Comments on Advanced NFL Stats: San Diego's Defensive Woes

No. The refs just love the Colts. That was obvious...

2008-11-23T23:06:00.000-05:00

No. The refs just love the Colts. That was obviously a fumble by Gonzalez.

Merriman is a huge factor, obviously. But the Chargers D is doing almost as well in other categories, just not in ints.

The chargers defensive stats are so poor this seas...

2008-11-23T22:56:00.000-05:00

The chargers defensive stats are so poor this season because Merriman is out for one, and the refs are totally against the chargers almost every game

Anon-Well put.

Good defenses will generally...

2008-10-31T15:31:00.000-04:00

Anon-Well put.

Good defenses will generally get more total interceptions because opposing offenses are behind and are passing more often.

I just tested another variable. I thought that a good run defense would make an offense more predictable and lead to a better interception rate. The (full-season) correlation between def YPC and def int rate is -0.08 (p=0.37), so not much connection there.

On the offensive side, there is some consistency in int rate. It auto-correlates (within season) at 0.27. It's not a lot, but helps put interceptions in general in perspective--overwhelmingly random, but 'thrown' far more than 'taken.'

a better defense will have a higher underlying pro...

2008-10-31T15:16:00.000-04:00

a better defense will have a higher underlying probability of making an interception, but this PR is something we never actually observe, we can only infer it from data. We'd expect a D with a .05 INT PR to intercept more passes than one with a .04 PR, but due to random variation this won't always be the case and so we might mistake the weaker defense for the stronger simply due to luck. I also think there are too many unknown variables at play to ever realistically model interceptions --> a ball that tips off a receiver's hands straight up in the air and is intercepted is the fault of the WR and pure luck to the DEF. This is probably as close to a random variable as you'll find in football. What are some others?

okk, so sacks don't correlate well, but as we seem...

2008-10-31T13:11:00.000-04:00

okk, so sacks don't correlate well, but as we seem to be agreeing, pass rush / pressure (however you measure it) creates INTs, not a sack in itself, although one would think the two are linked i guess.

anyway, i still think "good" defenses eg in yds per game or total points over a season will have higher INTs than "bad" ones.

Surely this must be the case?

Nick (again)

Anon-When I say they're "random," I should be more...

2008-10-31T08:56:00.000-04:00

Anon-When I say they're "random," I should be more precise. Really, I should say "they don't predict themselves well." In other words, having a high def int rate in the first half of the year does not mean a team will have a high rate in the 2nd half.

I looked at exactly what you asked in this article. I tried predicting late season def int rate with other related stats from the early season. Def pass eff correlated at r=0.01, and sack rate correlated at r=0.08.

Admittedly, sack rate is not equivalent to "pass rush." There are hurries and hits, etc. But we would expect some correlation between sacks and ints. But there isn't.

So statistically INTs are random, but does that me...

2008-10-31T06:42:00.000-04:00

So statistically INTs are random, but does that mean that the better ranked defenses over the years do not generate more INTs than poor ones? Intuitively, INTs come from (apart from the random) great plays by DBs, and a pass rush that pressures the QB, even if they don't get to sack him.
I know what your numbers say, but why, intuitively, do they not make sense? Not disagreeing, but a bit confused!
Kind regards,
Nick

I think the best way to explain defensive intercep...

2008-10-30T00:36:00.000-04:00

I think the best way to explain defensive interceptions is that they happen on the margins, because the offensive game plan already takes into account the talent of the defense. That is, if a DB is terrible, there will be more passes thrown that way, and more passes thrown in places where a good DB would have a play. Conversely, great DBs like Nnamdi Asomugha or Champ Bailey simply see fewer balls thrown their way, and the passes that do come their way are often safe, shorter passes that are harder to pick off.

That said, the .08 correlation that is left over, while quite small, may reflect a real and repeatable skill. You would obviously need a lot more data to prove that out, but my guess is that some corners, after adjusting for their general coverage and athletic talent, are better at hanging onto an errant pass. So while the opportunity and luck of having the QB throw a particular sort of inaccurate pass is the dominant factor, skill on the DB's part may play some role.