Since entering the league in 2011, Cam Newton has been one of the better quarterbacks in the NFL. But you wouldn’t know it if you looked at his completion percentage, which ranks as just the 36th best out of the 40 passers with the most attempts since the start of 2011 through week 3 of 2018.

In the past, I have offered up the idea that, in some instances, it may be appropriate to consider categorizing rushing attempts as equivalent to pass attempts completed to the quarterback. There are a couple of reasons for that. One, a mobile quarterback may scramble while an immobile quarterback would throw a check down to a running back; in other words, the plays are equivalent, but the immobile quarterback will have his completion percentage increased. Two, the perceived benefit to a quarterback with a high completion percentage is lower variability; rushing plays have lower variability, too, so labeling a rushing attempt as a pass to the quarterback helps reflect that.

One issue with this, though, is you need to remove kneels from the data. Thanks to Bryan Frye, we can do that. Let’s look at Newton. Since 2011, he’s completed 2,065 passes out of 3,515 attempts, a 58.7% completion rate. But he also had 783 carries for 4,545 yards (after removing kneels); if you count those as completed passes, his modified completion percentage would be 66.3%. If you want, you can also label his sacks as incomplete pass attempts. That would drop him down to 62.5%.

The table below shows the 40 quarterbacks with the most pass attempts since 2011. [continue reading…]

Previously:

• Week 1 (2018) Game Scripts
• Week 2 (2018) Game Scripts

The Buffalo Bills pulled off one of the most remarkable upsets in NFL history in week 3, beating the heavily-favored Vikings 27-6. That game was notable for a lot of reasons on the Bills side, but here’s something interesting on the Minnesota side: the Vikings had just six rushing attempts all game, including just one in each of the final three quarters of the game. Minnesota running backs finished the day with just four carries:

The Vikings had 59 passing plays, on the other hand, which means over 90% of all Minnesota plays in week 3 were passing plays (in fact, the two Cousins runs were one scramble and one fumbled snap, so the Vikings ended the day with only four designed running plays). How remarkable is that? Well, it’s the first time in the history of the NFL that it’s happened! The Vikings ran on just 9.2% of all plays (again, generously counting the two Cousins runs as running plays), breaking the previously low of 10.3%, set by the Cardinals against the… 2006 Vikings.

Also super pass-happy in week 3: the Indianapolis Colts, despite (or because?) Andrew Luck setting a career low in yards per completion.  The Colts-Eagles game was tied after the 1st and 3rd quarters, with Philadelphia holding just a 3-point lead at halftime.  In other words, this was a really close game…. and yet Indianapolis passed on 77% of its plays!  That is quite unusual, and represents an extremely pass-happy game (albeit one where the passes were mostly short passes that operated as runs).

The full week 3 Game Scripts data: [continue reading…]

Here was Matt Ryan’s stat line in week 3: 26 completions on 35 attempts, 359 passing yards, 5 TDs, 0 INTs, 146.3 passer rating, 12.08 ANY/A.

Somehow, Ryan’s Falcons lost. There have been 93 teams in NFL history to average at least 9.0 net yards per pass attempt, throw for 300 passing yards, and have a passer rating of at least 145. Those teams are now 91-2, with Atlanta on the winning side of things in a game against Pittsburgh in 2006.

Or how about this stat: what do you think the record is of quarterbacks who average 10 yards per pass attempt, with 5 TDs, and 0 INTs? Try 50-2, with Ryan now joining Dan Marino in one of his crazy games against the ’80s Jets.

Or what about this: 300 passing yards, no interceptions, and a passer rating of 145? Teams were 117-0 when their quarterbacks did that, until Ryan became the first one to ever lose such a game on Sunday.

So yeah, Matt Ryan had a remarkable game on Sunday, but his Falcons somehow lost. The reason? They were facing Drew Brees, who was nearly as good. This was the second time in three weeks that Brees and the quarterback facing the Saints were two of the best three quarterbacks of the week. If we want to play with a bunch of endpoints, consider: there have been 5 games this season, where a quarterback threw for 370 yards and 3 TDs, with a passer rating of at least 115 and a completion percentage of at least 73%.  Four of those five games came with the Saints on the field.

In between the best (Ryan) and 3rd best (Brees) performances of the week was Ryan Tannehill.  The Dolphins star had a wonderful game, even if his numbers were inflated a bit by a 74-yard touchdown “pass” and an 18-yard touchdown “pass” where the ball traveled about one foot from Tannehill’s hands each time.  Add in an Albert Wilson touchdown pass — he had a remarkable game, too — and it was one of the most efficient passing games in Miami history.

As for the league as a whole?  It was a less remarkable passing week than the historic numbers we saw in week 2, but it was still pretty darn efficient.  The league averaged 6.38 ANY/A, a passer rating of 94.6, and 257 passing yards per game. The table below shows the full week 3 results: [continue reading…]

Yesterday, Andrew Luck completed 25 of 40 passes for just 164 yards. That is an anemic 6.56 yards per completion average, making him just the 10th player to have such a low average on 25 completions. Once again, that’s Andrew Luck, who ranked 3rd in yards per completion as a rookie in 2012, ranked 5th in the metric in both 2014 and 2016.

Luck missed 9 games in 2015 with first a shoulder injury and then an abdomen/kidney laceration. In 2016, he had a frayed labrum in the preseason, but played in 15 games, missing just one game with a concussion. And in 2017, of course, Luck missed the entire year after shoulder surgery (to, you know, repair the shoulder that caused him to miss zero games in 2016).

