Writers of Pro Football Prospectus 2008

Most Recent FO Features

MurrayLat13.jpg

» Scramble for the Ball: With All the Fixings

An idiot's (two idiots'?) guide to Thanksgiving football, prepped and primed for the monsters-in-law who only watch these three games in a year.

13 Oct 2016

2016 ALEX: The Early Results

by Scott Kacsmar

Have you grown tired of watching a quarterback throw short of the sticks on third down? Are you concerned that your offense is doing this too often? This is why we created the ALEX stat last year. ALEX stands for Air Less EXpected, and measures the average difference between how far a quarterback threw a pass (air yards) and how many yards he needed for a first down. If a quarterback throws a pass to a receiver who is 5 yards behind the line of scrimmage on third-and-15, then that would be minus-20 ALEX. A pass thrown beyond the first-down marker would earn a positive ALEX figure. Most quarterbacks are wise enough to attack the sticks on fourth down, which is why ALEX is best applied on third downs to gauge the aggressiveness of a quarterback. As you will see, we have identified some quarterbacks with a strong year-to-year tendency for aggressive or ultra-conservative play.

Football Outsiders now features ALEX data on our quarterbacks page, which can be seen at any time. This is the first formal study of ALEX we have done for the 2016 season, and as always, the numbers are subject to change once corrections are made in the NFL's play-by-play, as well as the removal of certain incompletions such as passes thrown away or batted at the line. There may be some minor differences in the tables below compared to the above link due to these factors, as well as the exclusion of defensive pass interference plays. It's not that those penalty plays are not important examples of decision making, but our data on ALEX for 2006-2015 does not include penalties, so we are ignoring them in contemporary data to make valid historical comparisons.

We actually just had the highest and lowest ALEX plays of 2016 on Sunday in Week 5. Late in the first quarter in Denver, Matt Ryan tried to set up a bubble screen to Tevin Coleman on third-and-31, but the pass was not caught. That was minus-36 ALEX. Meanwhile in Cleveland, Tom Brady faced a third-and-6 before uncorking a 52-yard bomb in the air for Chris Hogan, who made the plus-46 ALEX catch.

So how do the numbers shake out after five weeks? The following table looks at every quarterback this season with at least 20 third-down passes. I wanted to use 30 attempts as the qualifier, but lowered the bar to make sure Trevor Siemian and Cody Kessler were included. The quarterbacks are ranked from highest to lowest ALEX. Also included are their conversion rate, DVOA, and average yards to go for a first down. Finally, Short% is the percentage of third-down attempts that were thrown short of the sticks (negative ALEX). ALEX and Short% generally have strong correlation (near minus-0.85), but Short% is a good way to account for the outliers like that minus-36 play for Ryan.

