Once again, it's The Boy from Rock M Nation. The amount of time between Parts Two and Three here was larger than I intended, but that's actually a good thing--I got sidetracked because somebody showed me a site where I can enter play-by-play for games about 4x or 5x faster. Instead of 8-10 a week, I can now do 40-50. So without the help of any volunteers, I should have all of 2007's games entered some time in late-April or early-May. Hooray! I still need volunteers, though, when it comes to playing with the data and seeing what becomes of it. As always, e-mail me at BillConnelly1 @ gmail.com if you're interested. Anyway, on to Part Three.
So today we're going to cover Defensive Success Rates, or Alvin Bowen's Law. This one probably won't be quite as interesting as Parts One or Two (I'm a great salesman!), but it's a pretty important concept.
I can basically pinpoint the moment I started down this "football metrics" road. It was spring 2007, and I was reading about Iowa State LB Alvin Bowen and the fact that he had roughly 113,536 tackles in 2007. He received plenty of post-season accolades, but I just kept thinking, "Yeah, but how many of those were actually good plays, and how many were just plays where somebody had to make the tackle? Bad teams always have somebody with an insane amount of tackles. If only there were a way to differentiate between plays and good plays..." Being a big sabermetrics fan (i.e. a nerd), I started looking around for stat nerd-oriented football books/websites. Long story short, I found Football Outsiders, I pilfered their Success Rate measure (covered in Part Two), I entered a crapload of Big 12 play-by-plays from 2006 and 2007, and it was time to play around with the data...in this case, using Alvin Bowen as the guinea pig, of course.
In 2006, Alvin Bowen recorded 155 tackles, a profane 12.92 per game, for an Iowa State team that was, shall we say, not great. How are we going to evaluate how many actual `good plays' he made, and how are we going to define `good plays'? Well let's go back to the Success Rate measure I used in Part Two. If an offense gains a certain number of yards on a certain down, it's a "successful" play. Well, wouldn't a "successful" play for the defense be preventing a "successful" play for the offense? Makes sense to me. So the rules for Defensive Success Rates are as follows (and if they look familiar, they are--they're just the inverse of the Offensive Success Rate rules):
1st Down: less than 50% of necessary first down yardage. If it's 1st-and-10, a gain of less than 5 yards equals defensive 'success'.
2nd Down: less than 70% of necessary yardage (rounded up to the nearest yard, of course). If it's 2nd-and-10, a gain of less than 7 yards is 'success'. 2nd-and-15? Less than 10 yards.
3rd and 4th Downs: less than 100% of necessary yardage. I figure this requires no explanation.
In all, any given play has about a 57% chance of defensive `success.'
During conference play in 2006, here are the top 10 Big 12 LBs according to Defensive Success Rate (for LBs who had at least 10 `successful' plays):
- Rufus Alexander (OU), 77.5%
- Rodrick Johnson (OSU), 74.6%
- Brent Slaughter (Tech), 74.1%
- Joe Mortensen (KU), 68.6%
- Jordan Dizon (CU), 67.7%
- Corey McKeon (NU), 67.2%
- Dedrick Harrington (MU), 65.0%
- Robert Killebrew (UT), 63.9%
- Bo Ruud (NU), 63.8%
- Rashad Bobino (UT), 63.2%
1st-Team: Alexander (#1 above), Ruud (#9), MU's Marcus Bacon (#18)
2nd-Team: KSU's Brandon Archer (#12), Alvin Bowen (#13), BU's Joe Pawelek (#32)
So obviously the coaches disagree with using Defensive Success Rate as a tool to evaluate effectiveness. Then again, so do I. Well, I disagree with using only Defensive Success Rate.
Brent Slaughter, #3 on my list above, had 13.5 tackles (to calculate this, I only give half-credit for assisted tackles) in conference play, while Alvin Bowen had 69.5. Slaughter making 10.0 successful tackles among his 13.5 doesn't make him a better linebacker than Bowen. LBs should, to some degree, still get credit for quantity along with quality. So what were the top 10 LBs according to Total Successful Tackles in conference play?
- Alvin Bowen, 42.5 (2nd-team)
- Jordan Dizon, 42.0 (HM...which was a total travesty, not to the level of Martin Rucker being 2nd-team All-Conference TE behind OSU's Brandon Pettigrew in '07, but still a travesty...Big 12 coaches are horrible voters)
- Rufus Alexander, 39.5 (1st)
- Tyron McKenzie (ISU), 35.0
- Dedrick Harrington, 32.5 (HM)
- Brandon Archer, 27.5 (2nd)
- Rodrick Johnson, 26.5
- Brock Stratton (Tech), 25.5 (HM)
- Patrick Lavine (OSU), 24.5
- Rashad Bobino, 24.0
It seems to me that if we were to add up the rankings of three separate categories (Defensive Success Rate, Total Successful Tackles, Total Tackles), we might get a pretty good indicator, no? Sure, there's some overlap in there (success rate is just Successful Tackles divided by Tackles), but still...let's see what happens.
Top 10 Big 12 LBs, 2006
- Rufus Alexander (sum: 8--#1 in Success Rate, #3 in Successful Tackles, #4 in Total Tackles...the lower the better)
- Jordan Dizon (9)
- Alvin Bowen (15)
- Dedrick Harrington (17)
- Tyrone McKenzie (21)
- Brandon Archer (26)
- Rodrick Johnson (28)
- Rashad Bobino (35)
- Brock Stratton (35)
- Patrick Lavine (39)
(You can tell by the Coaches' list that voters tend to only remember the big plays--hence Bacon and Ruud making 1st-team while Dizon, who was absolutely everywhere for four years, was relegated to Honorable Mention.)
You'll notice that Alvin Bowen was #3 on my LBs list. He is officially exonerated. In fact, that suggests that he was underrated by the coaches. I will therefore be renaming this Joe Pawelek's Law. Pawelek, you see, made it to 2nd-team All-Big 12 (and Freshman All-American) solely on the basis of his high number of tackles. But his Defensive Success Rate was only 42.0%, which is absolutely horrid for a linebacker and led to his being ranked 26th (out of 33) on my LBs list. A majority of his tackles were of the "Somebody had to make it..." variety, but he got all sorts of attention for it (as did 2006 Defensive Newcomer of the Year Misi Tupe, who ranked 32nd).
Really, this same concept could be used to judge Defensive Linemen as well (Top 5 Big 12 DLs in '06: OU's CJ Ah You, NU's Adam Carriker, MU's Lorenzo Williams, Tech's Keyunta Dawson, and CU's Abraham Wright), and even Safeties. (Cornerbacks are an entirely different ballgame--some of the best CBs hardly make any tackles at all, as the guys they're covering rarely catch the ball.)
Using concepts like Success Rates and Successful Tackles gives you a much better idea of which defensive players (at certain positions) actually excelled at making plays others didn't make, and which ones just made plays that almost anybody could have made. It gives guys like Jordan Dizon and Tyrone McKenzie a little more credit than they otherwise received (in '06, anyway) and puts the Joe Paweleks of the world in their place.
(Then again, in '07 Bowen had a 40.0% success rate and Joe Pawelek a 54.2% rate. So I'd say there's still room to grow here when it comes to predictive ability.)
The best part about all this, though, is that these measures are still in the formative stage. Anybody who thinks they see a better way of doing this can just e-mail me (or comment below), and we'll where it goes. This is one giant open book right now.
Next up (at some point): Part IV, Strength of Schedule...