Since he’s returned, Luck has had three of the worst games of his career in terms of yards per completion, culminating in yesterday’s career-low.  The graph below shows the yards gained per completed pass for each game of Luck’s career: [continue reading…]

It has been a horrendous start for the 2018 Arizona Cardinals.  On offense, the team is performing like an expansion team. Consider that through two weeks:

• The Cardinals rank 31st in ANY/A, ahead of only the Bills.
• Arizona ranks 31st in rushing yards, ahead of only the Saints.
• The Cardinals rank 31st in yards per play, ahead of only the Bills.
• Arizona ranks dead last in offensive yards, with just 350.

Things aren’t much better on defense.

• Arizona ranks 31st in ANY/A allowed, ahead of only the Saints.
• The Cardinals rank 28th in rushing yards per game allowed.
• Arizona ranks 28th in yards per play allowed and 30th in total yards allowed.

But perhaps it’s easiest to see how poorly Arizona’s done this year with a graph. I have plotted each of the league’s 32 teams in the graph below, with their total yards of offense through two games on the X-Axis, and total yards allowed on defense through two games on the Y-Axis. Obviously you want to be low and to the right — gaining a lot of yards while allowing few. Arizona, plotted below in Cardinals colors, is high and to the left. And like, really high and to the left: [continue reading…]

Last week, the Buffalo Bills had the worst quarterback in the NFL. And after switching from Nathan Peterman to Josh Allen, Buffalo had… the second-worst quarterback in the NFL in week two.  Sam Bradford was bad in week 1, but he was really bad in week two: the journeyman quarterback averaged under 5.3 yards per completion, the lowest of any quarterback since 2009.

Meanwhile, Ryan Fitzpatrick and Patrick Mahomes were once again two of the top three passers in the league.  The big story, of course, was that the league as a whole averaged an NFL record 6.99 ANY/A in week two; that means when Matthew Stafford throws for 347 yards and 3 TDs with no interceptions on 53 dropbacks, and averaged 7.07 ANY/A, he was basically average.  That is going to take some time to get used to.

The full week two results, below:

This Week In Completion Percentage Is Meaningless

Eli Manning ranks poorly in ANY/A this week despite a sparkling 75% completion percentage.  Do you know why? Because the Giants had one of the worst performances for a team that completed 75% of their passes, at least as measured by simple net yards per pass attempt. New York’s first five drives ended in punts, and the sixth ended in a fumble. The 7th drive was a field goal and the 8th drive was a 3-and-out, before the Giants scored 10 points on their final two possessions. Even still, New York averaged just 3.81 yards per play, the third worst performance by the Giants since the start of the 2015 season.

Stop the presses: Cleveland is a favorite this week. The Browns host the Jets on Thursday night, and as of Tuesday evening, are 3-point favorites. That is breaking news, and it would really be breaking news if the Browns won. The 2015, 2016, and 2017 seasons, along with the first two games of 2018, give us a nice 50-game sample for Cleveland. And over the team’s last 50 games, they have been favored just four times, and won only one of those games: a win over the Blaine Gabbert/Jim Tomsula 49ers.

On that day, Johnny Manziel was the team’s quarterback.  So yeah, it’s been a long time.  Of course, the Browns have only won 4 (and tied a fifth, of course) of those games, with the last coming back on December 24th, 2016.  That day, Robert Griffin III was the team’s quarterback.

So yes, it’s been a rough few years… or decade… or two decades, for Cleveland.  Over the team’s last 58 games, the Browns have been favored to win and actually won just one time — that game against the 49ers.  There have been a lot of bad days in Cleveland over the years, of course: the graph below shows each Browns game since 2008, along with the pre-game point spread.  Remember, a positive number means the team was the underdog.  In the graph below, you can see that most of the dots are above the 0 line, meaning Cleveland was usually an underdog.   The games that Cleveland won are in white dots with orange borders; the games that Cleveland lost are in brown dots with orange borders.

Do the Browns have a reason to be optimistic since the game is at home and the Jets will have to travel to Ohio on short rest? From 2012 to 2017, there were 87 games played on Thursday night excluding week 1, when both teams have equal rest. In those games, the home team won 49 times (a 0.563 winning percentage), although the home team was favored 52 times and a pick’em once. Of the 52 home teams that were favored, 37 won (0.711) and 28 covered (with 24 failing to cover). When the spread has been tight — the home team favored by between 1 and 3 points — the home team has won 9 of 16 games.

If the Browns manage to win the game, it will be just the second time Cleveland has been favored to — and actually won — a game in primetime.  The second-to-last time the Browns won a primetime game they were favored to win? They did it against a team that doesn’t exist anymore, with Bill Belichick as the head coach.

Week 2 is not yet in the books, but NFL teams had the greatest regular season day of passing the league has ever seen.  There were 28 teams playing yesterday and the group completed 711 passes out of 1,006 attempts — an astounding 70.7% completion rate! Those passes gained 7,848 yards, for a remarkable 7.8 yards per attempt average. And these players threw for 55 touchdowns against just 16 interceptions. The average team throws about 550 passes per season. On a per-550 attempt basis, this means the NFL yesterday was at 389 completions for 4,291 yards with 30 TDs and 9 INTs. And remember: this is the average quarterback.

As a whole, the league produced a 105.1 passer rating. Week 2 is not yet in the books, and the Thursday Night game drags the rating down slightly to 103.4, but no matter: week 2 will go down as the best week in NFL history by passer rating.