Rk Quarterback Team ALEX CONV% Rk DVOA Rk Passes Avg. Need Rk Short% Rk
1 Ben Roethlisberger PIT 6.8 52.0% 2 92.8% 1 50 5.9 31 22.0% 2
2 Aaron Rodgers GB 4.5 44.2% 11 26.0% 15 43 7.4 20 20.9% 1
3 Tyrod Taylor BUF 4.4 35.9% 25 42.1% 11 39 8.3 9 35.9% 7
4 Cam Newton CAR 3.1 43.9% 12 47.6% 8 41 8.3 11 29.3% 3
5 Brock Osweiler HOU 2.7 39.2% 16 3.7% 24 51 6.6 30 35.3% 6
6 Andrew Luck IND 2.2 44.7% 10 2.0% 26 47 7.9 16 34.0% 4
7 Eli Manning NYG 1.9 37.5% 18 31.6% 14 48 7.7 17 39.6% 11
8 Jameis Winston TB 1.7 40.4% 15 10.6% 22 52 8.1 12 34.6% 5
9 Ryan Fitzpatrick NYJ 1.4 36.7% 22 3.0% 25 49 7.3 21 49.0% 24
10 Kirk Cousins WAS 1.3 40.4% 14 -21.6% 29 47 6.6 29 36.2% 8
11 Blake Bortles JAC 1.1 26.2% 29 -61.6% 31 42 8.3 8 42.9% 14
12 Drew Brees NO 1.1 47.5% 6 43.5% 10 40 7.1 26 45.0% 18
13 Russell Wilson SEA 0.9 36.4% 23 13.1% 20 33 8.7 4 45.5% 19
14 Joe Flacco BAL 0.9 38.7% 17 18.0% 18 62 7.3 23 43.5% 16
15 Carson Palmer ARI 0.6 43.2% 13 48.4% 7 37 9.1 3 37.8% 10
Rk Quarterback Team ALEX CONV% Rk DVOA Rk Passes Avg. Need Rk Short% Rk
16 Alex Smith KC 0.5 37.5% 19 10.8% 21 40 7.2 25 40.0% 12
17 Philip Rivers SD 0.5 45.5% 8 53.0% 5 44 6.9 27 45.5% 20
18 Trevor Siemian DEN 0.4 47.8% 5 44.3% 9 23 8.3 10 52.2% 26
19 Matt Ryan ATL 0.4 45.0% 9 73.7% 2 40 8.1 13 42.5% 13
20 Brian Hoyer CHI 0.2 33.3% 26 7.4% 23 30 8.0 14 36.7% 9
21 Marcus Mariota TEN 0.2 37.0% 21 41.1% 12 46 8.7 5 43.5% 15
22 Case Keenum LARM -0.2 33.3% 27 15.8% 19 48 6.7 28 54.2% 29
23 Ryan Tannehill MIA -0.4 31.7% 28 -6.1% 28 41 7.3 22 46.3% 21
24 Dak Prescott DAL -0.5 52.3% 1 71.4% 3 44 7.6 19 47.7% 22
25 Derek Carr OAK -0.7 36.0% 24 21.4% 17 50 8.7 6 50.0% 25
26 Matthew Stafford DET -0.8 48.9% 3 24.6% 16 45 8.4 7 44.4% 17
27 Sam Bradford MIN -0.9 46.3% 7 67.0% 4 41 7.2 24 53.7% 28
28 Carson Wentz PHI -1.0 37.5% 20 49.8% 6 32 9.4 2 53.1% 27
29 Blaine Gabbert SF -1.2 25.5% 31 -58.9% 30 47 7.6 18 48.9% 23
30 Andy Dalton CIN -2.1 26.0% 30 -5.1% 27 50 9.6 1 60.0% 30
31 Cody Kessler CLE -3.3 48.0% 4 34.4% 13 25 8.0 15 60.0% 31

While we are only five weeks in, I immediately noticed some similarities to last year's results. Ben Roethlisberger is leading all quarterbacks in ALEX with the same plus-6.8 figure, which again would be the second-highest season since 2006, trailing only 2006 Michael Vick (plus-6.97). He is doing an even better job of hitting his passes this year. Roethlisberger's third-down passing DVOA and conversion rate are both in the top two, and he also has the shortest average need yards for the second year in a row. This season's ALEX top four of Roethlisberger, Aaron Rodgers, Tyrod Taylor, and Cam Newton all ranked in the top six last season. Andrew Luck aims to be in the top 10 in ALEX for the fifth year in a row.

Rodgers in particular has been as consistently high in ALEX as any quarterback in his era. He finished fourth or higher in every season from 2008 through 2014, and then fell a smidge to sixth in 2015. It is not surprising to see him at No. 2 so far this year, and the conversion rate is not as poor as last season when he ranked 29th. While the efficiency is not what it used to be, we can at least say Rodgers has good intentions of doing what he usually does on crucial downs. Rodgers' Short% of 20.9 percent this year would be the second-lowest season by any quarterback since 2006.

For as consistent as Roethlisberger and Rodgers are with a high ALEX, the duo of Alex Smith and Blaine Gabbert tend to be the same at the other end. Last year, Smith brought up the rear with the lowest ALEX season on record at minus-3.4. While he is struggling overall as a quarterback this season, Smith is actually just average in ALEX, which has not been the case since 2007. From 2009 to 2015, Smith ranked 30th or lower with a negative ALEX in each season. However, Gabbert has been up to his usual tricks on third down, as the change to Chip Kelly's offense did not stop him from having the third-lowest ALEX and the worst conversion rate of any quarterback this season. Gabbert has rightfully been benched for Colin Kaepernick in San Francisco.

