Aggressiveness Index Revisited

Aggressiveness Index Revisited
Aggressiveness Index Revisited
Photo: USA Today Sports Images

by Aaron Schatz

Football Outsiders introduced the concept of Aggressiveness Index back in Pro Football Prospectus 2006. The goal was to find a way to rank coaches based on their tendencies on fourth downs in a manner that was easy to understand but accounted for the different rates at which the average coach will choose to "go for it" in different situations. Although no NFL coach is as aggressive as the data suggests he should be, we discovered there is quite a wide range of fourth-down tendencies among coaches.

Aggressiveness Index numbers center around 1.0 and generally describe how much more (or less) likely each coach is to go for it on fourth down compared to his peers; for example, a coach with 1.20 AI is roughly 20 percent more likely to go for it than an average coach in equivalent situations. The Aggressiveness Index excludes obvious catch-up situations: third quarter, trailing by 15 or more points; fourth quarter, trailing by nine or more points; and in the last five minutes of the game, trailing by any amount.

This offseason, Jim Armstrong -- who created the original Aggressiveness Index and is responsible for keeping both drive stats and AI stats for Football Outsiders -- worked on a slightly updated version of the Aggressiveness Index formula. Changes include:

  • Aggressiveness Index now accounts for plays on the offense's own side of the field. Previously, AI only looked at plays in opponent territory, but a number of controversial fourth-down calls in recent years (including the infamous Bill Belichick fourth-and-2 as well as Mike Smith's fourth-and-1 against New Orleans in overtime) suggested that we would get a more accurate picture by including plays before the 50-yard line. Because the average rate of going for it on these plays is so low, going for it on your own half of the field (actually, technically, going for it between your own goal line and your own 48-yard line) will make a coach look extremely aggressive compared to average. Therefore, we are giving these plays only half strength in order to prevent them from overwhelming all other plays in the metric.
  • Plays in the final 10 seconds of the half were removed, to account for Hail Mary-type situations. For example, Leslie Frazier had previously ranked very high in Aggressiveness Index for 2010 because the Vikings had "gone for it" by throwing a Hail Mary pass on fourth-and-14 with five seconds left in the second half of a 2010 game against Chicago.
  • Baselines were changed to account for the surprising fact that the average head coach has become less aggressive in recent seasons.

We wrote about this lack of aggressiveness a year ago and nothing really changed in 2012. The public profile of football analytics has increased significantly over the last couple years, and the need for coaches to be more aggressive on fourth downs is the one thing that every single independent analyst agrees upon. In fact, not only do we all agree that coaches need to go for it more often on fourth downs, we also all agree that this strategy is even more important now than it was a few years ago because of the continued rise of overall offensive levels across the league. And yet, despite a few aggressive high-profile decisions that spurred discussion around the Internet, on the whole NFL coaches have become slightly more risk-averse.

Perhaps the most shocking example of this de-evolution in NFL strategy is the man who has long represted the ideal of aggressive fourth-down decision making: Bill Belichick. Last year during the regular season, the Patriots didn't attempt a single run or pass play on fourth-and-2, and they went for it only two times out of eight opportunities in long field-goal range (between the 31- and 37-yard lines). This trend carried over into the playoffs; against Baltimore, the Patriots had fourth-and-2 twice and settled for a punt and a 31-yard field goal. They also punted twice in long field-goal range, although these were "no man's land" situations, i.e. not just in deep field-goal range but in deep field-goal range on fourth-and-long when none of the options are mathematically any good.

As a result, Belichick ranks a shocking 23rd out of 34 head coaches in Aggressiveness Index for 2012. This is a massive change from most of Belichick's career. Belichick has the fifth-highest career AI of any head coach with at least three full seasons between 1991 and 2012, and he ranked in the top six for AI every year between 2004 and 2010 before falling to 11th in 2011 (although he was sixth in 2011 in the older version of AI). 2012 was only the third season out of 18 when Belichick ranked in the bottom half of the league; the others were 1994 (0.76 AI, 21st) and 2003 (.84 AI, 23rd).

