2012 Regression Accuracy Report

Mike Clay compares first and second half touchdown rates to determine the impact of regression.

| 3 years ago
vernon-davis

2012 Regression Accuracy Report


Back in early November, I kicked off this season’s touchdown regression analysis with a breakdown of each qualified quarterback, running back, wide receiver, and tight end’s touchdown rate over the first nine weeks of the regular season. In that piece, I listed some of the highest and lowest touchdown rates at each position. As always, the thought was that many of those noticeably high/low rates would regress towards the mean in the second “half” of the season.

With only three games left on the 2012-13 schedule, it’s as good a time as any to look back and see how we did.

Quarterback

Our original quarterback list included 33 names. Each of the 33 quarterbacks had attempted at least 100 passes during the first nine weeks of the season. Of the 33, we’re eliminating seven quarterbacks that failed to throw at least 50 passes from Week 9 on. Those quarterbacks are Alex Smith, Kevin Kolb, Blaine Gabbert, Michael Vick, Matt Hasselbeck, Matt Cassel, and John Skelton.

That leaves us with a solid sample size of 26 quarterbacks to look at.

  Weeks 1-9
Weeks 10-19
 
Player Att TD TD% Att TD TD% Change
Aaron Rodgers 311 25 8.0% 275 17 6.2% -1.9%
Peyton Manning 280 20 7.1% 321 20 6.2% -0.9%
Drew Brees 312 22 7.1% 306 21 6.9% -0.2%
Josh Freeman 227 16 7.0% 281 11 3.9% -3.1%
Ryan Fitzpatrick 245 15 6.1% 238 9 3.8% -2.3%
Russell Wilson 214 13 6.1% 200 16 8.0% 1.9%
Matt Ryan 286 17 5.9% 341 18 5.3% -0.7%
Ben Roethlisberger 274 16 5.8% 135 10 7.4% 1.6%
Tom Brady 304 16 5.3% 334 21 6.3% 1.0%
Jay Cutler 230 12 5.2% 183 7 3.8% -1.4%
Matt Schaub 230 12 5.2% 359 12 3.3% -1.9%
Andy Dalton 269 14 5.2% 254 13 5.1% -0.1%
Philip Rivers 248 12 4.8% 228 14 6.1% 1.3%

Our first chart shows the 13 quarterbacks who enjoyed a touchdown rate above the 4.6 percent league average over the first nine weeks (which I’m going to call the ‘first half’ going forward). We see here that 9-of-13 (69 percent) quarterbacks regressed, including each of the five highest marks.

Although I’m showing all qualified quarterbacks, the point of this exercise is, and always has been, to examine the regression probability of only the drastically high and low touchdown rates at each position. That being said, I’m going to use a threshold of 6.5 percent here since we see the biggest gap between Josh Freeman (7.0 percent) and Ryan Fitzpatrick (6.1 percent). That leaves us with four quarterbacks (Rodgers, Manning, Brees, Freeman), all who saw their touchdown rates regress in the second half. In my piece at the midway point, Freeman, Fitzpatrick, and Wilson were highlighted. Wilson was an obvious exception to our rule, but the other two showed the largest regression on this chart.

Regression Accuracy: 4-for-4 (100 percent)

  Weeks 1-9 Weeks 10-19
 
Player Att TD TD% Att TD TD% Change
Ryan Tannehill 221 5 2.3% 220 7 3.2% 0.9%
Matthew Stafford 328 8 2.4% 359 12 3.3% 0.9%
Cam Newton 213 6 2.8% 228 13 5.7% 2.9%
Brandon Weeden 302 9 3.0% 161 5 3.1% 0.1%
Andrew Luck 312 10 3.2% 315 13 4.1% 0.9%
Tony Romo 311 10 3.2% 313 18 5.8% 2.5%
Robert Griffin III 246 8 3.3% 141 14 9.9% 6.7%
Sam Bradford 233 8 3.4% 272 13 4.8% 1.3%
Christian Ponder 265 10 3.8% 187 8 4.3% 0.5%
Joe Flacco 260 10 3.8% 293 17 5.8% 2.0%
Eli Manning 302 12 4.0% 203 14 6.9% 2.9%
Mark Sanchez 251 10 4.0% 171 3 1.8% -2.2%
Carson Palmer 315 13 4.1% 220 9 4.1% 0.0%

Our next chart is the opposite of the previous one. It shows the touchdown rates belonging to the 13 quarterbacks below the 4.6 percent league average in the first half of the season. Of the 13, an impressive 11 (84 percent) regressed. Even more impressive, each of the 11-highest touchdown rates from the first half regressed by at least 0.1 percent. The only exceptions were the two quarterbacks nearest to league average, which basically makes them irrelevant to this study anyways. Of the five quarterbacks I highlighted in November, all five regressed in the second half.