Matthew Stafford and Sam Bradford are in the bottom six in ALEX for the second year in a row. However, life without Calvin Johnson and a move to Minnesota have led to better conversion rates for their offenses so far. But generally, the negative-ALEX approach struggles to maintain a good conversion rate over the course of the season. Andy Dalton is usually a top-16 ALEX quarterback, but his third-down struggles to get the ball down the field have hurt the offense this year.

A few quarterbacks are seeing a big difference between where they rank in DVOA versus conversion rate on third down. Andrew Luck is 10th in conversion rate, but only 26th in DVOA. His need yards are about average, though the Colts have been in third-and-short a league-high 23 times this season. Kirk Cousins (14th in conversion rate, 29th in DVOA) has had the third-shortest distance to go. On the other end, Tyrod Taylor is 25th in conversion rate, but 11th in DVOA. Much like last year, Taylor is utilizing a high-ALEX strategy, but not converting enough plays overall on third down. The loss of Sammy Watkins has not helped this season. Carson Wentz is 20th in conversion rate, but sixth in DVOA. He has faced the second-longest third downs, and the Eagles are the only offense to not throw a pass on third-and-short this season. Every other team has at least three such passes, but the Eagles have only had an NFL-low five plays on third-and-short. They would have loved a different call (or outcome) on Sunday's third-and-2 toss to Ryan Mathews in Detroit.

The distance to go is certainly a factor in how successful, and perhaps how aggressive quarterbacks are going to be on third down. Just like last year, I have provided a "heat map" to show how the quarterbacks stack up in the different third-down splits: short (1 or 2 yards to go), medium (3 to 6 yards) and long (7-plus yards). The colors indicate where a player is well above average (darker green) versus below average (darker red) based on standard deviations. Those conversion rates are also shown with a ranking.

Again, this is very early, and the sample sizes are small for these splits. I can only make some casual observations here, or point out how much Gabbert stinks, because anything else would be premature at this point. Trevor Siemian's third-and-medium numbers look insane, until you realize it is only three plays. Fine, they still look crazy regardless, but that is definitely going to change.

Last season, Roethlisberger blew everyone away with a plus-12.8 ALEX on third-and-short, but he has actually been below average so far this season in that area. Bradford has done an excellent job on third-and-medium, but third-and-long has been a struggle when he has to hold the ball a little longer behind that Minnesota offensive line. Brian Hoyer has a similar profile in his surprisingly good start with the Bears this year. Blake Bortles has been all around poor, but he has especially struggled to convert on third-and-long so far.

Let's conclude with the rookies, because there are certainly some eye-opening numbers here from this season's new faces. We talked a little about Wentz already, and I think the potential for him to remain a minus-ALEX quarterback is high because of what that offensive system asks of him. Dak Prescott in Dallas is not taking many risks, but he is converting at a higher rate on third down than any other passer this season. That is an interesting one to monitor, and another reason to justify keeping him as the starter over a healthy Tony Romo if you are in that camp.

Finally, there is Cody Kessler in Cleveland. The third-round pick is throwing the shortest passes of any quarterback in the league this season, so it is no surprise to see him dead last in ALEX too. However, the fact that he is fourth in conversion rate and 7-of-14 on third-and-long is amusing to say the least. This reminds me of the unexpected third-down success the Browns had last season with Josh McCown and Johnny Manziel, despite being a rather poor offense overall. I doubt Kessler's conversion rate will stay that high, but we'll continue to track that situation.

The question is, should we expect any change in the low ALEX and the generally low air yards from Kessler? I gathered the 31 qualified rookie seasons since 2006 to see where they stacked up in third-down ALEX with this year's trio. The following table orders these quarterbacks by descending ALEX, but all of the numerical rankings are for that particular season.

