Writers of Pro Football Prospectus 2008

19 Jan 2006

FO Greatest Hits: Third Down is the Charm for NFL Turnarounds

We take you back now to that long ago time called "August," when your humble FO editor penned an article for the New York Times that concluded like so: "Seattle and Carolina addressed their third-down issues in the off-season ... making Seattle and Carolina Super Bowl contenders and making Matt Hasselbeck the answer to the question, 'Who is this year's Drew Brees?'" Five months later, here we are.

Posted by: Aaron Schatz on 19 Jan 2006

37 comments, Last at 20 Jan 2006, 8:48pm by tmac


by Ruben (not verified) :: Thu, 01/19/2006 - 1:04am

Maybe you should start selling magic beans...

Question, though: in the article, you say Seattle and Carolina addressed their third-down issues in the off-season...The Panthers return many players from injury and added cornerback Ken Lucas, giving them enough depth to move the 2003 playoff star Ricky Manning into the nickel defense."

This doesn't impact 3rd down, but I am curious as to what it was the Pathers added to make them so freakish? I believe you when you say you know what it was, but I wanna know, too!

Did the opposition simply believe that Moose's replacement (I think we finally found the replacement prototype: "Panther Receiver Not Named Steve) would be as good as he was? Were the Panthers in fewer 3rd-and-long or second-and-long situations? Do tell (please)!

by Paul (not verified) :: Thu, 01/19/2006 - 1:11am

Since 28 teams' seasons are over and are looking to next season, which teams vastly overachieved on 3rd down this year and which ones underperformed? And will the draft be any help to them?

by Vash (not verified) :: Thu, 01/19/2006 - 1:14am

So are the Steelers still the team that's supposed to improve dramatically next year, despite winning 11 games and making the AFC Championship this year?

by Crushinator (not verified) :: Thu, 01/19/2006 - 1:23am

We take you back now to that long ago time called “August,� when your humble FO editor penned an article for the New York Times that concluded like so: “Seattle and Carolina addressed their third-down issues in the off-season … making Seattle and Carolina Super Bowl contenders and making Matt Hasselbeck the answer to the question, ‘Who is this year’s Drew Brees?’� Five months later, here we are.

So, what's new with Kevin Jones?

Sorry, I couldn't resist. =P

by DGL (not verified) :: Thu, 01/19/2006 - 1:47am

Paul - As Vash intimated, the team that FO picked as the "Offense Most Likely to Improve" based on the 2005 regular season was the Pittsburgh Steelers.

by RowdyRoddyPiper (not verified) :: Thu, 01/19/2006 - 2:02am

I'm actually kind of hoping this improvement for next year kind of kicks in this weekend...I mean it did say next year...and it was written in 2005.

by Tally (not verified) :: Thu, 01/19/2006 - 2:29am


The article specifically states that it was Carolina's DEFENSE on 3rd down that was the issue in 04 and which was addressed for 05. For Seattle, it was offense that was addressed.

by Israel (not verified) :: Thu, 01/19/2006 - 4:03am

I am trying to figure out what Cowher is supposed to do to make his third down better. Replace Verron Haynes on offense? That doesn't seem right because Verron does a good job. Change something - exactly what? - on defense? And are those things he can do now?

by Kal (not verified) :: Thu, 01/19/2006 - 4:06am

Well, one 3rd down performance that hopefully won't be repeated is Maddox's snaps. If you take away Ben's 3rd down performances while he was injured and Maddox's performances, I suspect you'd see an immediate improvement. You're also going to see some improvement with Willie Parker and Heath Miller, even as good as they were this year. Plus, Ben should be better too.

The only thing you probably won't have is Bettis.

by Pat (not verified) :: Thu, 01/19/2006 - 4:22am


It's offense-only. And the Steelers struggles on 3rd down occurred pretty early in the season. Look at the New England game (3/13? What?!), the Houston game (4/11?), San Diego (4/10), Cincinnati (4/11) and all of those were with the Steelers averaging 5 yards per play in those games!

A lot of those failed 3rd down conversions came from Roethlisberger throwing incomplete on 3rd down, or not throwing far enough. (And I mean 3rd and short/moderate, not 3rd and long).