For the purpose of our calculations, we’ll count any quarterbacks under 3.5 percent, which is the separation between Sam Bradford (3.4 percent) and Christian Ponder (3.8 percent).

Regression Accuracy: 8-of-8 (100 percent)
Running Total: 12-of-12 (100 percent)

Running Back

Our original examination of running backs included a sample of 46 backs who eclipsed 50 rush attempts in the first half of the season. Of those 46, we’re going to toss out the 10 backs who failed to carry the ball at least 35 times in the second half. They are Daniel Thomas, Isaac Redman, Donald Brown, Jonathan Stewart, Willis McGahee, Andre Brown, Kendall Hunter, Cedric Benson, Maurice Jones-Drew, and Ryan Williams. Unfortunately, we highlighted four of those backs in the original piece, but they simply didn’t see enough of a second-half workload to qualify for an accurate assessment.

Anyways, that leaves 36 backs to be examined.

  Weeks 1-9 Weeks 10-19  
Player Runs TD TD% Runs TD TD% Change
Arian Foster 192 10 5.2% 213 7 3.3% -1.9%
C.J. Spiller 78 4 5.1% 129 2 1.6% -3.6%
Ray Rice 131 6 4.6% 171 4 2.3% -2.2%
Doug Martin 154 7 4.5% 165 4 2.4% -2.1%
DeAngelo Williams 67 3 4.5% 106 2 1.9% -2.6%
LaRod Stephens-Howling 68 3 4.4% 42 1 2.4% -2.0%
Mikel Leshoure 92 4 4.3% 123 5 4.1% -0.3%
Michael Bush 77 3 3.9% 37 2 5.4% 1.5%
Shonn Greene 139 5 3.6% 137 3 2.2% -1.4%
Adrian L. Peterson 168 6 3.6% 202 6 3.0% -0.6%
Felix Jones 58 2 3.4% 53 1 1.9% -1.6%
Frank Gore 120 4 3.3% 162 5 3.1% -0.2%
Stevan Ridley 150 5 3.3% 155 8 5.2% 1.8%
Trent Richardson 152 5 3.3% 115 6 5.2% 1.9%
Reggie Bush 122 4 3.3% 106 2 1.9% -1.4%
Michael Turner 127 4 3.1% 109 6 5.5% 2.4%
Alfred Morris 164 5 3.0% 187 8 4.3% 1.2%
Ahmad Bradshaw 141 4 2.8% 80 2 2.5% -0.3%
Matt Forte 107 3 2.8% 141 2 1.4% -1.4%

Another chart, another success story. The backs with the seven-highest touchdown rates in the first half each regressed in the second half, six of them by a margin of at least 1.9 percent. Going deeper, 11 of the top 12 rates regressed. Overall, 74 percent (14-of-19) of rates above the 2.8 percent league average regressed.

Our cutoff here will be 4.0 percent, which comes between Mikel Leshoure (4.3 percent) and Michael Bush (3.9 percent).

Regression Accuracy: 7-of-7 (100 percent)
Running Total: 19-of-19 (100 percent)

  Weeks 1-9 Weeks 10-19  
Player Runs TD TD% Runs TD TD% Change
Vick Ballard 77 0 0.0% 156 2 1.3% 1.3%
Jonathan Dwyer 58 0 0.0% 98 2 2.0% 2.0%
Alex Green 86 0 0.0% 48 0 0.0% 0.0%
Daryl Richardson 62 0 0.0% 36 0 0.0% 0.0%
Steven Jackson 109 1 0.9% 149 3 2.0% 1.1%
Ryan Mathews 95 1 1.1% 89 0 0.0% -1.1%
DeMarco Murray 75 1 1.3% 86 3 3.5% 2.2%
LeSean McCoy 146 2 1.4% 54 0 0.0% -1.4%
Darren McFadden 139 2 1.4% 77 0 0.0% -1.4%
Jamaal Charles 132 2 1.5% 153 3 2.0% 0.4%
Pierre Thomas 64 1 1.6% 41 0 0.0% -1.6%
Rashad Jennings 63 1 1.6% 38 1 2.6% 1.0%
Fred Jackson 59 1 1.7% 56 2 3.6% 1.9%
Mark Ingram 54 1 1.9% 102 4 3.9% 2.1%
Chris D. Johnson 147 3 2.0% 129 3 2.3% 0.3%
BenJarvus Green-Ellis 142 3 2.1% 147 3 2.0% -0.1%
Marshawn Lynch 185 4 2.2% 166 9 5.4% 3.3%

