Writers of Pro Football Prospectus 2008

02 Apr 2010

Offensive Line Continuity Scores

In PFP 2007, we introduced OL Continuity Score, a measure of how much each team changes its offensive line starters during the season. I thought people might be interested to see the scores from 2009. There's definitely a correlation between poor offense and a lack of offensive line continuity, although the cause-and-effect relationship can go both ways. (The Colts are an exception, caused in part by the weirdness of Weeks 15-17.)

Continuity score starts with 48 and then subtracts:

  • The number of players over five who started at least one game on the offensive line;
  • The number of times the team started at least one different lineman compared to the game before; and
  • The difference between 16 and that team's longest streak where the same line started consecutive games.

The New York Jets were the only team with a perfect 48 score, meaning all five linemen starting all 16 games, after five different teams scored 48 in 2008. No team in the last decade has put up a 48 for three years in a row.

Year Team Continuity
Score
Longest
Stretch
Number of
Starters
Line
Changes
Cont. Score
2008
Change
2009 NYJ 48 16 5 0 48 0
2009 TB 40 11 6 2 35 5
2009 ARI 39 11 6 3 48 -9
2009 DAL 39 9 6 1 31 8
2009 CAR 38 10 7 2 25 13
2009 PIT 37 9 6 3 39 -2
2009 MIA 35 8 7 3 39 -4
2009 TEN 35 8 6 4 45 -10
2009 ATL 34 8 7 4 38 -4
2009 JAC 34 9 7 5 32 2
2009 NO 34 6 6 3 28 6
Year Team Continuity
Score
Longest
Stretch
Number of
Starters
Line
Changes
Cont. Score
2008
Change
2009 CHI 33 6 7 3 48 -15
2009 SD 33 7 7 4 33 0
2009 HOU 32 7 8 4 48 -16
2009 MIN 32 5 6 4 32 0
2009 NYG 32 6 7 4 48 -16
2009 PHI 31 6 7 5 30 1
2009 SF 31 6 7 5 29 2
2009 SEA 30 8 10 5 26 4
2009 BAL 29 4 7 5 27 2
2009 GB 29 7 8 7 30 -1
2009 NE 29 5 7 6 35 -6
Year Team Continuity
Score
Longest
Stretch
Number of
Starters
Line
Changes
Cont. Score
2008
Change
2009 CLE 28 3 7 5 30 -2
2009 CIN 27 4 8 6 37 -10
2009 KC 27 4 8 6 29 -2
2009 STL 27 5 9 6 25 2
2009 DEN 26 3 7 7 48 -22
2009 IND 26 4 8 7 29 -3
2009 DET 25 4 8 8 21 4
2009 OAK 25 4 9 7 29 -4
2009 WAS 23 4 11 7 29 -6
2009 BUF 21 3 10 9 30 -9

Thanks to intern Jared Cohen for putting this together, and to Jason McKinley for inventing continuity score in the first place.

Posted by: Aaron Schatz on 02 Apr 2010

35 comments, Last at 08 Apr 2010, 7:20am by Bowl Game Anomaly

Comments

1
by are-tee :: Fri, 04/02/2010 - 3:42pm

I'd be interested to see the continuity stretching from the previous year also. If I'm not mistaken, the Jets have started the same five linemen for 35 straight games, including playoffs.

2
by Led :: Fri, 04/02/2010 - 4:25pm

I'm pretty sure Mangold, Brick and Moore started all 16 games in both 2006-2007. Moore hasn't missed a game since 2004. *knocks on wood*

3
by are-tee :: Fri, 04/02/2010 - 4:40pm

That's correct, although Faneca and Wooody didn't join the team until 2008.

4
by >implying implications (not verified) :: Fri, 04/02/2010 - 5:44pm

Of the possible teams for McNabb:
ARI - 39
MIN - 32
PHI - 31
SF - 31
STL - 27
OAK - 25
BUF - 21

Looks like the best place for him might be Arizona, given that he also summers there. Oakland not quite as bad as Buffalo, but still understandable why he wouldn't want to be traded into that madhouse.

