should be the #2 team in the country is going to win those games over 95% of the time. Granted, this slightly disadvantages the Tigers as they had a slightly harder bottom eight, but you can include those games if you want to do more heavy lifting. For now, let’s just focus on each team’s toughest five games.

Ohio State will have gone undefeated against Wisconsin, Michigan State, Michigan, Iowa, and Northwestern. Is that more or less impressive than going 4-1 against Alabama, Missouri, LSU, Texas A&M, and Georgia? One way to can answer this question is by looking at a team’s win probability in each game.

Let’s assume that Ohio State has an SRS rating of 62.1. Why that number? You’ll see why in a minute. When the Buckeyes hosted the Badgers (SRS of 53.8), how likely was Ohio State to win? If we give three points for home field, that would make the Buckeyes 11.3-point favorites. And we can use the following formula to determine how likely an 11.3-point favorite is to win a given game:

(1-NORMDIST(0.5,-(home_fav),13.86,TRUE)) + 0.5*(NORMDIST(0.5,-(home_fav),13.86,TRUE) – NORMDIST(-0.5,-(home_fav),13.86,TRUE))

Based on this formula, an 11.3-point favorite would win 79.2% of the time. Against Michigan State (48.8), Ohio State would be a 13.3 point favorite if the Buckeyes had an SRS rating of 62.1, which translates into an 83.1% win probability. For Michigan, Iowa, and Northwestern, the spreads and win probabilities would be 15.4/86.7%, 20.3/92.8%, and 22.6/94.8%, respectively.

Now, what are the odds that Ohio State would win all five of those games? That is simply the product of 79.2%, 83.1%, 86.7%, 92.8%, and 94.8% — which is 50%. That’s not a coincidence, of course: the reason I picked 62.1 is because that’s what rating Ohio State would need to have in order to have a 50% chance of going undefeated against those five teams. In reality, the Buckeyes have a rating of 56.1, which indicates that — like just about every undefeated team — they were a little bit lucky to go undefeated (assuming, of course, that they beat Michigan State).

Now, let’s use that same 62.1 rating number to go through Auburn’s schedule. At home against Alabama (rating of 56.4), a team with an SRS rating of 62.1 would be a 5.7-point favorite, and have a 65.9% chance of winning. In Atlanta against Missouri (55.7), the team would be a 6.4-point favorite, and have a 67.8% chance of success. The team would be 8 point favorites in Baton Rouge — the game Auburn lost — against LSU (51.1), and have a 71.8% chance of winning. The games at Texas A&M (48.9) and at home against Georgia (48.5) would have 76.9% and 88.4% chances of victory.

Now, the odds of winning all five of those games is just 21.8%, which is a very long-winded, mathematical way of saying what we all know: Auburn faced a harder schedule. But what are the odds of going 5-0 or 4-1 against that schedule? Well, the odds of going 4-1 is just a bit more complicated.

• The probability of beating Missouri, LSU, A&M, and Georgia, but losing to Alabama, is 11.3%;
• The probability of beating Alabama, LSU, A&M, and Georgia, but losing to Missouri, is 10.4%;
• The probability of beating Alabama, Missouri, A&M, and Georgia, but losing to LSU, is 8.6%;
• The probability of beating Alabama, Missouri, LSU, and Georgia, but losing to A&M, is 6.6%; and
• The probability of beating Alabama, Missouri, LSU, and A&M, but losing to Georgia, is 2.9%.

Therefore, the likelihood of going 4-1 is 39.6%; that means the likelihood of a team with an SRS rating of 62.1 going 4-1 or 5-0 against those five teams is 61.4%. While there are many assumptions implicit in this post, the conclusion is that it is harder to do what Ohio State will do if it wins tonight than what Auburn will do.  Adding in the bottom 8 opponents for each team won’t change the numbers much (you can run the numbers using the above formula).

What would change the numbers is changing the ratings of some of the team’s opponents.  If, for example, Alabama had a rating of 69 instead of 56.4, then a team of a a quality equal to 62.1 would win that game only 38.9% of the time, and the odds of going 4-1 or 5-0 against that schedule would be 50/50. But that’s a pretty significant increase to Alabama’s grade, of course.

