One of the most important factors to winning games is pressuring the opposing quarterback and totaling up sacks. Ethan Young has developed a formula, drawing on data from 2009 to 2015, for accurately predicting future sacks totals, to see which players are expected to increase their sack totals, and which players we can expect to see a decrease in sack totals.
Few positions are judged on a single statistic like EDGE rushers are with sacks. But there are some problems with that thought process when we start digging into it. Because they are so few and far between (the top-32 pass rushers in sacks per game average about 0.75), sacks are an extremely small sample from which to draw statistical results. And with such small sample sizes, factors like luck and randomness can significantly affect outputs.
By using pressure data as a sack peripheral, we not only draw from a sample that is about three times as large, but we also can use those results to project future sack performance by transforming the data into a stat called Expected Sacks (XS).
You may think that the best indicator of future sack production would be past sack production, but XS projections have actually been more accurate at predicting future sacks than the previous season sack totals have for each of the last six seasons, which is as far back as the data goes.
| Projection Method | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | Average |
| XS Correlation | 0.56540 | 0.53658 | 0.61227 | 0.62401 | 0.66881 | 0.56242 | 0.59491 |
| Past Sack Correlation | 0.49200 | 0.50975 | 0.58929 | 0.57449 | 0.66579 | 0.51946 | 0.55846 |
So, XS has been better at projecting future sack totals than actual sacks have over an extended stretch. While that inherently has value, applying these concepts to the outliers with huge differences between their XS and sack totals uncovers detailed results.
Let’s look at the outliers year by year. Keep in mind that XS doesn’t factor in player development or regression, so results that look off usually stem from that. Players can also occasionally overperform in consecutive seasons, so we will see a few examples of such instances.
[dt_divider style=”thick” /]
2009
2009 featured only one player in the positive outlier zone. But the Giants’ Mathias Kiwanuka didn’t get a chance to meet his projection, only playing three games in 2010 because of a neck injury. The four sacks in less than one-fifth of a season was certainly promising, but we will never know how it would have ended up otherwise.
Buffalo’s Aaron Schobel and Cincinnati’s Antwan Odom also didn’t play that season. I’m happy with most of the rest of the results, although Minnesota’s Jared Allen is a little off. But we can talk about his results as we move on to future years.
[dt_divider style=”thick” /]
2010
Interestingly enough, Allen just missed this list. He also had a positive change, leaving him still due for some regression. Tennessee’s Jason Babin and Dallas’ DeMarcus Ware really defied their odds as well, and the 2011 results validate that.
[dt_divider style=”thick” /]
2011
Interestingly enough, Allen just missed this list. He also had a positive change, leaving him still due for some regression. Tennessee’s Jason Babin and Dallas’ DeMarcus Ware really defied their odds as well, and the 2011 results validate that.
There were no players in the positive outlier tier in 2011, but the closest three were:
Boom! Allen, Ware, and Babin finally regressed. Denver’s Von Miller really blossomed in 2012, but he and San Francisco’s Aldon Smith reappear on next season’s list as well. Dumervil proved to be a miss, as he didn’t appear on the next list after pulling in a -2 / -1.5. That isn’t a terrible result, but not enough considering.
[dt_divider style=”thick” /]
2012
Man, what could have been for San Diego’s Melvin Ingram. A torn ACL ended his season and a chance at fulfilling a lofty projection.
That J.J. Watt result is awesome, given the magnitude and the match. Miller was on pace for just under 9.0 sacks, while Smith was on pace for just under 12.5, before they both get hurt. Obviously we didn’t get a full sample with either one, but I’m happy with how those projections faired. Carolina’s Greg Hardy was pretty frustrating, and you’ll see what I mean in the 2013 results.
[dt_divider style=”thick” /]
2013
Minnesota’s Brian Robison is a pretty weird outlier, and the only explanation I can think of is a significant decline in ability this is supported by his raw pressure totals, which fell 56.6% in just one year. And while Detroit’s Willie Young certainly blossomed in a move to Chicago, he actually found himself on the other side of the list in 2014, which makes sense given his overshoot.
Unfortunately, neither Hardy nor Indianapolis’ Robert Mathis had a chance to test their XS projection. Hardy is particularly frustrating, considering the results of his previous XS projection. As for the Bills duo, it took an extra year for their regression. We don’t see Jerry Hughes again, but he fell to 5.0 sacks the following season. And Mario Williams showed up again on the next list.
[dt_divider style=”thick” /]2014
Clearly, Oakland’s Khalil Mack really developed during his second season, allowing him to overshoot his projection by a considerable margin.
2014 was a great season for XS. First off, we see Chicago’s Willie Young and Buffalo’s Williams appear again. While their results weren’t surprising, Williams really seemed to decline in 2015. Kansas City’s Houston was on pace for about 11.5 sacks if he had played 16 games. New York Giants star Jason Pierre-Paul only played half a season, which is part of why he was so far under his projection. Even though Pierre-Paul’s peripherals pointed to this regression, he was actually more efficient at creating pressure in 2015 than he’d been throughout his entire career. He’s someone to watch this season, and that shows up in our new results.
[dt_divider style=”thick” /]
2015
This season, XS projects Cleveland’s Paul Kruger and Pierre-Paul to have sizable increases in their sack totals. Remember, XS uses inputs from the previous season, and Pierre-Paul only played eight games last year after injuring his hand in a fireworks mishap. If he stays on the field for all 16 games, he could very well double his 2016 projection.
Miami’s Cameron Wake should be very interesting to watch. As is the case with Pierre-Paul, his projection is affected by the fact that he only played seven games last season. If he can stay healthy, Wake will easily surpass his projection due to the additional volume he’ll gain. But there is no way he will maintain the one sack per game pace he was on last year.
Some other interesting results:
I’m excited to see what Atlanta’s Vic Beasley Jr. and Arizona’s Markus Golden can do in their second NFL seasons. With some sophomore development, they should absolutely crush last year’s sack totals.
It’s nearly impossible to exactly predict future totals in this stat category, but XS serves as a useful peripheral for identifying outliers that are due for major sack progression or regression. Based on the data available, XS correctly predicted the direction of future sack performance 89.16% of the time, with that percentage jumping to 95.18% when adjusting with immediate correction the following year. While it’s certainly not perfect or revolutionary, XS serves as a very effective sack peripheral statistic and projection model.
Hmm. A 0.59 correlation is not meaningfully different than a 0.56 correlation. What formula are you using, btw?