I'm not sure what that really means - it might mean that Roethlisberger needs a bit more time to learn to read coverage well. It might also mean that Cowher isn't calling plays that can easily pick up first downs. In the Cincinnati game, he did go a little nutzoid with the two gadget plays on 3rd down and short!

by The Load (not verified) :: Thu, 01/19/2006 - 7:26am

I think some of the Steelers' problems with 3rd down come frmo playcalling. For much of the season, 3rd and 3 was a shotgun formation. With a power running team, that surprises me.

I also think they don't use Heath Miller enough as a receiver (he has beautiful hands), and well as not using their backs enough on passing downs. The only time Willy Parker gets the ball on a pass play is a screen. With Miller working the middle short, ARE stretching the field, and Parker coming out into the flats, I could see opposing DC having fits...

I think when some of these issues are worked out, there may be better 3rd down completions.

by tmac (not verified) :: Thu, 01/19/2006 - 8:01am

Move over Shaun Alexander,Jerome Pathon is the real MVP.

by tmac (not verified) :: Thu, 01/19/2006 - 8:04am

And if so, why didnt the Falcons finish bettter than 8-8?

by DGL (not verified) :: Thu, 01/19/2006 - 11:29am

#13: FOMBC.

by someone (not verified) :: Thu, 01/19/2006 - 11:56am

Remind me what your predictions were for Jacksonville, and on what basis, again?

by Dr. Regressing Towards Being Mean (not verified) :: Thu, 01/19/2006 - 1:03pm

Aaron, do you think the tendency towards improvement for teams that are weak on third down is primarily regression towards the mean on relatively small sample sizes, primarily real improvement on third down, or a more-or-less equal combination of the two?

by Dman (not verified) :: Thu, 01/19/2006 - 1:05pm

I don't think aaron is claiming that the 3rd down problem will be solved by play calling or personell. I believe his point is that decreased performance on third down versus first or second is a statistical outlier that will tend to regress to the mean (i.e. better third down conversions). Which would of course translate in to better performance and more wins. Thats my understanding of it anyway.

by Pat (not verified) :: Thu, 01/19/2006 - 1:38pm

Jacksonville's prediction was very, very wrong in terms of record, but that has a lot more to do with the fact that Houston and Tennessee were projected to do quite well. Aaron was notably nervous about the Tennessee prediction, since the Titans replaced their defense with one that they bought at the 7-11 for a buck and a quarter.

Who knows why the Houston projection was wrong, though I think it might have something to do with the fact that the staff down there is incompetent.

by pawnking (not verified) :: Thu, 01/19/2006 - 1:43pm

I believe that a combination of factors play into a team's predictive improvement, but FO is the first which uncovered the importance of third down under of over performance.

Take in Steelers. If their third down performance means they underperformed this year, we also have to factor in Pythag over or underperformance, Key losses or acquisitions in the offseason, and of course schedule in 06 vs. 05.

Let's also remember that the Steelers barely made the playoffs this year. It's not as if we're looking at improvement on a 14-2 team.

by Pat (not verified) :: Thu, 01/19/2006 - 1:47pm

It should also be noted that I think the projections (which include more than just regression to the mean) have problems with dealing with organizational disasters - that is, front offices/coaches that suck - see New Orleans, Houston, and Detroit. The system just keeps wanting them to improve (and really, so do the rest of us), but... they don't.

And of course, it can't predict injury disasters, like the Jets and the Eagles. But other than that, the predictions for a lot of teams were very spot on - and most of them didn't involve "let's assume everyone stays the same."

by JMM (not verified) :: Thu, 01/19/2006 - 1:48pm

Some numbers to consider re Pitt's 3rd down performance.