Rookie Quarterbacks: ALEX on Third Down (Min. 50 Attempts)
Quarterback Team Year Passes CONV% Rk Short% Rk ALEX Rk
Matt McGloin OAK 2013 61 32.8% 37 34.4% 7 4.1 1
Andrew Luck IND 2012 155 47.1% 12 30.3% 3 3.7 3
Vince Young TEN 2006 101 30.7% 37 30.7% 5 2.7 9
Russell Wilson SEA 2012 113 47.8% 11 38.9% 16 2.2 12
Matt Ryan ATL 2008 143 46.2% 10 37.8% 14 2.0 11
Josh Freeman TB 2009 86 33.7% 30 37.2% 13 1.9 11
Geno Smith NYJ 2013 132 43.2% 16 43.2% 31 1.7 18
Andy Dalton CIN 2011 139 39.6% 23 41.0% 20 1.5 15
Jameis Winston TB 2015 143 42.7% 22 30.8% 4 1.2 21
Marcus Mariota TEN 2015 95 43.2% 20 35.8% 13 1.1 23
Colt McCoy CLE 2010 61 44.3% 16 42.6% 21 1.1 19
Derek Carr OAK 2014 168 40.5% 30 40.5% 19 1.1 23
EJ Manuel BUF 2013 92 31.5% 38 41.3% 24 0.9 28
Joe Flacco BAL 2008 132 40.9% 21 44.7% 24 0.9 23
Mark Sanchez NYJ 2009 130 37.7% 23 36.9% 12 0.9 21
Cam Newton CAR 2011 108 38.0% 26 46.3% 28 0.9 21
Matt Leinart ARI 2006 94 41.5% 20 51.1% 35 0.4 30
Trent Edwards BUF 2007 78 32.1% 36 48.7% 33 0.4 26
Quarterback Team Year Passes CONV% Rk Short% Rk ALEX Rk
Bruce Gradkowski TB 2006 87 33.3% 36 50.6% 34 0.4 31
Nick Foles PHI 2012 64 34.4% 33 45.3% 27 0.3 25
Brandon Weeden CLE 2012 147 36.1% 31 54.4% 35 0.2 26
Sam Bradford STL 2010 157 41.4% 22 46.5% 27 0.2 27
Matthew Stafford DET 2009 106 41.5% 20 41.5% 23 0.1 27
Ryan Tannehill MIA 2012 135 40.0% 23 43.7% 24 0.0 29
Robert Griffin WAS 2012 99 35.4% 32 47.5% 29 -0.1 30
Christian Ponder MIN 2011 80 46.3% 10 50.0% 30 -0.3 29
Dak Prescott DAL 2016 44 52.3% 1 47.7% 22 -0.5 24
Mike Glennon TB 2013 113 37.2% 33 45.1% 33 -0.6 37
Teddy Bridgewater MIN 2014 100 46.0% 16 52.0% 31 -0.6 32
Blake Bortles JAC 2014 117 35.9% 33 47.9% 27 -0.7 33
Carson Wentz PHI 2016 32 37.5% 20 53.1% 27 -1.0 28
Blaine Gabbert JAC 2011 121 36.4% 30 52.1% 32 -1.7 34
Jimmy Clausen CAR 2010 92 32.6% 30 47.8% 30 -2.2 32
Cody Kessler CLE 2016 25 48.0% 4 60.0% 31 -3.3 31
AVG - - 104 39.6% 23.6 44.0% 23.1 0.6 23.2

This year's trio is on the low end in ALEX, but only seven passers in this table ranked in the top 16 in their given season. On average, these rookies were all around 23rd for the season in conversion rate, Short%, and ALEX. Personally, I do not think most of the quarterbacks that have entered the NFL since 2006 have been good players. Andrew Luck and Russell Wilson have Hall of Fame potential, and Matt Ryan and Joe Flacco have been dependable long-term starters, but the rest leave a lot to be desired.

The fact that Gabbert and Jimmy Clausen, arguably the two worst players in this table, are sandwiched between Wentz and Kessler is not very encouraging for their outlooks. We are just going into Week 6, but be sure to keep up with our ALEX updates this year to see how things change.

Posted by: Scott Kacsmar on 13 Oct 2016

7 comments, Last at 15 Oct 2016, 1:04am by DRohan

Comments

1
by Raiderjoe :: Fri, 10/14/2016 - 12:58pm

I dlook at low ALEX guys as being wussies (and/or having dopey head coaces/offensive coordinaotrs)) coempared to most average quarterback. is this rightg way of looking at it?

2
by Tomlin_Is_Infallible :: Fri, 10/14/2016 - 1:47pm

"25 Derek Carr OAK -0.7"

heheheheh

--------------------------------------
The standard is the standard!

4
by Raiderjoe :: Fri, 10/14/2016 - 3:11pm

oh, didn't even llok at whole list. yes, sometimes Carr and raiders a little fascinated with the dinka nd dunk.