How do we know? Because prior to this week, the best single week in NFL history by passer rating came in week 10 of the 2010 season, at 94.8. The next three after that: week 12 of 2016 (94.1), week 3 of 2014 (93.9), and week 2 of 2015 (93.8). This week will blow every other week out of the water, and it’s not hard to see why: NFL defenses have been neutered. Penalties are out of control, according to some defenders, and this impacts the passing game in two ways. Quarterbacks are safer than ever in the pocket, as defenders are wary of getting a flag for hitting quarterbacks too low, too high, or even in the middle (a new rule prohibits a defender from using “all or part of his body weight to land on the quarterback immediately after the ball was thrown”). As a result, what used to be a hurried pass or a sack is now an opportunity for a quarterback to stand tall in the pocket and deliver an accurate throw.

In the secondary, things are just as tough, as defensive backs can’t hit defenseless receivers (and a “receiver is considered defenseless throughout the entire process of a catch, up until the player is capable of avoiding or warding off impending contact.”) The NFL also decided to more strictly enforce the illegal contact rule, meaning “a defender cannot initiate contact with a receiver who is attempting to evade him” after five yards. In other words, it’s harder for defensive backs to stop receivers from getting open, and it’s harder for pass rushers to sack quarterbacks.

This didn’t quite show up in week 1 (teams had a passer rating of 82.4). But that just makes this all the more remarkable. Last season, the league average passer rating was 85.1. It was slightly below that in week 1. And then, all of the sudden, a league that never hit a 95.0 passer rating for a week suddenly has a week where teams have produced a 103.4 passer rating. This remarkable turn of events needs to be monitored, as now the average quarterback is putting up Aaron Rodgers numbers (no, really: Rodgers has a career passer rating of 103.9).

On Sunday, there were 28 quarterbacks who threw passes. If you had a passer rating under 105, you ranked outside of the top 14. A remarkable 16 of 28 quarterbacks had a passer rating over 100 (17 if you remove a Nick Foles spike to stop the clock), and 23 of 28 had a passer rating over 95!

Just 5 of the 28 quarterbacks had a passer rating below the 2017 average of 85.1, which might be the most remarkable of all these stats. Welcome to the 2018 NFL, September 16th edition, where over half of the quarterbacks completed over 70% of their passes.

The single most important question facing the NFL right now: is this new level of efficiency here to stay? If so, we are about to enter an era of football that is significantly different than the league that Tom Brady entered in 2000, where the average team completed 58.2% of their passers and had a passer rating of 76.2. But this is what I suspect the NFL wants: quarterbacks and receivers are less likely to get injured, teams will continue to shift towards passing over running, and quarterbacks will look like superstars with record-breaking numbers.

In the last 8 days, Ryan Fitzpatrick — sorry, journeyman, Harvard graduate Ryan Fitzpatrick — just produced two of the best games of his career.  He had one of the best games in recent memory in week 1, completing 21 of 28 passes for 417 yards and 4 TDs with no interceptions or sacks.  Today, against the defending Super Bowl champions in week 2 of the 2018 season, Fitzpatrick went 27 of 33 for 402 yards with 4 TDs, although he did add one interception and two sacks for nine yards.

Fitzpatrick averaged 17.75 ANY/A in week 1 on 28 dropbacks; if we use the NFL average ANY/A from 2017 as our estimate of what the average will be in 2018, this means he produced 332 Adjusted Net Yards of Value over average. Today, he averaged 12.49 ANY/A on 33 dropbacks, which is a still remarkable 230 ANY over average.

Fitzpatrick started his career with the Rams in 2005, then the Bengals, then the Bills, Titans, and Texans. In Houston, Fitzpatrick had the other remarkable game of his career: he went 24 for 33 for 358 yards and 6 TDs, averaging 14.48 ANY/A in a game against one of his many former teams, the Titans. Given the league average in ’14 of 6.14 ANY/A, this means Fitzpatrick had 276 ANY of value over average.

Then Fitzpatrick went to the Jets, and while he had some good games, it also produced the worst game of his career: a 6-interception disaster against the Chiefs that produced -356 ANY over average.

The graph below shows — color-coded, of course — each game of Fitzpatrick’s varied career.

Are Fitzpatrick’s last two games a mirage? You take a look at the graph and tell me…

The Game Scripts are back! Below are the Game Scripts from week 1:

[continue reading…]

There are seven new head coaches this year: Jon Gruden in Oakland, Matt Nagy in Chicago, Pat Shurmur in New York, and Frank Reich  in Indianapolis.  On the defensive side, Mike Vrabel is in Tennessee, Steve Wilks is in Arizona, and Matt Patricia is in Detroit.  In week 1, all seven teams lost, and other than Nagy’s Bears (who nearly upset the Packers despite being 6.5-point underdogs), none of the teams even covered against the spread.  Five of the seven teams lost at home.

The table below shows the 15 teams that lost in week 1, and how they fared against the spread: [continue reading…]

The Ravens blew out the Bills on Sunday, 47-3, giving Baltimore the biggest win of week one. Baltimore was a 7.5-point favorite, so the 44-point win means that the Ravens covered by 36.5 points.

That’s a lot. Last year, the Rams shocked the Colts with a 46-9 win, covering by 33.5 points.

Last night? The Jets began the Sam Darnold experience in the most ugly way possible: the rookie quarterback from USC threw a pick six on his first pass, something Jets fans are all too familiar with.