10
by tuluse :: Sat, 04/03/2010 - 3:43am

nevermind

8
by Tim Wilson :: Fri, 04/02/2010 - 10:34pm

Why are we looking at this relative to its benefit to Donovan McNabb? Also, his current team had a poor score last year and that offense seemed to do all right with him at the helm.

5
by FireSnake78 (not verified) :: Fri, 04/02/2010 - 7:24pm

Useless. Besides the Jets with a perfect score, this stat looks random to me (at least the way how it is computed leaves more questionsmarks than it answers).

6
by Thok :: Fri, 04/02/2010 - 7:52pm

The Colts are an exception, caused in part by the weirdness of Weeks 15-17.

That's a funny way of spelling Peyton Manning.

7
by Temo :: Fri, 04/02/2010 - 10:22pm

The point is that they rested starters in those weeks, making it seem like they didn't have offensive line continuity.

31
by Bobman :: Mon, 04/05/2010 - 2:17pm

Actually, I am pretty sure LT Charlie Johnson sat out the Jets game, making it a little more logical to pull Manning eventually to preserve his health, just in case. So they DID have an OL blip that game, though your larger point is well made.

(and that is mainly Weeks 16-17... in Week 15 it was Mathis and Freeney who were on a pitch count until the 4th quarter, when they played a lot and the Colts suddenly shut down the Jags. And resting DL starters shouldn't affect their OL continuity scores.)

9
by Jeff Fogle :: Fri, 04/02/2010 - 11:11pm

I was thinking Indy played their starters the first half of Week 15. Can intern Jared Cohen estimate what Indy's score would have been if they hadn't tinkered? Thanks in advance if possible.

12
by DZ (not verified) :: Sat, 04/03/2010 - 9:53am

Actually, I think the discontinuity for the Colts goes deeper than just those last two games.

Plus, Indy's line wasn't actually very good last year, so it's not surprising to see them near the bottom.

32
by Bobman :: Mon, 04/05/2010 - 2:19pm

I was surprised to see that the 2009 line had a lower score than 2008, when Lilja missed 16, Saturday missed 4, and a whole lot of other guys missed, plus Manning was gimpy for almost two months.

Perhaps a healthier Addai and manning made it seem as if the line was doing better. Numbers don't lie, do they?

11
by Brendan Scolari :: Sat, 04/03/2010 - 5:18am

Ok, I might be missing something obvious here, but isn't this stat completely meaningless? At least if the point is to correlate offensive line continuity with offensive line performance it is, for two obvious reasons:

1. Teams with higher score had their line stay healthier than teams with lower scores

2. Teams with higher scores didn't have to bench, rotate, or platoon players because of poor performance or inexperience as much as teams with lower scores

If you're trying to use this as a study going along with the old football saying about continuity on the offensive line leading to better play, it seems too flawed to work. Of course teams who have players play well enough not to get benched and stay reasonably healthy are going to have better performance than teams that don't have those two things. It seems the only way to accurately measure that sort of possible effect would be to see if teams that replaced lineman with different but equally skilled players had a drop-off in performance, but of course that would be incredibly tough to measure.

Forgive me if I've missed the point of the study.

13
by Dean :: Sat, 04/03/2010 - 9:58am

They did observe that it was correllation, not causation.

17
by BDC :: Sat, 04/03/2010 - 1:23pm

But what is the point of all this? If any other news site posted something or wrote something that showed correlation without causation (you run to win the game anyone???) we would all be laughing it up about what idiots they are and how much smarter we are then that, etc.

What precisely makes this any different, other then it was "our" site that posted it?

19
by tuluse :: Sat, 04/03/2010 - 3:03pm

Sometimes it's worth it just to measure what's happening, even if it doesn't add greater understanding. They are quantifying how much change offensive lines went through.

21
by BDC :: Sun, 04/04/2010 - 11:30am

I'll remember that when we're mocking someone for the 10 millionth time for quoting a stat like, "team x has a record of y when they run the ball at least z number of times".

23
by spatne (not verified) :: Sun, 04/04/2010 - 12:05pm

Your example is different, though, because it includes a condition (running z number of times) and a result (a W-L record of y). And when people quote stats like that, they're (at least tacitly) stating that running causes winning.