He doesn't seem like the kind of guy to ever react to the pressure of conventional wisdom, but is it possible that a few high-profile fourth-down failures have actually made Belichick more risk averse? That certainly seems to be the case with Mike Smith. Smith has a career AI of 1.31 but ranked dead last in 2012 with 0.62 AI. The Falcons only went for it once in 91 qualifying opportunities in 2012, and that opportunity itself barely qualifies. It was a handoff to Michael Turner on fourth-and-1, up 13 points on Philadelphia in Week 8 with only 20 seconds left in the game. They didn't convert.

With Belichick and (especially) Smith coming down with a case of the wussies, and Sean Payton suspended for the year, the 2012 leaders in Aggressiveness Index look much different than in other recent years. Jacksonville's Mike Mularkey finished the year as the league's most aggressive coach with 1.61 AI. That includes decisions like going for it on fourth-and-10 from near midfield against Houston in overtime and going for it on fourth-and-3 near midfield against Tennessee with a minute left in the second quarter. It doesn't include a lot of other aggressive fourth-down decisions which don't qualify for Aggressiveness Index because the Jaguars were losing by more than two touchdowns. I would love to believe that Mularkey was the league's most aggressive coach because the Jacksonville front office has taken a very public stance on using analytics, but you might be surprised to learn that Mularkey was nearly as aggressive during his first year as head coach in Buffalo, finishing second with 1.26 AI in 2004 before dropping to 12th with 0.85 AI in 2005.

(Digression: One thing I have learned in talking to a lot of front office people who are interested in analytics is that there is very little correlation between how much analytical work is being done in a front office and how much the head coach's on-field decisions seem to reflect the general precepts that have developed in the football analytics community over the last decade. For most teams doing analytics, the impact is coming in draft and free-agency decisions, and the difference that analytics can make between one free-agent signing and another can be very subtle. Eventually we'll get to the point where a lot of head coaches have buy-in, but we aren't there yet, even on teams where the salary cap analyst is regularly reading Football Outsiders and fully understands Brian Burke's fourth-down calculator. Digression over.)

Here are the Aggressiveness Index numbers for all head coaches in 2012:

Aggressiveness Index, 2012
Rank Coach Team AI Overall
Rate x Rank Coach Team AI Overall
1 Mularkey JAC 1.61 7 107 6.5% x 18 Philbin MIA 0.88 5 100 5.0%
2 Allen OAK 1.57 6 97 6.2% x 19 Munchak TEN 0.88 6 99 6.1%
3 Rex Ryan NYJ 1.42 8 102 7.8% x 20 Vitt NO 0.86 3 54 5.6%
4 Lewis CIN 1.31 15 119 12.6% x 21 Kubiak HOU 0.86 6 120 5.0%
5 Fisher STL 1.25 7 109 6.4% x 22 Crennel KC 0.85 7 95 7.4%
6 Whisenhunt ARI 1.14 4 123 3.3% x 23 Belichick NE 0.81 6 97 6.2%
7 McCarthy GB 1.14 11 108 10.2% x 24 Arians IND 0.81 4 73 5.5%
8 Reid PHI 1.06 11 95 11.6% x 25 Gailey BUF 0.78 2 90 2.2%
9 Shurmur CLE 1.04 6 115 5.2% x 26 Coughlin NYG 0.77 6 90 6.7%
10 Turner SD 1.03 6 107 5.6% x 27 Frazier MIN 0.75 4 104 3.8%
11 Tomlin PIT 1.02 10 107 9.3% x 28 Schiano TB 0.74 4 102 3.9%
12 L.Smith CHI 1.01 6 109 5.5% x 29 Garrett DAL 0.69 4 88 4.5%
13 Shanahan WAS 1.00 13 105 12.4% x 30 Kromer NO 0.67 0 32 0.0%
14 Carroll SEA 1.00 9 101 8.9% x 31 Pagano IND 0.67 0 22 0.0%
15 John Harbaugh BAL 0.98 6 114 5.3% x 32 Fox DEN 0.66 3 96 3.1%
16 Jim Harbaugh SF 0.92 8 107 7.5% x 33 Schwartz DET 0.65 2 98 2.0%
17 Rivera CAR 0.90 5 94 5.3% x 34 M.Smith ATL 0.62 1 92 1.1%

A couple of other interesting notes here:

  • Marvin Lewis was extremely aggressive in long field-goal range, going for it on seven out of 10 opportunities between the 31- and 37-yard lines.
  • Ken Whisenhunt comes out near the top of AI even though he only went for it once out of 11 qualifying fourth-and-1 opportunities, primarily because of a fake punt against the Jets and a fake field goal against the Bears. Yes, we count fakes as "going for it" for the purposes of Aggressiveness Index; it's hard to think of a play that would qualify as more aggressive than a fake from fourth-and-long.
  • Andy Reid, one of the least aggressive coaches of the past two decades, swam surprisingly upstream against the recent trend of coaches becoming more conservative; the Eagles went for it on eight of 12 qualifying fourth-and-1 opportunities.
  • No, I still don't know what's up with Jim Schwartz. One of these weeks when I don't feel buried by book editing and the Lions aren't in OTAs, I will call him up and ask him.

[ad placeholder 3]

As far as I remember, Football Outsiders hasn't published a full list of career AI since that first article back in Pro Football Prospectus 2006. We now have a new updated version of the metric and a lot more years in our database, so this would be a good time to run some career numbers for you all to enjoy and discuss. I've long referred to the most aggressive coaches as the "Holy Triumverate of Bills," Parcells, Belichick, and Cowher. The Holy Triumverate doesn't appear quite as high in this list as they would have a couple years ago; Cowher in particular drops once we include more seasons, because he was much less aggressive in his early years in Pittsburgh. This matches a trend that we generally see and probably could use a little more specific research: coaches tend to become more aggressive as they gain more experience, perhaps because they are less worried about being second-guessed by the press and/or ownership.

While analytics folks generally believe that the better coaches in the NFL will tend to be more aggressive, this list of coaches from 1991 through 2012 clearly demonstrates that there's a lot more to being a great NFL coach than simply making the mathematically correct decision on fourth downs. Some of the best coaches of the last two decades have been among the most conservative, particularly Reid and the other coaches of the Mike Holmgren coaching tree such as Jon Gruden and Holmgren himself. Meanwhile, the most aggressive coaches of the last two decades include Bruce Coslet, Jim Haslett, and... uh, well, just see for yourself who comes out at number one.

Aggressiveness Index, 1991-2012 (minimum three seasons)
Rk Coach Years AI Rk Coach Years AI Rk Coach Years AI Rk Coach Years AI
1 Kotite 6.0 1.66 x 22 Rhodes 5.0 1.20 x 43 Don Shula 5.0 1.06 x 64 Green 12.9 0.85
2 Ditka 5.0 1.64 x 23 Fisher 17.4 1.20 x 44 Nolan 3.4 1.05 x 65 McGinnis 3.4 0.85
3 Haslett 6.8 1.64 x 24 Phillips 8.7 1.20 x 45 Vermeil 8.0 1.04 x 66 Gibbs 6.0 0.85
4 Mularkey 3.0 1.63 x 25 Davis 3.7 1.18 x 46 Kubiak 7.0 1.03 x 67 Schwartz 4.0 0.85
5 Belichick 18.0 1.59 x 26 Del Rio 8.7 1.18 x 47 Martz 5.3 1.02 x 68 Levy 7.0 0.85
6 Coslet 6.8 1.58 x 27 Spagnuolo 3.0 1.18 x 48 Fontes 6.0 0.99 x 69 Ross 8.6 0.84
7 Parcells 11.0 1.58 x 28 R.Ryan 4.0 1.17 x 49 Marinelli 3.0 0.99 x 70 Crennel 5.2 0.83
8 Payton 6.0 1.51 x 29 Whisenhunt 6.0 1.17 x 50 Capers 8.0 0.97 x 71 Reeves 12.8 0.80
9 Seifert 9.0 1.51 x 30 Caldwell 3.0 1.16 x 51 McCarthy 7.0 0.96 x 72 Edwards 8.0 0.79
10 Johnson 7.0 1.43 x 31 Cowher 15.0 1.16 x 52 Mariucci 8.7 0.96 x 73 Switzer 4.0 0.79
11 Schottenheimer 14.0 1.36 x 32 Lewis 10.0 1.14 x 53 Bugel 4.0 0.96 x 74 Morris 3.0 0.77
12 Williams 3.0 1.34 x 33 Mora Jr. 4.0 1.13 x 54 Turner 14.8 0.96 x 75 Holmgren 17.0 0.77
13 Carroll 7.0 1.33 x 34 Sparano 3.8 1.13 x 55 Coughlin 17.0 0.95 x 76 Glanville 3.0 0.76
14 Jauron 8.9 1.33 x 35 Marchibroda 7.0 1.13 x 56 Mora Sr. 9.5 0.94 x 77 Reid 14.0 0.73
15 M.Smith 5.0 1.31 x 36 Tice 4.1 1.13 x 57 Shanahan 17.0 0.93 x 78 Jones 3.6 0.73
16 Dungy 13.0 1.28 x 37 Erickson 6.0 1.12 x 58 Gailey 5.0 0.93 x 79 Fox 11.0 0.73
17 Campo 3.0 1.26 x 38 Wyche 5.0 1.10 x 59 Tomlin 6.0 0.92 x 80 Infante 3.0 0.72
18 David Shula 4.4 1.25 x 39 Fassel 7.0 1.09 x 60 Flores 3.0 0.89 x 81 Billick 9.0 0.72
19 Tobin 4.6 1.22 x 40 L.Smith 9.0 1.09 x 61 Riley 3.0 0.89 x 82 Gruden 11.0 0.72
20 Mangini 5.0 1.22 x 41 John Harbaugh 5.0 1.08 x 62 Sherman 6.0 0.89 x 83 Pardee 3.6 0.68
21 Wannstedt 10.6 1.22 x 42 Shell 5.0 1.06 x 63 Childress 4.6 0.89 x 84 Knox 4.0 0.67

