In researching my weekly post at the Post, I took a look at team tackles by position. In the off-season, I created a nifty stat called Tackle Factor (TF--available at the individual stats pages) that attempts to make sense of tackle statistics. Conventional tackle stats can be very misleading. Porous defenses allow more plays than stout defenses, creating additional tackle opportunities for below-average defenders. Additionally, some positions should be expected to have more tackles than others.
I wanted to see how the Redskins' defensive lines were performing. I figured that the larger the share of a team's tackles that a line accounts for, the better it's probably playing. I created the list below, which shows the percentage of tackles by each position group for each team.
Obviously, there is going to be some expected differences between defenses that are primarily 3-4 schemes and primarily 4-3 schemes. This is due to the relative number of defenders at each position and each position's role in the respective schemes. Additionally, the numbers can be skewed by late-game strategies of either running out the clock or desperation passing. Still, we might learn something by looking at where the tackles are being made.
Team | DL | LB | CB | S |
ARZ | 20% | 38% | 21% | 21% |
ATL | 21% | 40% | 22% | 17% |
BLT | 23% | 32% | 22% | 23% |
BUF | 25% | 35% | 18% | 23% |
CAR | 26% | 37% | 19% | 17% |
CHI | 19% | 36% | 28% | 18% |
CIN | 15% | 39% | 20% | 26% |
CLV | 23% | 32% | 24% | 21% |
DAL | 15% | 45% | 23% | 17% |
DEN | 18% | 40% | 25% | 17% |
DET | 28% | 25% | 23% | 24% |
GB | 14% | 42% | 26% | 18% |
HST | 19% | 34% | 20% | 26% |
IND | 24% | 28% | 30% | 18% |
JAX | 17% | 28% | 31% | 24% |
KC | 21% | 39% | 24% | 16% |
MIA | 16% | 32% | 27% | 25% |
MIN | 23% | 35% | 25% | 17% |
NE | 16% | 40% | 22% | 22% |
NO | 23% | 31% | 22% | 25% |
NYG | 31% | 20% | 21% | 27% |
NYJ | 20% | 33% | 23% | 25% |
OAK | 31% | 23% | 14% | 32% |
PHI | 33% | 27% | 17% | 24% |
PIT | 9% | 51% | 23% | 18% |
SD | 16% | 39% | 19% | 26% |
SEA | 30% | 22% | 23% | 25% |
SF | 24% | 36% | 21% | 20% |
SL | 26% | 26% | 22% | 26% |
TB | 21% | 33% | 23% | 23% |
TEN | 25% | 23% | 19% | 32% |
WAS | 13% | 35% | 25% | 28% |
Average | 22% | 34% | 22% | 23% |
Brian, Have you adjusted your Tackle factor for run/pass ratio?
This is an interesting first cut. PIT sticks out like a sore thumb for overworking their LBs and letting their DL block-but-not-tackle. The Ravens, who run a very similar look, are almost exactly average because of guys like Haloti Ngata who block and tackle. The Patriots (also regarded as excellent) have a mix like the Steelers', and the Jets' mix looks like the Ravens' -- if that surprises you, Google "Rex Ryan".
I think the most telling numbers here are that lots of different teams -- all with good defenses -- get their Safeties' tackle percentage down to 17% or 18%, which means everyone else is doing their job. The Redskins rely more heavily on their safeties (only Tennessee and Oakland are higher), which says to me that the DLs/LBs aren't setting the edge against the run, and the DBs are missing tackles at the point of reception.
One thing that the mix doesn't seem to address: cornerbacks will have fewer tackles if they can either convince the QB not to throw (with good coverage) or defend/INT a pass. Conversely, a set of DBs with lots of tackles is giving up pass yards even if they knock the guy over as soon as the pass is complete. [Then again, look at IND -- their CBs have lots of tackles because their offense forces other teams to throw. The Jets are closer to the mean because Revis and Cromartie force opposing teams to *not* throw.]
I wonder about how these statistics relate to points allowed, yards allowed, etc. That'd be an interesting study.
From a Lions' fan this confirms everything I've seen this season. Very good DL, and awful LBs.