And then? It was a script for the ages for the Jets. Darnold went 16/20 for 198 yards with 2 TDs and 0 INTs the rest of the way. Free agent acquisition Isaiah Crowell rushed for 102 yards and 2 TDs on just 10 carries. Quincy Enunwa returned from a devastating injury and had 6 catches for 63 yards and a touchdown. 2016 first round pick Darron Lee, who has had an up and down career, recorded two interceptions, including a pick six. 2017 first round pick Jamal Adams picked up his first career interception. Trumaine Johnson, the team’s big acquisition at cornerback in free agency, had an interception; so too did the team’s big acquisition at cornerback in 2017, Morris Claiborne. Oh, and the Jets added a 79-yard punt return from Andre Roberts.

The Jets won 48-17, with the game ending as both team took knees near the Lions goal line. New York was a 7-point underdog, meaning the Jets covered by a whopping 38 points. That’s more than the Ravens this year or the Rams last year; in fact, it’s the fourth largest cover in opening week history!

In 1987, the Bucs, as 2-point home underdogs to Atlanta, beat the Falcons 48-10, covering by 40 points.

In 1998, in a game that is near and dear to Jets fans’ hearts, New York stunned Seattle 41-3 as 6.5-point underdogs (44.5 point cover) in the opening game of the Bill Parcells era.

Finally, in 1989, the Browns shut out the Steelers 51-0 despite being just 2-point favorites.

For New York, it was the 4th-largest cover in franchise history (or, at least, going back to 1978), behind the Seattle game in ’98, this upset over Houston in 1988, and this upset over the 2002 Chargers.

It was a game that was shocking on just about every level.  The Jets hadn’t scored a defensive or special teams touchdown since 2013; they did both on Monday night.  Since 2000, the Jets had thrown 5 or more interceptions in a game three times, but hadn’t done so to an opposing quarterback since 1999.  They did that to a Pro Bowler in Matthew Stafford (4 INTs) and his backup (Matt Cassel) on Monday Night. The Jets draft history has been disappointing in recent years, but then the team’s last three first round picks all had big nights.

Will this last for the Jets? Probably not. But for one night, it was the game Jets fans dream about. [continue reading…]

Things got really ugly this week in Pittsburgh, as Le’Veon Bell remains in a dispute with the Steelers over his franchise tag. Bell is currently refusing to play under the tag, and there is no immediate end in sight.

There is a long-term end in sight, tho: Bell will need to report for the final six games of the season in order to accrue another season of play; otherwise, Pittsburgh could just franchise Bell yet again for the same $14.5M tag after the season.

Bell will come back in November at the latest, which will also make him available to play in the postseason. And that’s where all of this could get interesting. Is it possible that Pittsburgh might wind up better off in the playoffs (assuming they get there, and an opening day tie against the Browns doesn’t engender confidence) if Bell doesn’t have a full workload behind him?

I’m thinking back to Bob Sanders and the 2006 Colts. The hard-hitting safety was one of the best defensive players in the NFL in his prime, but was rarely healthy. In 2005, he was an All-Pro safety; in 2007, he was the AP Defensive Player of the Year. In between? He missed most of the 2006 season due to injury, and the Colts defense suffered for it. Indianapolis ranked 21st in yards allowed, 23rd in points allowed, and 32nd in rushing yards allowed. [continue reading…]

Nathan Peterman is the Bills week 1 starter, beating out A.J. McCarron (who signed a 2-year, $8.1M contract with Buffalo in March, and received $4M from the Bills without ever taking a regular season snap before being traded to Oakland on three days ago) and rookie Josh Allen, taken with the 7th pick in the 2018 Draft.

Which is… well, unusual to say the least.  Peterman does not have a strong pedigree nor a track record of success. He was the 5th round pick in the 2017 Draft; as a rule of thumb, 5th round picks don’t start week 1 games for teams unless they have had some success. So, how did he do last season as a rookie?

[continue reading…]

I looked at all players (excluding quarterbacks) since 1990 with at least 50 carries in a season. I then grouped those players into 50-carry increments based on their number of carries that season: i.e., 50-99 carries, 100-149 carries, and so on. The chart below shows that data, along with the average number of carries for each group, the average number of rushing yards, and the average yards gained per carry:

Those results are probably not very surprising. The players with the most carries have the most rushing yards, and the yards per carry average tends to increase, too. This is in some ways an example of survivorship bias: the players who are performing the best will continue to keep getting carries, moving them into the higher-carry buckets.

Now, what happens the next season? Take a second and think about what you expect…. [continue reading…]

There were 644 fumbles during the 2017 season, with 276 being recovered by the opposing team. That’s a 43% fumble recovery rate for the defense, and a 57% fumble recovery rate for the offense; in other words, exactly 3 out of every 7 fumbles were recovered by the opposing team, and 4 out of every 7 fumbles were recovered by the fumbling team.

The Baltimore Ravens fumbled 19 teams, but only lost 4 of those fumbles. So the Ravens offense recovered 79% of those fumbles, when we would have “expected” them to recover 10.9 of those fumbles. Therefore, Baltimore recovered 4.1 more fumbles than we would have expected. On defense, Baltimore’s opponents had 22 fumbles, and Baltimore recovered 12 of those 22 fumbles. So the Ravens defense recovered 55% of all fumbles by opponents, when we would have “expected” them to recover 9.4 of those fumbles; therefore, Baltimore’s defense recovered 2.6 more fumbles than expected. Add it up, and there were 41 times that the football hit the ground during Ravens games, and Baltimore recovered 27 of them, which was 6.7 more than expected.