214.3 NFL average 3rd dw attempts
243 Max 3rd dwn attempts (Balt)
187 Min 3rd dwn attempts (Indy)
192 Pitt's 3rd dwn attempts

298 NFL average 1st dw's per game
190 Min 1st dwns (San Fran)
363 Max 1st dwns (Indy)
298 Pitt 1st dwns

Pitt had few 3rd down plays, but an average amount of 1st downs.

by Israel (not verified) :: Thu, 01/19/2006 - 1:56pm

Pat (#10) - I asked Aaron about this a few weeks ago and he said it's the same phenomenon on defense as on offense. It's the Steelers defense which has been getting killed on third down.

by Israel (not verified) :: Thu, 01/19/2006 - 1:58pm

Dman (17) - I asked just that question in Aaron's mailbag a couple of weeks ago, without using the specific term "statistical outlier." He hasn't published a mailbag piece since and I am looking forward to his remarks.

by Just Another Falcon Fan (not verified) :: Thu, 01/19/2006 - 2:21pm

From the Atlanta Journal-Constitution:

"The Falcons led the NFL in rushing both years, improved dramatically this season to lead the league in third-down defense (allowing a conversion rate of 28.5 percent), and tied with Cincinnati for second in offensive third-down conversion rate (42.9 percent).

But making lots of good plays is not as important as making a handful of really big ones in close games. This season, with an influx of young players (especially on defense), the Falcons didn't make enough."

When the defense collapsed, it felt like the opposing offenses didn't have many 3rd and short situations; they either had already picked up the first down or were facing a 3rd and long.

by pawnking (not verified) :: Thu, 01/19/2006 - 3:37pm

JAFF, let's not forget that DVOA had Atlanta being one of the most over-acheiving teams in the NFL last year, outperforming their Pythag and DVOA projections considerably. Were they a 8-8 team last year that got lucky? I'd say that was likely.

by someone (not verified) :: Thu, 01/19/2006 - 4:28pm

"Jacksonville’s prediction was very, very wrong in terms of record"

You can say that again. As for the Tennessee/Houston thing, yeah whatever, why not add that the dog ate your homework too. BTW Same thing could be said for Arizona/St Louis - don't know what FO projections were for both teams but they are widely considered to be disappointments this year. Probably more of a reason than cutting Koren Robinson for Jerome Pathon, which is the prime non-statistical reason for the Seattle improvement given in the article. How'd that work out again? As for Carolina, yeah, FO was really ahead of the curve in predicting them as NFC favourites. I don't recall anybody else doing that.

by NF (not verified) :: Thu, 01/19/2006 - 4:45pm

When I first read this extra points post, I thought it said "this year's Drew Bledsoe" and was very confused until I re-read it.

by mactbone (not verified) :: Thu, 01/19/2006 - 6:08pm

Actually, FO predicted Philadelphia as overwhelming favorites.

It predicted Green Bay's decline, mostly the offense getting older and Ahman Green completely losing it - I can't believe people are seriously considering him a major FA next year, now he's fumbling machine that can't run.

It predicted the Jets decline, again due to age and a crumbling running back.

Also, it's important to note that nobody can be perfect when predicting the future of a semi random event like a football game. If Aaron knew his projections were perfect, I don't think we'd have this site.

by Jim A (not verified) :: Thu, 01/19/2006 - 7:23pm

Re: 16/17 and whether the effect is mainly due to regression toward the mean of a small sample size (3rd down plays). I hypothesized that regression toward the mean would also be observed on second down to a lesser extent. I asked Aaron about this and he said that a similar effect was not observed on second downs at all. If that's true, it suggests that the third down unpredictability is caused by more than just regression toward the mean. Perhaps it has to do with a higher turnover rate among the players who primarily see action on third downs, for example the #3 and #4 WRs and the nickel and dime backs.

by Pat (not verified) :: Thu, 01/19/2006 - 7:41pm

Actually, FO predicted Philadelphia as overwhelming favorites.

Yes, but the loss of a starting QB lowers a team's chances of winning games by something like 20%. It's flat huge. You can't predict something like that at all.

Well, that's not true - you could include depth of the QB spot, but then that wouldn't predict Philly tanking, it would just increase their standard deviation.

Both the Jets offense and the Eagles offense were entirely wrong, and that's from QB injury, in my opinion.