The continuity score is only a condition. It does not state or even imply that teams won or lost because of that condition, or that their OLs were better or worse. It's just a measure of continuity. It's a fair question to ask what the point is. After all, if you can't link OL continuity to some measure of productivity, why is it important? But I don't think it's fair to say that what FO has done is create a "kneel to win" type of metric.

30
by tuluse :: Sun, 04/04/2010 - 7:55pm

"team x has a record of y when they run the ball at least z number of times".

That is measuring two things and implying there is a causal relationship. Continuity score is more like just counting the number of rushing attempts a team has a in a season.

16
by jimbohead :: Sat, 04/03/2010 - 1:07pm

I think it would be more interesting to see someone differentiate between those two effects. Obviously, linemen get injured, and its mostly random (though some are injury prone). What about changes due to performance? How often does that actually happen? And when it does happen, given that every lineman is always a little dinged up, how often is it disguised as an injury issue?

Personally, I can't think of a good example from this last season where a lineman got benched.

20
by tuluse :: Sat, 04/03/2010 - 3:04pm

Frank Omiyale and Orlando Pace.

33
by Bobman :: Mon, 04/05/2010 - 2:22pm

Benching ALMOST: In preseason, Caldwell named former backup C Johnson the starter over recent 2nd round pick (and two year starter) Tony Ugoh.

So it doesn't count as an in-season benching, but sure put the Colts fans in turmoil.

14
by Dean :: Sat, 04/03/2010 - 10:03am

I have to think that the Rams, of all teams, could have a much better OL this year.

If Jason Smith develops, that solidifies LT.
Jacob Bell was a big money signing 2 years ago who played well when he was healthy, but nobody notices when you're 6-42 over the last 3 years.
Last years big money signing, Jason Brown, looks like he's worth the money at C.
Hank Fraley is no star, but he should certainly make a servicable RG.
Alex Barron (if he signs his tender), for all his faults, is an extremely gifted tackle, and will manhandle the DE opposite him - except for the 2 plays a game when he takes a bonehead penalty.
Adam Goldberg, John Greco, and Mark Setterstrom are all young and showed some ability last year, and give the unit plenty of depth. Greco could push for a starting spot.

This ain't the '77 Raiders, but it has the makings of a squared away line. Now if they draft some skill players and some defense...

15
by jimbohead :: Sat, 04/03/2010 - 1:03pm

That was my impression as well. I think a lot of the offensive futility last year had to do with receivers just never being able to get open. Bulger isn't actually that bad of a QB, and when he had receivers a couple years ago, he had a pretty decent quick release. Granted, at this point he's not worth the incredible salary bump he's getting, but from a line perspective, Bradford or whoever goes behind center probably won't get the Carr or Smith experience.

As long as they can get a receiver, that is. Who in 2nd rd and beyond can play the slot?

18
by Drunkmonkey :: Sat, 04/03/2010 - 1:29pm

I have a question about the score totals. If a starter goes down in mid-season, then shouldn't his loss be accounted for over the rest of the season? I mean, it seems from the description of how the score is devised, that once a new player steps in to replace the injured starter, after the first week the replacement player starts, he is seen as the regular starter for the rest of his duration in the starting lineup according to this measurement. But shouldn't the fact that the new starter is only seen as a replacement until the regular starter gets back be taken into account?

I know that what I'm getting at would basically suggest that this is a useless statistic, unless you want to show a good reason why the Jets did so well this year, but at the same time, you can't really compare the Jets' score with anybody else to show that this was a reason for success, since nobody else has any correlation. But the reason I bring this up is just that I don't think continuity is really represented on a long-term basis, which is what a team is probably really trying to get with their offensive line, if you just count replacement players as starters after their second week in the starting lineup.

22
by Aaron Schatz :: Sun, 04/04/2010 - 11:44am

Man, are you people bitchy. At this point, we may need to limit the discussion threads to one or two "main research projects" per year. Apparently, "we thought this was as an interesting little stat" is just not enough for some of you. What's the point of the discussion threads if they consist of nothing but complaints?