3
by Jerry P. :: Fri, 10/14/2016 - 2:57pm

I think it really depends on how these values are all calculated. I am assuming an incomplete pass on 3rd and 5 is -5 ALEX so you get wrecked if your receivers have drops on 3rd and long but maybe it doesn't work that way.

My favorite play in football, the bubble screen, would also heavily penalize the QB who is ordered to throw them. There seems like there are a lot of scenarios where the QB gets punished by play-calling.

CORRELATIONS

ALEX to CONV% = .2352
ALEX to DVOA = .2133
CONV% to AVG. NEEDED = -.2789
ALEX to AVG. NEEDED = -.4125

So the strongest connection is between having less average yards to go and having a good ALEX. Makes sense to me.

Does this mean the key to having a good ALEX is run the ball on early downs since it is ALEX neutral in the downs where passes are less likely to be beyond the first down marker? Then pass on hopefully 3rd and short and always have positive yards.

So...

1st and 10 - 3 yard run - 0 ALEX
2nd and 7 - 3 yard run - 0 ALEX
3rd and 4 - 5 (air) yard pass - +1 ALEX
1 first down, 11 yards gained and +1 ALEX

1st and 10 - 3 yard bubble screen! - -10 ALEX! (pass thrown to the LOS)
2nd and 7 - 3 yard bubble screen! - -7 ALEX (pass thrown to the LOS)
3rd and 4 - 5 (air) yard pass - +1 ALEX
1 first down, 11 yards gained and -16 ALEX

Is this the correct calculation or am I screwing this up bad?

5
by Vincent Verhei :: Fri, 10/14/2016 - 4:59pm

I'm sorry, but you're making many mistakes here.

I am assuming an incomplete pass on 3rd and 5 is -5 ALEX so you get wrecked if your receivers have drops on 3rd and long but maybe it doesn't work that way.

False. It depends on how far the pass travels, not whether the ball is caught or not. If it's third-and-5 and you throw to a receiver at the line of scrimmage, that's minus-5 ALEX. If he's 50 yards downfield, that's plus-45 ALEX. All this is measuring is the length of the throw, nothing else.

My favorite play in football, the bubble screen, would also heavily penalize the QB who is ordered to throw them. There seems like there are a lot of scenarios where the QB gets punished by play-calling.

This is correct. But there's lots of room for interpretation here. Perhaps the coach is calling for a bubble screen because he doesn't trust the quarterback to drop back and read through progressions and find the open man.

If I were in charge, in a perfect world, we would break plays down into three categories: runs, passes, and screens. But I'm not in charge, and the world isn't perfect, and so screens get lumps in with all passes.

1st and 10 - 3 yard run - 0 ALEX
2nd and 7 - 3 yard run - 0 ALEX
3rd and 4 - 5 (air) yard pass - +1 ALEX
1 first down, 11 yards gained and +1 ALEX
1st and 10 - 3 yard bubble screen! - -10 ALEX! (pass thrown to the LOS)
2nd and 7 - 3 yard bubble screen! - -7 ALEX (pass thrown to the LOS)
3rd and 4 - 5 (air) yard pass - +1 ALEX
1 first down, 11 yards gained and -16 ALEX
Is this the correct calculation or am I screwing this up bad?

Well, first of all, running plays aren't "zero" ALEX. They do not effect ALEX at all. If you throw a 40-yard bomb and then a fullback dive, the team's ALEX will be 40.0, not 20.0. In the six plays you've listed here, the average ALEX is minus-3.8 (minus-15 total air yards divided by four passes), not minus-2.5 (minus-15 total air yards divided by six plays).

For the passes, you've got it right, but there is a reason that in this column, we only look at ALEX on third downs.

7
by DRohan :: Sat, 10/15/2016 - 1:04am

ALEX only measures 3rd down passes, right? So I would think the above sequence would average to +1.0.

6
by Scott Kacsmar :: Fri, 10/14/2016 - 5:22pm

Speaking of screens, the Texans actually ran one (minus-21 ALEX) on fourth-and-16 on Sunday, down by 18 points. It gained 4 yards. Since 2006, only 14 fourth-down passes had a minus-20 ALEX (or lower), and only two plays were thrown more than 3 yards behind the LOS. Joe Flacco and Jason Campbell threw those two passes when their teams trailed by 3 points in the fourth quarter.

I can't even begin to describe what Bill O'Brien had in mind here.