For a team to have a 50% chance of winning at least four out of five games against Alabama, Missouri, LSU, A&M, and Georgia, they would need a rating of 59.8. But a team with a rating of 59.8 would only have a 40.5% chance of not dropping a game to Wisconsin, Michigan State, Michigan, Iowa, or Northwestern.

Of course, I’ve followed college football long enough to not wait until Sunday to make this post. That’s because there is only a 30% chance of both Ohio State and Auburn winning today. We could perform the same analysis for Missouri, but the results would only look worse for the SEC crowd, as those Tigers have had an easier schedule than Auburn.  Assuming a rating of 62.1, a team would have a 36.8% chance of beating Auburn, Georgia, South Carolina, A&M, and Ole Miss, and a 78.0% chance of winning at least four of those games. In fact, a team would only need a rating of 56.0 to have even odds of going 5-0 against those teams.

The more interesting case, however, is Florida State. Assuming a rating of 62.1, the Seminoles would have a 69.8% chance of winning in Clemson, and then over a 90% chance of winning every other game (Duke will be the second toughest game of the year for FSU). That means a 62.1 SRS team would have a 53.0% chance of going 5-0 against Clemson, Duke, Florida, Pittsburgh, and Boston College; a team that had only a 50% chance would need a rating of 61.4, slightly lower than what Ohio State has produced.

That doesn’t mean Ohio State is more deserving of a spot than Florida State in the BCS National Championship Game, as FSU’s dominance is an element that can’t be overlooked. But I wouldn’t argue with you if you said that it was easier for FSU to go undefeated than it is for Ohio State.

For the second straight week, an NFC West team produced a monster game script. This week, it was Seattle dominating New Orleans and taking control of the NFC. The Seahawks can clinch homefield advantage throughout the playoffs by simply winning the team’s final two home games, and Seattle appears to be (again) getting hot just in time for the postseason. Among passers with at least nine starts, Russell Wilson has the second best ANY/A average behind Peyton Manning, and the team should have a healthy Percy Harvin for the playoffs. In other words, it’s going to take an incredible effort for a team to knock off the Seahawks, who also rank #1 in DVOA.

Six teams in week 13 won with negative Game Scripts, with Matt Ryan leading the biggest comeback of the week. In surprising twists, the Patriots and Cowboys trailed early before toppling the Texans and Raiders, while the Vikings came from behind late to defeat the Bears. The Jaguars won in the final minute in one of the more exciting games of the week, while the Giants won (in somewhat controversial fashion) after falling behind 14-0 early in Washington. Below are the Game Scripts data from each game in week 13: [continue reading…]

Nobody knows what power rankings are supposed to mean. And frankly, nobody cares. They just want to see lists. Are power rankings supposed to simply reflect records, in which case, what is the point of doing them? For example, I have cracked the code to ESPN’s power rankings:

• Step 1 – Rank teams in descending order of wins.
• Step 2 – Move San Francisco ahead of Kansas City (Chiefs are trending down!), San Diego ahead of Miami (even though Miami has won two straight, we had them really low two weeks ago, so we can’t move them that high), and move Tampa Bay ahead of Cleveland (Bucs are trending up, Browns are trending down!).
• Step 3 – For team with same number of wins, rank randomly, or based on the the best way to generate discussion.

I don’t see the point in doing power rankings that read just like the NFL standings page. Are power rankings supposed to reflect which teams we think are the best going forward? Perhaps you would like Advanced NFL Stats’ ratings, but that leads to situations where a team like the Ravens is ranked 25th despite being in line for a playoff perth. Which, of course, is either totally acceptable or makes no sense at all, with no middle ground.

Are power rankings supposed to reflect which teams have the best odds of winning the Super Bowl? You might as well use Football Outsiders’ playoff report and call it a day.

Instead, I’m going to make power rankings based on this method of measuring how each team played in each game relative to the performance by the team’s opponents in the rest of its games. The lower the rating, the better. You can view the historical ratings using this formula here.

[continue reading…]

This week at the New York Times, I look at what offensive records may be set this season.