24
by langsty :: Sun, 04/04/2010 - 12:06pm

u mad

27
by Marko :: Sun, 04/04/2010 - 5:32pm

You want to complain! Look at these shoes. I've only had them three weeks and the heels are worn right through.

http://www.youtube.com/watch?v=kQFKtI6gn9Y

25
by capt. Anonymous (not verified) :: Sun, 04/04/2010 - 12:08pm

BDC,
I don't think F.O had drawn any conclusions from this data set. We would have to mock them if they took this info and then made a statement to the effect of the NYJ obviously being the best offense in football.

26
by Jeff Fogle :: Sun, 04/04/2010 - 12:22pm

Happy Easter everyone...

What's the point of having a discussion thread if everything is praise?

This is a stathead site. By my count, four of the 21 posts prior to Aaron were clearly "bitchy" to use his word (5-11-17-21). #6 could be taken that way...though I'm still not sure what #6's main point was. Did Manning sit out games 16-17, but the offensive line started? (Pinning that down would help with the particular issue of whether or not there was weirdness in weeks 15-17.) That's a BI (Bitchiness Index) of 22% if you count #6 as the fifth example.

Tough to justify limiting discussion threads to a couple per year based on that I think. "Nothing but complaints?"

We work toward the truth through analysis and peer review. 22% critical comments isn't an outlandish percentage, particularly when the basic points made in 5-11-17-21 had some veracity to them with this particular stat sample.

The whole world isn't a Larry David sitcom. People aren't your enemies just because they disagree with you.

29
by Thok :: Sun, 04/04/2010 - 7:41pm

though I'm still not sure what #6's main point was.

Even if you take into account week 15-17 wonkiness, the Colts still have a relatively low offensive line continuity given their DVOA (assuming perfect continuity for those weeks only pushes them to average consistency overall, while their offense should also be corrected to something like 3rd in offensive DVOA.)

Basically, having Peyton Manning also takes care of some of the problems arising from not having offensive line consistency. I could have said it with a bit less snark.

34
by Bobman :: Mon, 04/05/2010 - 2:28pm

The last two MVPs bear you out on that. Possibly the worst two OL years of the decade for him, clearly the worst two rushing years for the team, and still that guy manages.

As Will Allen will attest, there's a bargain made with the devil there. Proof will be next year when it's just Jeff Saturday and two UDFA OTs on the line, Manning, and 7 receivers. Still an incredibly low sack rate (surprising) and a low success rate on 3rd and short (not surprising). Fortunately, they only have seven 3rd and shorts all season....

28
by NHPatsFan (not verified) :: Sun, 04/04/2010 - 5:54pm

Okay, it's a lazy Sunday so I took the Offensive Line stats and these continuity scores and tried to fiddle with ways to put them together.

I tried to come up with an OLine Composite score that I calculated by inverting (1 becomes 32 etc.) ranks for Adjusted Line Yards, Power, Stuffed, and Passing. I gave Power and Stuffed only 0.5 weight, because they're already somewhat reflected in ALY and Passing 1.5 weight, because passing is so much more prevalent than rushing. Finally, I summed the total scores to come up with an OLine Composite Score that doesn't mean much except to say "who performed best" relatively speaking for one season.

I then calculated a Continuity / Composite Score ratio that I found kinda interesting, if not exactly meaningful, but before passing that along, I wanted to get some feedback on the Composite Score calculation, because if that's B.S., then there's no point putting up the end results.

Suggestions for a fix to the OLine Composite Score are welcome, then I'll publish what I did with that and the Continuity scores.

35
by Bowl Game Anomaly :: Thu, 04/08/2010 - 7:20am

Giving extra weight to Passing is a mistake if you are trying to assess how good an OL was (rather than how effective the whole offense was). Pass protection is something the QB has a lot of control over, and the OL is much more responsible for its run-blocking performance than its passing performance. This is borne out by the fact that when QBs switch teams they bring their sack rates with them, and that different QBs on the same team tend to have different sack rates (as documented by the P-F-R blog).

I'm also not convinced that ALY is a really good measure of OL run-blocking, although I guess it's better than raw yards per carry.