On the other side of things we have the New Orleans Saints. Despite being one of the best teams in the NFL last season, New Orleans had really bad fumble luck (perhaps that’s why they “only” went 11-5 despite being the most efficient team in the league). On offense, the Saints had 19 fumbles but only recovered 9 of them; that’s 1.9 fewer fumbles recovered than expected. And on defense, New Orleans forced 20 fumbles but only recovered 5 of them, which was 3.6 fumbles below average! Together, the Saints recovered just 14 of 39 fumbles in all games, which was 5.4 fewer fumble recoveries than expected. [continue reading…]

A question for the crowd today: which is a better way to present data on the NFC East passing attacks since 2002? The measure today will be Adjusted Net Yards per Attempt.

We can do it in graph form, like this, with each team having color-coded lines and ANY/A on the Y-Axis:

Or we can do a heat graph that shows the actual data, with blue shading for the best years and red shading for the worst years:

If you wanted to tell a story about the NFC East passing since 2002, what story would you tell, and which graph is more useful?

Who are the worst passing teams of the last three years? The Browns are the obvious and correct choice as the worst passing attack, and you won’t hear arguments from many if you include the Texans, Ravens, and 49ers in that group. Those four teams have all averaged 5.7 or fewer net yards per passing attempt since 2015, the simplest measure of passing efficiency.

But the fifth-worst team over the last three years is a shocker: it’s the Green Bay Packers. Yes, Aaron Rodgers missed most of last season (and Brett Hundley was terrible in his stead), but you may not remember that the Packers offense had a lot of struggles in 2015 playing without Jordy Nelson, who has been very instrumental to Rodgers’s success. Yes, Rodgers had his always glowing TD/INT ratio in 2015, but he ranked 32nd out of 36 qualifying passers in NY/A that season.  And Rodgers’s struggles creeped into September of 2016, too, before he finally turned things around.

Still, we think of him as Aaron freakin’ Rodgers, so it’s jarring to see that — even with half a season of Hundley — Green Bay ranks in the bottom five of any passing stat.  To be sure, NY/A has always been Rodgers’s weakest stat, since his TD rate and INT rate are what have buoyed his success. So let’s instead look at Adjusted Net Yards per Attempt, a stat that Rodgers remains the career leader in since 1970.

The graph below shows the trailing 16-game ANY/A average of the Packers passing attack over each 16 game period beginning with the 16th game of the 2008 season (Rodgers’s first as a starter).  The Packers line is in green and yellow; the league average is in black.  As you can see, things are not trending in the right direction, and even as of the middle of 2016, the T16G average was pretty ugly: [continue reading…]

Let’s travel back to 1990. On average, most defensive backs weighed between 172 and 210 pounds, most linebackers between 225 and 250 pounds, and most defensive linemen between 260 and 290 pounds. The graph below shows the amount of AV produced by defensive players at each weight in 1990:

If you look carefully, you’ll notice a few low spots on the graph. Very little AV is coming from players who weighed between 211 and 220 pounds, and also at 256 to 259 pounds. Let’s graph this another way: below, I show the percentage of all defensive AV produced by players who are X pounds or lighter. For example, about 11% of AV is produced by players 187 pounds or lighter, about one-third of AV is produced by players 210 pounds or lighter, and 34% is from players 220 pounds or lighter. The graph gets very flat between 211 and 220 pounds, indicating the lack of players in that range: [continue reading…]

Paxton Lynch was the 26th pick in the 2016 Draft, but it looks like his Broncos (and perhaps NFL) career is coming to a close. As a rookie, he sat on the bench behind Trevor Siemian, who was taken with the 25t0h pick in the 2015 Draft and vastly outplayed Lynch during the summer of ’16 to win the job. And in 2017, Lynch missed time with shoulder and ankle injuries, but he also was the third most productive QB on a bottom-5 passing offense, looking worse than both Siemian and Brock Osweiler. Lynch has now fallen to third on the team’s depth chart even with Siemian and Osweiler gone: Case Keenum was brought in to be the team’s starter, and Chad Kelly — the 253rd pick in the ’17 Draft — has moved ahead of Lynch on the depth chart.

That’s right: the Broncos 1st round pick in 2016 has now been beaten out by 7th round picks from both the 2015 and 2017 drafts.

Let’s assume Lynch finishes 2018 with zero touchdown passes.  That would give him 4 career touchdown passes through three seasons.  The graph below shows quarterbacks drafted from 1992 to 2015 plus Lynch, with the X-Axis representing draft slot and the Y-Axis showing touchdown passes through three years.  Lynch is the red dot. [continue reading…]

Back in August, I asked you 38 questions that served as prop bets for the 2017 NFL season. Thanks to the tireless work of Jeremy De Shelter, who helped compile all the results. Earlier this offseason, I looked at Part I and Part II. Let’s move on to Part III…

Number of playoff (non-Super Bowl) games won by the visiting team, +0.5
Maximum number of TDs thrown by Kirk Cousins in a single game

Cousins topped out at 3 passing TDs in a single game, done four times in 2017.  Meanwhile, road teams won… 3 games in the 2017 playoffs, with Atlanta and Tennessee winning on the road in the Wild Card round, and Jacksonville winning in Pittsburgh in the second round.

That Steelers loss was critical: only 39% of you picked the road playoff wins side, which means the Jaguars upset gave the minority group the win.