Peyton Manning remains on schedule to break the single-season touchdown and passing records. With 41 touchdowns in 12 games, he needs 10 in his final four games to break Tom Brady’s single-season record of 50 passing touchdowns. The tougher record will be the yardage mark, set by Drew Brees in 2011 with 5,476 yards. Manning is on a pace for 5,500, leaving little margin for regression.

But we have come to expect superlative performances from Manning. Much more surprising is that Philadelphia’s Nick Foles has several records in his sights:

■ Foles, the Eagles’ backup quarterback to begin the season, has thrown for 19 touchdowns with no interceptions. The record for most touchdowns to start a season without an interception is 20, set by Manning this year.

■ Foles has thrown for a touchdown on 9.7 percent of his passes, the highest rate in the league. Since 1970, only three quarterbacks — Brady in 2010, Steve Young in 1992 and Ken Anderson in 1981 — have led the league in both touchdown rate and interception rate in the same season.

■ The highest touchdown rate in a season was produced by Sid Luckman of the Bears in 1943 (13.9 percent). Foles will not be able to get to that record, but he could set a post-World War II record (10.0 percent, set by another Eagle, Adrian Burk, in 1954) or a postmerger record (9.9 percent, by Manning in 2004).

■ The single-season passer rating record was set by Aaron Rodgers in 2011 at 122.5; Foles, remarkably, has a 125.2 rating with four games remaining.

■ The record for most pass attempts without an interception is 127, set by the Colts’ Paul Justin in 1996. Even if Foles throws one interception, he can set the record for interception rate among qualifying passers as long as he throws 245 or more passes. That record is held by Damon Huard, who threw one interception on 244 passes in 2006 (0.4 percent).

One other quarterback has a possible record in view: San Diego’s Philip Rivers has completed 70 percent of his passes this season. He would need a strong finish to break the record of 71.2 percent set by Brees in 2011. Even if he falls short of Brees, if Rivers continues to complete 70 percent of his passes, he will join Brees, Ken Anderson, Steve Young, Joe Montana and Sammy Baugh as the only quarterbacks to complete such a high percentage in a season.

You can read the full article here.

A couple of weeks ago, I presented another way to do team rankings. The method implicitly incorporates strength of schedule and margin of victory without having to do any hard math. For example, assume Team A hosts Team B and wins by 7 points. After adjusting for home field, Team A gets credit for winning by 4 points. The next step is to measure how Team B fared in its other 15 games. If Team B lost by more than 4 points in 4 other games, and won (or lost by less than 4) in its other 11 games, that would mean Team A had the 5th best result of the season against Team B. Therefore, we give Team A 5 points for this game. It’s that simple. You get credit for beating your opponent by more than other teams beat that opponent.

I don’t have a cool name for this sort of system, but I’m sure someone out there has been using this methodology for a long time and has already given it a name. So if you know it, post it in the comments. But I thought it would be fun to run through this method for every team since 1932. That’s what I’ve done in the table below. Keep in mind, though, that it’s only appropriate to compare teams who played the same number of games in a season. In a 9-game season, a team is obviously going to produce a much lower grade than a team in a 16-game season.

Here’s how to read the table below, which shows each team since 1932. It lists the top team in 2012, then the top team in 2011, then the top team in 2010, and so on, but you can use the search or sort functions to run whatever queries you like. In 2012, the Broncos ranked 1st in this system playing in the NFL (yes, that means I’ve got AFL and AAFC teams in here, too). The Broncos had an average score of 4.4 points. Denver had a win percentage of 0.831 that season, while playing 16 games (useful information when sorting), a 13-3 record. What’s the GR1 column? That means there was 1 Game where the Broncos Recorded a 1 — i.e., by delivering the biggest beatdown of the season (I also included games in this category if one other team delivered an equally-dominant performance against them). The Broncos ratings each week had a Standard Deviation of 3.1. I’m not quite sure what to do with the standard deviation column, but it was easy enough to include and might help you identify great teams that sat players in week 17.
[continue reading…]

New Orlean’s Drew Brees is officially listed as six feet tall. Seattle’s Russell Wilson is officially listed as 5’11. That means the average height of the starting quarterbacks in tonight’s game is 71.5 inches, tied for the shortest average in any game since 1964. In fact, it’s been twelve years since a game has featured two quarterbacks of such short stature, when in week two of the 2001 season, Doug Flutie (5’10) and the Chargers beat Anthony Wright (6’1) and the Cowboys.