Other notes:

  • Five coaches went for it on more than half of all qualifying fourth-and-1 situations in our database, led by Bill Parcells (57.4 percent) as well as Don Shula (56.4 percent), Sam Wyche (52.5 percent), Belichick (52.4 percent), and Jack Del Rio (50.4 percent). Jerry Glanville (19.4 percent) was the coach least likely to go for it on fourth-and-1.
  • Coaches who went for it at least 20 percent of the time on fourth-and-2 include Cosley, Payton, Mularkey, and Mike Nolan. Four coaches never went for it on a qualifying fourth-and-2: Lindy Infante, Chuck Knox, Art Shell, and Tony Sparano.
  • Dave Campo (61.1 percent) and Cosley (60 percent) were by far the coaches most likely to go for it in long field-goal range (between the 31 and 37) while Wyche almost always took the long field goal, going for it just 7.9 percent of the time.
  • Although Jim Mora Sr. may not have been one of the most aggressive coaches of the last two decades, aggressiveness seems to run in other coaching families. Both Shulas are above 1.0 AI, as are both Rex and Buddy Ryan (1.50 AI in 1991-92) and both John and Jim Harbaugh (1.06 AI in 2011-12).


48 comments, Last at 03 Jun 2013, 10:31pm

#1 by The Ancient Mariner (not verified) // May 20, 2013 - 10:55am

Interesting--this supports the sense we have in Seattle that Carroll was much more risk-averse last season after a couple high-profile burnings the previous two seasons.

Points: 0

#2 by Jimmy // May 20, 2013 - 11:05am

That is all.

Points: 0

#3 by Ryan W (not verified) // May 20, 2013 - 11:13am

I'd be curious to see the 2012 McCarthy stats split into two categories: before it was clear that Mason Crosby was in a huge slump vs after it was clear that Mason Crosby was in a huge slump. The Packers seemed significantly more aggressive during the latter.

Points: 0

#4 by RickD // May 20, 2013 - 11:15am

Yes, Belichick punted from the Ravens' 34 in the AFC Championship game. The 34!

One play after the punt, the Ravens were at the 28.

Points: 0

#31 by herewegobrowni… (not verified) // May 20, 2013 - 8:18pm

Shurmur was big on plays like that both in '12 (@ Colts) and '11 (vs. Steelers,) and each decision arguably cost the game (although in the latter case, there's no guarantee that even Dawson makes a FG that long given the 50 mph winds.)