Punts by Jets opponents, -5
Jets team offensive passer rating.

The Jets were supposed to be terrible on offense and not too bad on defense, making this a tricky one to analyze.  The votes here were pretty split, with 53% of contestants voting for the Jets offensive passer rating side to be higher.  Jets opponents had 87 punts in the league last year the 4th most behind Jacksonville, Arizona, and Denver.

Meanwhile the Jets passing attack was surprisingly…. decent? New York finished with an 86.1 passer rating, slightly above the league average of 85.1, and 15th-best in the NFL.

In other words, this was a good line where the hook made the difference! The line finishes 82 vs. 86.1, swinging the side towards the brave 47% who backed the Jets passing attack. [continue reading…]

Yesterday, I looked at Norm Van Brocklin and how he set the single-game passing record with 554 passing yards.  This was done way back in 1951, in a game against the New York Yanks.  In that post, I noted that big passing games in Van Brocklin’s era tended to come against bad teams in blowouts, while big passing games now come in more competitive games. Let’s investigate that a bit more today.

In the 1950s, there were 10 games where a team threw for at least 400 gross passing yards (that is, without deducting sack yards). In those games, the average team threw for 444.5 yards, while the opponent had just 159.3 passing yards. And the 400-yard passing team led by, on average, 5.4 points, 13.3 points, 24.6 points, and 26.6 points after each quarter.

In other words, those were one-sided affairs where the winning team was able to name its score (and number of passing yards).  Van Brocklin’s game against the Yanks is a good example; in modern times, this is much less common, with the Patriots/Titans snow game from 2009 being an outlier (New England passed for 442 yards, while Tennessee had negative passing yards even without including sacks!).

Let’s compare that to the 2010s. There have been 126 passing games of 400+ yards, with an average of 436.6 passing yards. On average, the opponents in those games had a 300-yard game — 306.9 passing yards, to be exact. And the games were almost always close, with the margin being within 2 points at the end of each quarter (in fact, it was negative for the big passing team). [continue reading…]

As noted yesterday, red zone performance is mildly correlated with team success. One team that was a big outlier last year was the Steelers. Pittsburgh’s offense made it into the red zone 63 times, but converted those trips into touchdowns just 32 times (50.8%), slightly below-average. On defense, the Steelers allowed touchdowns on 24 of 39 red zone trips, a 61.5% rate that was the 5th-worst in the league. Of course, the Steelers were actually one of the best teams in the league.

On one hand, this seems kind of silly: of course red zone performance alone won’t tell us much about a team’s record! It says nothing about a team’s turnover rate, how effective the team is at producing or preventing big plays, overall team efficiency, or special teams. Perhaps the most remarkable part is that it does tell us quite a bit.

The table below has a lot of information, so to analyze it, let’s look at the best red zone team of the modern era, the 2005 Seattle Seahawks. That year, the Seahawks offense had 60 red zone trips and scored a touchdown on 43 of them, a conversion rate of 71.7%.  Given that 53% of red zone trips yield a touchdown, this means Seattle’s offense scored 11.1 more red zone touchdowns than expected. On defense, Seattle faced 47 red zone trips, and allowed just 19 touchdowns, a 40.4% conversion rate that was 6.0 touchdowns better than average. Therefore, overall, the team’s red zone performance yielded 17.1 touchdowns better than average, the most of any team since 2002.  Based on the best-fit formula derived yesterday, a team’s expected winning percentage based on its red zone team value is 0.0154*(Team Value) +0.500.  Since Seattle’s red zone value was +17.1, we would have expected the Seahawks to win 76.3% of their games; they actually won 81.3% of their games, a difference of 0.049.

[continue reading…]

Today’s guest post comes from Tom Nawrocki, a longtime fan of the site and not of the Dallas Cowboys. What follows are his words….

I grew up in the 1970s, watching the NFL and hating the Dallas Cowboys, as all right-thinking Americans did. The Cowboys were consistently strong throughout the decade; they made the Super Bowl after the 1970, 1971, 1975, 1977 and 1978 seasons.

What was additionally frustrating for us Cowboy haters was the way they kept adding top-flight talent throughout the decade. Despite the fact that they were a perennial playoff team, they seemed to have a top draft choice nearly every season. In the middle of this run, the Cowboys had the No. 1 or No. 2 pick in the draft for three out of four years, and the fact that they made the most of those choices helped give their dynasty a second wind. But how were they able to get those top draft picks, when they were successful every year? [continue reading…]

In June, I published the Game Scripts results from every game of the 2017 season. A team’s Game Script is simply the average points differential over every second of every game. The largest Game Script of the season came in the Rams/Seahawks game in Seattle last year that the Rams won by a score of 42-7. Just as impressive as the final score was how Los Angeles got up on Seattle early, which is how a team gets a very high Game Script: LA was up 13-0 after the first quarter and 34-0 at halftime, and finished with a Game Script of +23.4. This means, on average, the Rams lead by 23.4 points over the course of the 3600 seconds in that game.

Not surprisingly, LA only passed on 36.8% of plays in that game. In general, as Game Script goes up, Pass Ratios go down. In 2017, for every point of Game Script, a team would be expected to pass about 0.72% less often. A team with a Game Script of 0 would be expected to pass on 57.6% of plays; if the Game Script was -10, it would be expected to pass on 64.8% of plays, at -5, 61.2%, at +5, 54.0%, and at +10, 50.4%.