The other two games since 1990 where the average height of the starting quarterbacks was below six feet also involved Flutie facing a 73-inch quarterback: a 24-21 win in 1999 against Pittsburgh and Kordell Stewart and a 17-16 win year earlier against Mark Brunell and the Jags.

Twenty-five years ago, two other Flutie vs. 6’1 Quarterback games make the list: this game against Jim McMahon and this one against Dave Krieg.

You have to go back to 1978 to find a game before tonight where (1) the average height of the starting quarterbacks was under six feet and (2) Doug Flutie was not involved. Fran Tarkenton (6’0) and Pat Haden (5’11) met five times in the mid-to-late ’70s, and Billy Kilmer (6’0) also faced Haden in the final game of the 1977 season.

Kilmer and 5’11 Bob Berry met three times in the early ’70s, and Sonny Jurgensen (5’11) faced Gary Cuozzo (6’0) and Tarkenton twice each. The only other games of the post-merger era were Len Dawson (6’0) vs. Berry in 1972 and Bill Nelsen and Edd Hargett in 1971. [continue reading…]

We are out of words. You should be dead, Auburn, because we saw you die. And here you are, breathing in the flesh, able to say this: you made the Alabama Crimson Tide kick the winning touchdown for you.

It’s hard to top that recap from EDSBS of one of the greatest games in college football history. Two weeks after pulling off the ending of the season — the Prayer at Jordan-Hare — Auburn gave us the ending of our lives. Entering week 14, Alabama had fielded the best special teams in the nation; on Saturday, all of the Tide’s goals were ripped from them following three missed field goals and a game-winning field goal return touchdown.

In a second, Alabama lost to its most bitter rival. With that, the Tide lost the SEC West division title, which means the team won’t have a chance to win the SEC Championship or the BCS Championship (barring the unthinkable). In an odd twist, the most dominant team of our era has now won just one division title in the last four years.

Of course, the SRS is not so sensitive to missed field goals that are returned for touchdowns. The Crimson Tide ranked third in last week’s SRS, a ranking which felt one spot too low. Following the Iron Bowl loss, Alabama’s rating dropped from 61.1 to 59.4, moving Nick Saban’s team down to… third. The beauty of a predictive system is that it need not change due to a close road loss to a top team, and that’s what happens here. Auburn jumps from 14 to 11 but no higher, as a 14-point road loss to LSU, a 4-point home win against MSU, and a 7-point home win against Washington State still count.

Ohio State, in fact, actually drops one slot, as the close win in Ann Arbor dropped the Buckeyes behind idle Oklahoma State. Does that mean the Buckeyes don’t deserve to be in the BCS National Championship Game if they defeat Michigan State? Of course not. Last year, Notre Dame was ranked 6th on December 9th in the SRS, but the Fighting Irish surely deserved a spot in the BCSNCG by virtue of being the lone undefeated (and eligible) team in college football. Ohio State deserves the same treatment this year.
[continue reading…]

Cleveland’s Josh Gordon caught 14 passes for 237 yards and a touchdown against the Steelers last week. Against the Jaguars this afternoon, Gordon caught 10 passes for 261 yards and two scores. In the process, he became the first player to ever record back-to-back 200+ yard receiving games, and set an NFL record with 498 receiving yards in two games.

The table below shows the 53 players to record 350 receiving yards in back-to-back games from 1960 to 2012. Until this year, Houston’s Andre Johnson had the modern record for receiving yards in consecutive games, set just last season. Then Calvin Johnson had 484 yards in two straight games, setting a record that stood for all of five weeks.

You probably didn’t know it, but Cam Newton is having a down year, at least statistically.

Carolina’s defense has been outstanding, of course, so an 8-3 record and a seven-game winning streak have overshadowed any flaws in Newton’s game. The Panthers have held an average lead of 5.05 points per second this year, the third best rate in the league. As a result of that high Game Script, Newton is asked to do less on offense, but that doesn’t explain the declining efficiency numbers. Newton’s taking slightly more sacks and his rushing numbers are down across the board, but the biggest decline comes with respect to yards per completion.
[continue reading…]