As has been asserted here several times, line play in the NFL is vital. Many people have said for a long time that games are won and lost on the offensive and defensive lines. Well, it's rarely that simple, but it is true that how good your line is will certainly go a long way in predicting how good your team will be.
The Football Know It All has developed a system using three categories to determine the quality of a defensive line. Today, I will lay out the criteria and list the defensive line rankings of each team for 2007. In my next post, I will take into account acquisitions, key losses, aging, and draft picks to give a best estimate for how the defensive lines will perform in 2008.
The new defensive line rating will be called Defensive Line Index (DLI). The three categories that make up DLI are:
1. Sacks- I wish that QB pressures was an official stat, because I believe that it is a fairly more accurate assessment of how much pressure a D-line puts on a quarterback. However, sacks are still a good measure of how much pressure a D-Line exerts on quarterbacks. Furthermore, teams that sack the quarterback a lot give up fewer points and win more games. There is a 0.55 correlation between the number of wins a team has and its sack total. The top five teams in sacks in 2007 all made the playoffs, and the top 2 teams faced off against each other in the Super Bowl.
2. Rushing yards/carry in the first half- Why only look at rushing yards in the first half? Because in the first half, all offenses are trying to score points, whereas in the second half, teams with big leads simply run more and try to eat up time. This leads to lower yards/carry. Thus, teams that are often behind in the second half give up fewer yards/carry in the second half, because the run is more predictably coming. This phenomenon helps them put up better run defense numbers than they deserve. Incidentally, yards/carry allowed in the first half has a -0.31 correlation with wins, whereas overall yards/carry allowed only has a -0.08 correlation. Clearly, First Half Yard/carry is the better index.
3. Percentage of tackles by D-Line- The better the D-Line, the more often it will tackle the ball carrier. Lesser D-Lines miss more tackles and depend more on linebackers and defensive backs to tackle ball carriers. % D-Line Tackles had a 0.23 correlation with winning percentage.
Using these three criteria, I used a regression for each category relative to its relationship to win percentage and adjusted weights in order to account for the differing numerical values of different categories. Below are the results for the 2007 rankings for DLI. The average score was 27.4, with a standard deviation of 7.0:
1. New York Giants- 44.5
2. Seattle Seahawks- 38.9
3. Minnesota Vikings- 37.6
4. Dallas Cowboys- 37.0
5. San Diego Chargers- 35.9
6. Baltimore Ravens- 34.4
7. Tampa Bay Buccaneers- 33.4
8. Green Bay Packers- 32.2
9. Philadelphia Eagles- 31.7
10. New England Patriots- 31.2
11. Chicago Bears- 28.3
12. Tennessee Titans- 28.1
13. Washington- 27.9
14. New Orleans Saints- 27.8
15. Jacksonville Jaguars- 27.4
16. Kansas City Chiefs- 25.8
17. Houston Texans- 25.3
17. San Francisco 49ers- 25.3
19. Detroit Lions- 25.1
20. Pittsburgh Steelers- 24.9
21. Indianapolis Colts- 24.8
22. Arizona Cardinals- 24.7
23. Carolina Panthers- 23.8
24. St. Louis Rams- 23.4
25. Miami Dolphins- 22.8
26. Denver Broncos- 22.2
27. New York Jets- 20.8
28. Buffalo Bills- 20.6
29. Cleveland Browns- 20.1
30. Cincinnati Bengals- 16.8
31. Oakland Raiders- 15.4
32. Atlanta Falcons- 14.4
The DLI, by the way, has a very strong correlation of 0.50 with winning percentage and a correlation of -0.52 with points allowed. These correlations demonstrate that DLI has high predictive validity as a predictor of overall defensive and team performance.
Big Changes Coming
10 years ago
1 comment:
Interesting. I like what you've done but here are a couple suggestions. There are lots of LBs and DBs who get sacks too, so did you separate them? I'd also use sack rate (sacks per attempt) because some teams might have high sack totals not because they have a good pass rush, but because their offense tends to put them well ahead and opponents need to pass more often. Last point is that rushing average is very heavily skewed by a small number of very long runs. The very long runs occur not because of DL failures but because of LB/DB failures. I'd recommend using median rush yds allowed or % of runs greater than X yds. I think FO does something similar for offenses.
Post a Comment