You could tell Shurmur also lost his job from Haslam's reaction in the Colts game (although it made sense to not fire him in the middle of the season.)

Points: 0

#5 by nat // May 20, 2013 - 11:33am

The correlation between the 2012 Aggressiveness Index and Overall Opportunities is 0.44.

Something is wrong with the index. Since it claims to be a rate stat, it should be independent of opportunities, shouldn't it?

Or am I missing something?

Points: 0

#24 by Revenge of the NURBS (not verified) // May 20, 2013 - 3:54pm

The coaches from the Saints and Colts only coached partial seasons last year, and they're all on the right side of the table. That could be part of it.

I also gather that there's some weighting of opportunities, that not all opportunites are created equal. Going for it on 4th and 15 at your own 15 is probably more aggressive than going for it on 4th and 1 at the opponent's 35. Notice #25 and #26, Gailey and Coughlin. Same number of opportunities, Coughlin went for it 3X as often, but the index rates them basically the same.

But even so, it's clear that there are a lot more guys with opportunites > 100 on the left side of the table than the right. It does look like there's some element of counting going on.

Points: 0

#35 by nat // May 21, 2013 - 9:42am

The correlation is much, much higher than last year (0.44 vs 0.13) so I think it's something added to the formula this year that caused this.

The answer to Aaron's question about the fate of Belichick's cojones might well be "Jim Armstrong cut them off with New Math".

Aaron: did Jim run the numbers with last year's formula? If so, how do the two approaches compare?

edits: Jim Armstrong did these stats, not Aaron; last year's correlation was 0.13 not 0.12)

Points: 0

#36 by nat // May 21, 2013 - 10:03am

To give a little more depth to the concern about this year's formula:

In 2010, the correlation of AI to the Attempt Rate was higher than that of AI to the absolute count of attempts. (0.68 to 0.56) That's as it should be, since AI is supposed to measure a per opportunity tendency to be aggressive (i.e. a rate), rather than be a cumulative stat.

In 2011, the same thing happened. (0.65 to 0.49) AI fit the attempt rate much better than it did the number of attempts. That's all good.

This year, this is reversed. AI is better correlated to the count of attempts than it is to the attempt rate, 0.59 to 0.55.

It would be very good if Aaron and Jim ran the numbers again with the old formula. I know it might be awkward - I don't know if the new numbers are carved in stone for upcoming publications. But it seems like this formula change may have introduced a cumulative feature into a rate stat.

This could also be a fluke. That's straightforward to check, although it takes a little work: Run the numbers for prior years to see whether the new formula tracks attempts or attempt rate for those years. But also, it may be possible to see the cause in the formula itself.

I love the concept of an Aggressiveness Index. If my concerns are justified, it would be great to get the index fixed for this year or at least to let us know the problem. If my concerns are unfounded, then there was something really interesting going on in 2012, and I'd love to hear more about that, too.

Points: 0

#28 by RickD // May 20, 2013 - 7:06pm

I don't think you can work backwards from a presumption of statistical independence to a conclusion of faulty data entry.

I don't see any obvious flaws in the table.

Points: 0

#32 by nat // May 20, 2013 - 9:53pm

No one said anything about data entry.

There are many possible explanations. Perhaps the method of selecting and weighting opportunities favors teams that are more aggressive. Perhaps the weighting was done wrong in some way. Perhaps the formula is partially cumulative. But it is curious, and could use an explanation.

Points: 0

#37 by RickD // May 21, 2013 - 2:28pm

I think if you look at the rate and the opportunity columns by themselves you will see correlation between opportunities and rate. I don't think the correlation is an artifact of the AI formula as much as it's there already in the raw data.

I would also like to see the formula explained in greater detail.

As for the change in the correlation from last season to this season, that might be partially explained by the change in how the data was compiled.

Points: 0

#38 by nat // May 21, 2013 - 2:47pm

Here are the correlations between opportunities and rate for the last three years...
2010: 0.08
2011: 0.10
2012: 0.48 (Wow! What happened here?)

A small correlation (e.g. 2010, 2011) makes some sense. After all, aggressive teams extend some drives, giving them a few more opportunities over the season. But this new, much larger correlation looks like something has been screwed up in the new formula. It probably has to do with how the opportunities are selected and weighted. But I'm open to other ideas.