Last year, the Dolphins were the most pass-happy team in the NFL. Miami finished the year with 602 pass attempts, 33 sacks, and just 360 runs; in other words, the Dolphins passed on 63.8% of all plays last year. If you take the calculate the average pass ratio for the Dolphins in each game, Miami has a 63.9% pass ratio (I have decided to use an average of the average approach for this post, rather than an average of the gross). That was 1.74 standard deviations above average last year, since the average pass ratio was 57.6% and the standard deviation was 3.6%. However, the Dolphins had an average Game Script of -5.27, which was 1.65 standard deviations below average. The average Game Script, by definition, is 0, and the standard deviation last year was 3.20.

So Miami wasn’t particularly pass-happy once you account for Game Script. But you know who was? The team with the Hall of Fame quarterback, Hall of Fame tight end, and star young wide receiver. And in addition to Tom Brady, Rob Gronkowski, and Brandin Cooks, three of New England’s top four running backs are known just as much (if not more) for their receiving prowess than their rushing ability: Dion Lewis, who caught 91% of his targets last year, Rex Burkhead, who had 264 rushing yards and 254 receiving yards, and James White, who had 56 receptions and just 43 carries.

By traditional numbers, New England ranked 16th in the league in pass ratio. But the Patriots are the Patriots, a team that generally plays with the lead. New England finished the year with a Game Script that was +1.66 standard deviations above average and a pass ratio that was 0.29 standard deviations above average. Add those two numbers, and New England’s Pass Identity was +1.95, easily the strongest in the league. The table below shows the results from every team last season. [continue reading…]

The 1991 Eagles had a very bad passing offense. Philadelphia was one of four teams to finish with an ANY/A below 4.00, wasting a legendary defense along the way. You may have already known that, but here’s something you may not have known: Jim McMahon was Philadelphia’s starting quarterback that season, and he had a good season, ranking 12th in ANY/A and 13th in passer rating.

So yes, Philadelphia’s starting quarterback was an above-average passer despite Philadelphia’s passing attack being terrible. How did that happen? Well, it’s pretty simple: McMahon was responsible for 61% of the Eagles pass attempts that year, but also 66% of the team’s completions, 71% of the passing yards, 71% of the passing touchdowns, and only 47% of the sacks and just 41% of the interceptions. McMahon averaged 5.59 ANY/A; the rest of the Eagles passers averaged 0.77 ANY/A! That’s not a typo:

Thought of another way, non-McMahon passers had 226 dropbacks.  Using McMahon’s average of 5.59 ANY/A, we would “expect” McMahon to have produced 1,263 Adjusted Net Yards on those dropbacks.  In reality, other Eagles passers produced just 175 Adjusted Net Yards on those dropbacks, in large part due to 16 interceptions on 202 pass attempts.  This means McMahon “would have” produced 1,088 more Adjusted Net Yards than his backups.

That’s… a lot.  In fact, it’s the second-most in history using the following methodology:

• 1) Calculate the ANY/A for each passer on each team.  So for 1991 McMahon, it’s 5.59.
• 2) Calculate the ANY/A for the rest of that team’s passers for each passer.  For 1991 McMahon, it’s 0.77.
• 3) Subtract the result in step 2 from the number in step 1. This leaves us with 4.82 in the case of ’91 McMahon.
• 4) Multiply the result in step 3 by the smaller number of (a) that passer’s number of dropbacks and (b) the total number of dropbacks by the rest of the team.  In ’91 McMahon’s case, we use (b), which is 226, to get a result of 1,088.

The biggest difference in NFL history wasn’t McMahon, but Dan Fouts on the 1983 Chargers.   He averaged 7.32 ANY/A and threw 347 passes.  His backup, Ed Luther, averaged 3.77 ANY/A and threw 287 passes.  Fouts threw 20 TDs and 15 INTs, while Luther had 7 TDs and 17 INTs!  Non-Fouts passers in 1983 on San Diego had 308 dropbacks and averaged 3.66 ANY/A, exactly half of Fouts’ average!  Therefore, Fouts had 1,130 Adjusted Net Yards above expectation that season.

Three players from 2017 make the top 75 using this methodology, and my hunch is you could guess them pretty easily.  Then again, this formula isn’t supposed to shock you: it’s just a way of measuring which teams had a really good passer play about half a season, and really bad passers the rest of the season.

Last week, I looked at how many players were active for at least one game by class year, both before and after the current CBA was adopted in 2011. I also analyzed the production of players (by class year) from 2005 to 2010 and from 2012 to 2017 using Approximate Value and looking at just defensive players.

I’m short on time today, but wanted to do a quick search. I looked at all players who were in years 5-9 of their career and played in a given season. Then, I looked to see how many of those played in a single game the next year. The results are consistent with the other studies I’ve done: players in years 5-9 are losing jobs the following year. About 25% of players who were in their 5th, 6th, 7th, 8th, or 9th years in 2016 were not on an active roster in 2017. That rate used to be about 20%.

More time to come, but let me know if you have other thoughts on how to measure this.

Yesterday, I looked at how many players were active for at least one game by class year, both before and after the current CBA was adopted in 2011. Today, I want to do the same thing — analyze the production of players (by class year) from 2005 to 2010 and from 2012 to 2017. But instead of using games, I’ll use Approximate Value, and instead of using all players, I will use just defensive players.