Here’s where we finally start to hit a snag. Unfortunately, there are always going to be backs that don’t have a ton of breakaway speed, but spend a lot of time getting the ball between the 20s. We see here that, of our 36 backs, only four failed to find paydirt in the first half. Of those four, two finished the season without a touchdown. That’s already two in the ‘miss’ column. Adding fuel to the fire are Ryan Mathews, LeSean McCoy, Darren McFadden, and Pierre Thomas. They combined to score zero second-half touchdowns on the ground.

On the plus side, Vick Ballard and Jonathan Dwyer each scored twice in the second half. Steven Jackson and Demarco Murray both scored three times. Overall, only 59 percent (10-of-17) of rates below the 3.8 percent league average regressed.

Even worse, with a cutoff of 1.2 percent, which is between Mathews (1.1 percent) and Murray (1.3 percent), our qualified accuracy is dismal.

Regression Accuracy: 3-of-6 (50 percent)
Running Total: 22-of-25 (88 percent)

Wide Receiver

Our original study examined 58 wide receivers with 40-plus targets during the first half of the season. Of those 58, nine failed to reach the 30-target mark in the second half. Eliminated from the study are Dwayne Bowe, Dexter McCluster, Michael Jenkins, Jordy Nelson, DeSean Jackson, Kevin Ogletree, Titus Young, Nate Burleson, and Percy Harvin.

That leaves us with strong sample of 49 wideouts.

  Weeks 1-9 Weeks 10-19  
Player Targ TD TD% Targ TD TD% Change
James Jones 60 8 13.3% 43 7 16.3% 2.9%
Randall Cobb 57 6 10.5% 53 2 3.8% -6.8%
Eric Decker 68 7 10.3% 62 6 9.7% -0.6%
A.J. Green 79 8 10.1% 88 3 3.4% -6.7%
Mike A. Williams 51 5 9.8% 69 4 5.8% -4.0%
Josh Gordon 41 4 9.8% 48 1 2.1% -7.7%
Vincent Jackson 63 6 9.5% 74 2 2.7% -6.8%
Torrey Smith 56 5 8.9% 58 5 8.6% -0.3%
Sidney Rice 46 4 8.7% 39 3 7.7% -1.0%
Marques Colston 71 6 8.5% 55 4 7.3% -1.2%
Nate Washington 49 4 8.2% 37 0 0.0% -8.2%
Brandon Marshall 86 7 8.1% 95 4 4.2% -3.9%
Mike Wallace 62 5 8.1% 54 3 5.6% -2.5%
Andre Roberts 63 5 7.9% 44 0 0.0% -7.9%
Julio Jones 64 5 7.8% 73 5 6.8% -1.0%
Victor Cruz 95 7 7.4% 42 3 7.1% -0.2%
Cecil Shorts 43 3 7.0% 58 4 6.9% -0.1%
Denarius Moore 59 4 6.8% 51 3 5.9% -0.9%
Demaryius Thomas 64 4 6.3% 81 7 8.6% 2.4%
Jeremy Maclin 49 3 6.1% 69 4 5.8% -0.3%
Roddy White 66 4 6.1% 82 4 4.9% -1.2%
Miles Austin 67 4 6.0% 48 2 4.2% -1.8%
Michael Crabtree 51 3 5.9% 78 8 10.3% 4.4%
Steve Johnson 69 4 5.8% 75 2 2.7% -3.1%
Kendall Wright 61 3 4.9% 34 1 2.9% -2.0%

And we’re back. Of the 25 wide receivers with a touchdown rate above the 4.9 percent league average, 22 (88 percent) regressed in the second half. That really is an amazing rate. Of the 18 wideouts with a rate above 6.5 percent in the first half, 17 (94 percent) had a lower second-half mark.