And amazingly enough, if you have a model that doesn't take things into account, and say "I think this will be wrong because of this", and it turns out that in the end, you were right (Tennessee's defense did, in fact, suck horrendously) you can excuse yourself from the poor prediction.

by Dman (not verified) :: Thu, 01/19/2006 - 7:47pm

Perhaps the correlation is related to play calling then. That is to say that play calling might be significantly different on 3rd down to produce this effect. I'm looking forward to that mailbag too israel.

by Comrade Jason (not verified) :: Thu, 01/19/2006 - 9:39pm

So what does it mean that Seattle only completed 39.6% of their third-downs this year, making them very average in that respect? Does that mean their offense will likely improve to, or is that number too connected to the fact that they had (I believe) the 2nd most first downs and the 2nd fewest third downs? (Behind Indy in both cases, obviously.)

by tmac (not verified) :: Fri, 01/20/2006 - 1:10am

Maybe Carolina was improved this year because they avoided (Jenkins excluded) the debilitating injury bug that hit them last year. Seattle started holding on to thrown footballs (due in no small part to #81 exiting) and learning how to hold a lead.

by tmac (not verified) :: Fri, 01/20/2006 - 1:22am

Denver's improvement is similar to Seattle's in that they had the same trouble. They blew a 13 point 4qtr lead at the Giants and let the Patriots mak a game of it late after having an eighteen point lead. They jumped on Philly 28-0 and watched the lead shrink to 28-21 with Philly in posession at the Bronco 30. McNabb fired for the end zone and Foxworth intercepted. Denver went on to win 49-21.Since that point they have been a team that once it gains control maintains it.

by admin :: Fri, 01/20/2006 - 1:40am

Lots of things to respond to here, but let me point one thing out, and tell you the real reason I posted this.

There is a problem that human beings have called the "broken tile syndrome." Imagine that you walk into a church with an ornate painting on the ceiling, but one tile is broken. When the average person looks at that ceiling, the first thing that will draw their eye is the broken tile, not the thousands of perfectly painted ones. It is human nature to focus on what is wrong rather than what is right.

With that in mind, it is true that the book was pessimistic about the Jaguars and I was wrong about Jacksonville personally. If you remember our preseason projections, you know that I correctly picked 5 of 6 AFC playoff teams, and if you read King Kaufman's column, you know that I picked more correct playoff teams than any other preseason prognosticator. So I can live with being wrong about the Jaguars, even if some guy from Jacksonville thinks this proves that I am an idiot who knows nothing about football, that we should shut down FO and I should go back to doing something else for a living.

Using the online handle "someone" sure makes it easy to anonymously take potshots at a website that provides you with pages and pages of content at the cost of absolutely zero dollars.

Now, the real reason I posted this. This article isn't just funny because Seattle and Carolina ended up in the championship game, or because Jerome Pathon was third on the Seattle depth chart at the time. It also ended up being wrong, sort of. (See, I'm not perfect. Shocker!) Seattle's third-down offense improved from 27th to 4th. But Carolina's third-down defense actually didn't improve that much, at least if you judge by rank. Carolina's third-down defense went from 33.4% (30th) to 10.4% (25th). A good DVOA improvement, but certainly not in line with the other teams written about in that chapter.

Sort of interesting, I think, that the Panthers ended up in the NFC Championship game anyway.

by someone (not verified) :: Fri, 01/20/2006 - 5:51pm

"Using the online handle “someone� sure makes it easy to anonymously take potshots at a website that provides you with pages and pages of content at the cost of absolutely zero dollars."

Genuinely, I have no idea what you're talking about with this. It sounds like you're confusing me with someone else. I'm not from Jacksonville for a start so please don't imply my particular criticism in this thread is based in some way on some irrational parochialism. I actually think you do a lot of good stuff on this site but if you're going to continually take shots at Peter King, Don Banks et al and in so doing leech their content for your readers and your hits, then you can take some criticism in response too when you try to thump your chest in such a selective way. Its actually a type of compliment that the site is worthy of such comment.

BTW, you're still doing the selective reporting thing - 5 out of 6 AFC playoff contenders? Congrats, now tell us how you did on the NFC.

by tmac (not verified) :: Fri, 01/20/2006 - 8:48pm

re 35 and 36 You guys better take it easy before somebody(one) starts crying.