On the X-Axis is all defensive players by class year, with the 2012-2017 players in blue and the 2005-2010 players in blue.  On the Y-Axis is percentage of AV by all players.  In general, second, third, and fourth year players produce the most AV, and that was true prior to the current CBA.  But the new CBA imposes an artificially low cap on the amount that can be paid to players in their first four seasons (with few exceptions).  As a result, teams seem to be giving more playing time and roster spots to players in their first four years, with 7th-8th-9th year players losing playing time: [continue reading…]

A few years ago, I looked at Eli Manning and the HOF in the context of an ANY/A Gray Ink test. What do I mean by that? Here’s what I did:

• Step 1) Calculate each quarterback’s ANY/A for each season of his career where he had enough pass attempts to qualify for the passing title (14 attempts per team game). ANY/A, of course, is calculated as follows: (Passing Yards + PassTDs * 20 – INTs * 45 – Sack Yards Lost) / (Pass Attempts + Sacks).

• Step 2) For each quarterback, award him 10 points if he led the league in ANY/A, 9 points if he finished 2nd, 8 points if he finished 3rd, … and 1 point if he finished 10th. A quarterback receives 0 points if he does not finish in the top 10 in ANY/A or does not have enough pass attempts to qualify.

• Step 3) For each quarterback, add his “points” from each season to produce a career grade.

I decided to update that post (written in 2014) and make a few changes.

1) I have included the results from the last four seasons.

2) I included sack data from 1960 to 1968, using estimated individual sack data based on team sack data.

3) I have lumped together the AAFC/AFL with the NFL in each season as if it was all one big league.

4) I have pro-rated the values based on the number of qualifying passers in each season. So let’s say a player ranks 4th in ANY/A in 2017, where there were 32 qualifying passers. He gets the average of 7 points and 7 points * 32 divided by 32, which is of course just 7 points. Now, let’s say a player ranks 4th in ANY/A in a 10-team league. He gets the average of 7 points and 7 points * 10/32 (2.18), which is 4.6 points. This is not a special formula, but a “good enough” metric that helps discredit players in smaller leagues, but not overly so (hence the average idea).

Okay, enough words. Here are the results. I have also included the raw totals, so you can see what happens if I didn’t pro-rate the values. Finally, I subjectively included some HOF labels at the end: actually in the HOF (not very subjective), Future HOFer, Borderline, and then either Not Eligible or a No for the HOF.

The future HOFers rank 1-4-5-13-18 by this litmus test: there’s not much to debate there.

Among the actual HOFers, only George Blanda — who is probably the least qualified quarterback of the “modern” era to make the HOF — ranks outside of the top 50. Bob Waterfield and Ken Stabler are the next lowest quarterbacks, and that’s consistent with how I’d view them. As a litmus test, this does a decent job for being a (somewhat) quick and dirty way to measure HOF play.

Among the Not Eligible guys, only Carson Palmer is in the top 60. He ranks 37th, on the back of a #1 season in 2015 with the Cardinals (+10.3 points, since there were 34 qualifying passers that year), a #3 season with the Bengals in 2005 (+8.25, also 34), and a 6th and 7th rankings in ’06 and ’14 (+5, +4.06). That’s not really a HOF career by any stretch, but it’s a memorable career.

Among the pure no guys — those who have been passed over and didn’t receive my subjective borderline label — you have Charlie Conerly at 26, Trent Green at 27, Boomer Esiason at 29, John Hadl at 31, Tommy Thompson at 33, and Milt Plum at 35. All good quarterbacks, occasionally great ones, who are HOVG (at worst) type players.

And then we get to the borderline guys. I gave 10 players that label, including three guys who entered the league in the last ten years and who are probably too young to really evaluate. Among the other 7…

Philip Rivers, Ken Anderson, and Tony Romo all are in the top 25, and make sense to be discussed together. None of the three won a Super Bowl, all three had fantastic efficiency numbers, and all three are more favored by the analytics crowd than the non-analytics crowd. Statistically, based on regular season efficiency, all three are clear HOF players. But, of course, that’s not the HOF test.

Steve McNair, Donovan McNabb, and Randall Cunningham are all borderline guys, too, and are underrated by an analysis like this that ignores rushing. They rank 68th, 72nd, and 104th. McNair (2003) and Cunningham (1998) each have a first place finish in ANY/A, while McNabb has a 2nd (2006) and 3rd (2004) place spot. McNair also has a 5th (2001) and a 9th (1999), while Cunningham has a 10th (1990), but that’s it, and that’s the problem. McNabb and McNair each have three finishes at 11 or 12 (worth zero) and McNabb has three more at 13/14; this analysis ignores solid seasons and rushing, which is going to hurt these guys a lot.

And then, of course, we get to Eli Manning. He was the inspiration for this post three years ago, and not much has changed since.  Manning has a a 5th-place finish from 2011 (+6.1), and three 10th place finishes (2009, 2012, and 2015, each worth 1 point).  He also has a 12th (2014) and a 13th (2010) place finish, but those are his only other top-15 seasons.  Manning performs horribly in the ANY/A Gray Ink test for HOFers.  He’s far behind Blanda, let alone the Stabler/Waterfield floor of quarterbacks.

The graph below shows the data data but with color-coded labels: black for HOF or future HOFer,red for not in, orange for not eligible, and large green dots for the borderline guys.  The running quarterbacks (McNair, McNabb, Cunningham, and Newton) are at the back of the pack with Manning (the second farthest dot to the right); the three “stats stars” are to the left, and then you have Ryan who currently ranks just outside of the top 40 (the X-Axis is rank; the Y-Axis is prorated value).

What do you guys think?