Of course, every year, there’s one major exception to the rule. This year, it was James Jones, who somehow managed to score 15 times over 18 games for Green Bay. His 16.3 touchdown rate in the second half was actually greater than his league-leading 13.3 percent first-half mark.

Of the six wideouts we highlighted in November, all but Jones regressed in the second half.

Our cutoff here will be nine percent, which separates Vincent Jackson (9.5 percent) from Torrey Smith (8.9 percent).

Regression Accuracy: 6-of-7 (86 percent)
Running Total: 28-of-32 (88 percent)

  Weeks 1-9 Weeks 10-19  
Player Targ TD TD% Targ TD TD% Change
Davone Bess 60 0 0.0% 36 1 2.8% 2.8%
Josh Morgan 40 0 0.0% 36 2 5.6% 5.6%
Calvin Johnson 81 1 1.2% 118 4 3.4% 2.2%
Steve L. Smith 66 1 1.5% 61 3 4.9% 3.4%
Brian Hartline 63 1 1.6% 55 0 0.0% -1.6%
Donnie Avery 63 1 1.6% 54 2 3.7% 2.1%
Antonio Brown 61 1 1.6% 37 4 10.8% 9.2%
Anquan Boldin 54 1 1.9% 72 4 5.6% 3.7%
Justin Blackmon 52 1 1.9% 77 4 5.2% 3.3%
Kenny Britt 46 1 2.2% 39 3 7.7% 5.5%
Hakeem Nicks 45 1 2.2% 51 2 3.9% 1.7%
Wes Welker 80 2 2.5% 99 4 4.0% 1.5%
Dez Bryant 65 2 3.1% 72 10 13.9% 10.8%
Reggie Wayne 95 3 3.2% 101 2 2.0% -1.2%
Andre Johnson 60 2 3.3% 115 2 1.7% -1.6%
Lance Moore 52 2 3.8% 48 4 8.3% 4.5%
Greg Little 48 2 4.2% 39 2 5.1% 1.0%
Malcom Floyd 48 2 4.2% 34 3 8.8% 4.7%
Jeremy Kerley 47 2 4.3% 41 0 0.0% -4.3%
Danny Amendola 46 2 4.3% 48 1 2.1% -2.3%
Brandon Lloyd 68 3 4.4% 70 2 2.9% -1.6%
Larry Fitzgerald 90 4 4.4% 58 0 0.0% -4.4%
Andrew Hawkins 44 2 4.5% 33 2 6.1% 1.5%
T.Y. Hilton 43 2 4.7% 58 5 8.6% 4.0%

Our second chart is further confirmation that touchdown regression is in full force when it comes to wide receivers. The bottom four and 12 of the lowest 13 touchdown rates from the first half regressed in the second half. Overall, 17-of-24 (71 percent) of rates below the 4.9 percent league average were higher in the second half.

Our primary outlier here is Brian Hartline, who failed to find the endzone after Week 4. He’s the only wideout with a touchdown rate below 3.2 percent in the first half to not score at least once from Week 10 on. Of the six wide receivers we highlighted in November, all but Hartline regressed in the touchdown department.

We’ll go with a cutoff point of 1.7 percent, which is the separation between Antonio Brown (1.6 percent) and Anquan Boldin (1.9 percent).

Regression Accuracy: 6-of-7 (86 percent)
Running Total: 34-of-39 (87 percent)

Tight End

Although our tight end sample is actually lower than what we had to work with at quarterback, there is still something to be learned from the data. Back in November, we started out study with 27 tight ends that had 30-plus targets through nine weeks. Of the 27, three failed to reach the 20-target mark in the second half. The three we’ll eliminate are Coby Fleener, Ron Gronkowski, and Fred Davis.

A sample of 24 tight ends remains.

  Weeks 1-9 Weeks 10-19  
Player Targ TD TD% Targ TD TD% Change
Vernon Davis 32 4 12.5% 29 1 3.4% -9.1%
Heath Miller 50 6 12.0% 43 2 4.7% -7.3%
Kyle Rudolph 45 5 11.1% 44 4 9.1% -2.0%
Scott Chandler 39 4 10.3% 31 2 6.5% -3.8%
Owen Daniels 53 5 9.4% 70 1 1.4% -8.0%
Anthony Fasano 35 3 8.6% 26 2 7.7% -0.9%
Jimmy Graham 60 5 8.3% 71 4 5.6% -2.7%
Antonio Gates 40 3 7.5% 36 4 11.1% 3.6%
Tony Gonzalez 62 4 6.5% 66 5 7.6% 1.1%
Marcedes Lewis 32 2 6.3% 43 2 4.7% -1.6%
Martellus Bennett 49 3 6.1% 39 2 5.1% -1.0%

We see strong results here. Each of the seven-highest tight end touchdown rates during the first half of the season were lower in the second half. All seven rates dropped by at least 0.9 percent, with Vernon Davis, Heath Miller, and Owen Daniels each seeing rates at least seven percent lower.