One way to improve the LB tackle % would be to show the percentage of plays they make when the DL *doesn't* make the play. Pittsburgh's LBs don't look as effective when you figure they only tackle .51/(1-0.9) = 56% of the time the DL doesn't do it.
Good idea, especially if we segregate run plays and pass plays.
I'm learning this works best if I just put up some interesting data and let smarter people do the analysis.
Interesting stuff.
50% is high for PIT, but its also hard to compare 4-3 teams with a 3-4 team like PIT. I'm not an expert by any means, but isnt the point of the 3-4 that the line just shows gap discipline and gets doubled? The 3-4 is a system where the LB are expected to make the plays.
I'm not saying that the Lions' LBs have played well, but I don't think the numbers above say they've been poor, especially not if we look a little more closely at the context.
First, the DL are making a greater-than-average share of tackles: only four teams have seen more tackles from their DL (caveats about 3-4 defenses apply). Some of those plays are plays where a LB might have been in position to make the tackle if the play had reached him, but a DL got there first.
Second, it's not like the Lions have seen a disproportionate number of running plays. Unlike last season, they've been in every game but one this year, and they've actually had the ball late in games (as opposed to watching their opponents grind the clock). If I have the numbers right, the Lions have seen run plays 44.5% of the time, compared to 42.9% across the NFL ... and that represents only 6 more running plays, just one per game. The LBs aren't letting through a host of running plays that they should be stopping, not from these stats, anyway.
By the way, this doesn't really tell you too much about secondary play, mostly because if a DB blows a tackle, usually there are two alternatives: another DB or a TD. (There are exceptions, of course, but on medium- to long-range pass plays, those options plus the receiver stumbling/going out of bounds are about it.) So even though we know the Lions' secondary is very bad, it won't show up here as it might if you were displaying, say, net yards per attempt or something like that.
Of course, if it's difficult to learn much from these numbers, it's because the stat on which they're based is simple: you either get a tackle or you don't, and each tackle counts the same. Julian Peterson and Ashlee Palmer are third and fourth on the team in solo tackles (behind Louis Delmas and C.C. Brown???), but we don't know if they're making tackles at the point of attack or ten yards downfield. And while the tackle numbers suggest that the "other" LB might be a problem, well, we already know that, because it's not DeAndre Levy (who's still injured) or Jordon Dizon (who's still injured) or Zack Follett (who's now injured). If you get 3 or 4 deep at any position other than WR or CB, well, you're not going to get much there.
As an old football guy, all I've got to say is at least 1/2 of your teams tackles should be made by LBs and DEs. If your safeties are making 1/3 or more your D is in big trouble.
You could model a team's pass and run defenses as two parallel series of gates, each with a reliability percentage. You could pre-calculate a team's Run Reliability and Pass Reliability based on past data, and then look at their projected opponent's offensive run/pass percentage to get a good idea of what to expect on Sunday. Another way to slice the data would be to look at the run and pass reliability scores for each positional grouping, as you've done above, but to segregate the two. If you graphed them as eight adjacent bars, I bet that you'd see "fingerprints" for successful defenses, and that a 3-4 like BAL/PIT/NYJ would have similar fingerprints.
So, off to reliability numbers... first, pass plays. Eliminate plays that go out of bounds, and then you would have several outcomes (best to worst):
- Sack, for which I would give the DL credit for a pass-tackle.
- Incomplete pass, for which I would credit the DEs with an equivalent pass-tackle.
- INT, for which I would credit the DE/S with more than one pass-tackle equivalent... but I don't know what number to use here.
- Complete pass, pass-tackle by DE (or LB dropping into coverage)
- Complete pass, pass-tackle by S
- Complete pass for score, which is a Pass Defense Failure... you actually get this number for free at the end from (1-Rpass)
Handle these as serial reliabilities.
Run plays are actually easier to account for. The DL should stop the run; if they don't, the LBs should (duh) back the line and make the tackle; if they don't, the DBs and S have to step up; and if they don't, the offense scores. So your combined reliability against the run is RRun = R(dl)+[(1-Rdl)*Rlb] + [1 - (1-Rdl)*Rlb] * Rdbs, and (1-Rrun) is your rate-of-failure-vs-run.