Overall, nine of the 11 rates above the 5.5 percent league average for tight ends regressed in the second half. The exceptions, Antonio Gates and Tony Gonzalez, were within two percentage points of the league mean.

Our cutoff point is a cool 10 percent, which separates Scott Chandler (10.3 percent) from Owen Daniels (9.4 percent).

Regression Accuracy: 4-of-4 (100 percent)
Running Total: 38-of-43 (88 percent)

  Weeks 1-9 Weeks 10-19  
Player Targ TD TD% Targ TD TD% Change
Tony Scheffler 31 0 0.0% 52 1 1.9% 1.9%
Rob Housler 35 0 0.0% 30 0 0.0% 0.0%
Jason Witten 80 1 1.3% 66 2 3.0% 1.8%
Brandon Pettigrew 57 1 1.8% 38 2 5.3% 3.5%
Brent Celek 55 1 1.8% 28 0 0.0% -1.8%
Greg Olsen 47 1 2.1% 49 4 8.2% 6.0%
Jermichael Finley 45 1 2.2% 48 1 2.1% -0.1%
Jacob Tamme 41 1 2.4% 42 1 2.4% -0.1%
Ben Watson 30 1 3.3% 44 2 4.5% 1.2%
Brandon Myers 50 2 4.0% 51 2 3.9% -0.1%
Jermaine Gresham 50 2 4.0% 49 3 6.1% 2.1%
Dennis Pitta 50 2 4.0% 47 6 12.8% 8.8%
Jared Cook 44 2 4.5% 24 2 8.3% 3.8%

Our final chart takes a look at tight ends with exceptionally low touchdown rates in the first half.

Of the 13 tight ends shown, only eight regressed in the second half (62 percent). Of course, what counts are the names at the top of the list.

Very much like our running back list from earlier, we see at least two situations where a player with an already-low scoring rate failed to score in the second half of the season. Like Daryl Richardson and Alex Green earlier, Rob Housler failed to score at all this season. That’s a miss. Brent Celek scored only once and that occasion came prior to Week 10. Of the six lowest marks from the first nine weeks, four were higher in the second half.

Our final cutoff point of the day will be 1.5 percent, which fits between Jason Witten (1.3 percent) and Brandon Pettigrew (1.8 percent).

Regression Accuracy: 2-of-3 (67 percent)
Running Total: 40-of-46 (87 percent)

Conclusion

I can’t stress it enough that the primary goal here is to focus on the very high and very low rates at each position. That said, we learned today that even rates close to league average tend to regress towards the mean. Consider that of the 68 “high” rates from the first half, 54 (or 79 percent) were lower in the second half. Of the 67 “low” rates, 46 (69 percent) were higher.

That means that, if someone were to give you a player’s touchdown rate at the midway point of the season, you could guess as to whether or not it will be higher or lower in the second half with 75 percent accuracy.

And, as if that isn’t impressive enough, our running accuracy total turned out to be 87 percent. Of the 46 extremely high and low touchdown rates, 40 were lower/higher, respectively, after Week 9.

For the third year in a row, touchdown regression has proven to be as close to a guarantee as you’ll find in football statistics. In the future, before you to decide to buy low or sell high, be sure to check in on each player’s touchdown rate. After all, those six points could be all that separates you from a league title.

Touchdown regression history:

Introduction – November, 2010
Accuracy Checkup – December, 2010
2010 Accuracy, Looking Ahead – January, 2011
WR/TE TD Regression – June, 2011
RB TD Regression – June, 2011
QB TD Regression – July, 2011
2011 Regression Alert – November, 2011
2011 Accuracy – January, 2012
2012 Regression Alert – November, 2012

Check back during the offseason for a breakdown of players you can expect to enjoy/suffer touchdown regression during the 2013 season.

Follow Mike Clay on Twitter: @MikeClayNFL

Comments are closed.