Good thought on your part to point out the opps-rate correlation as being key. Well done.

As always with statistical analysis, this could just be a fluke. But I doubt it. Trying the old formula for 2012 and/or the new formula for the earlier years would tell us more.

Points: 0

#43 by Jim Armstrong // May 23, 2013 - 12:55am

To clarify how AI is computed, we computed league-wide averages for all situations, combining yardlines and, in some cases distances-to-go, where go-for-it rates were essentially the same. We divided each coach's rate by the league average for that situation, and then averaged all of these, weighted by the number of opportunities each coach had at each situation.

The reason for weighting by opportunities is to give proportionate credit to larger sample sizes. For example consider a coach who went for it 8 out of 16 times on 4th-and-1 and also 1 out of 2 on 4th-and-5. The former is slightly more aggressive than average while the latter is extremely aggressive (by rate alone). But I think the former tells you more about how aggressive he's been than the latter, and weighing the two situations equally would give too much credence to that one time he went for it on 4th and 5.

The issue noted with the correlation to opportunities appears to be due partly to adding the own-side of the field plays. It could be argued that these are weighed too heavily even at half-strength and are still overwhelming the results. Of the 25 qualifying own-side attempts in 2012, only 3 came from coaches ranked in the bottom half of AI. But then again, those are the most aggressive calls, right? In this sense, the formula has become more cumulative, perhaps too much so. But I wouldn’t necessarily agree that AI, simply by definition, should track closely to the overall rate. Also, there is definitely some flukiness to 2012. The correlations of AI to opportunities in 2010 and 2011 with the new formula are 0.27 and 0.05, respectively. And the correlations of opportunities to rate are 0.13 and -0.24.

It does appear that the teams with the most opportunities in 2012 tended to have a disproportionate number of them on their own side of the field. I’m not sure that makes them more likely to go for it there other than just having more opportunities to try it at least once. I also suspect that these teams tend to be worse overall, which would lead to more own-side opportunities, but they also may be more willing to take chances on fake punts.

I do think including the own-side plays does add value to AI, though that turned out to be not quite as straightforward as it seemed.

Points: 0

#44 by nat // May 23, 2013 - 8:02am

I think it's great that you're exploring fourth down aggressiveness.

I think my root problem with your approach is that in effect you are computing an AI for each play, and averaging those. As you found out, that begins to give weird results when you include situations where the average attempt rate is very low. You would not have had that issue had you simply compared a coach's attempts to the expected value for his mix of opportunities.

You attempted to fix the problem by discounting certain opportunities. But that factor is arbitrary, and can only be right by accident, like a broken clock is right twice a day.

In effect, you did this
   Average ( attempt1 / expected1, ... )

Instead of this

   Count( attempts) / Sum( expected1, ...)

The second is merely how a coach's fourth down attempts compares to what an average coach would have in the same mix of situations. I have no idea how to describe the first.

Points: 0

#45 by Jim Armstrong // May 23, 2013 - 9:56am

Thanks for the feedback. I did consider your suggestion at one point. Given the challenges of the expanded data set, I may give it another try now.

Points: 0

#46 by nat // May 23, 2013 - 1:37pm

See what happens.

I can understand the path you took. One of the points of FO-style analysis is to allow you to slice and dice the data many ways. That's easiest when there's just one number to keep for each play. To do what I suggested requires that you keep two: Attempt (= 0 or 1)and the league wide attempt rate for that situation (0 <= R <= 1).

Keep at it. This is very interesting stuff to look at. Thanks.

Points: 0

#6 by dmstorm22 // May 20, 2013 - 11:34am

Mike Mularkey was the Buffalo coach in 2004 and 2005, not 2003 and 2004, so either the years are wrong but the AI is right, or you're including the last Gregg Williams year.

Also, I was noticing BB's conservatism more this year, but not only for 4th down plays. End of the Ravens game, needing one first down to close it out, the Pats went run on first down twice in a row (before passing on 2nd, so it wasn't too bad). Then against the Cards they went really conservative in average field goal range. Against Seattle needing a first down or two to ice the game, they went run, run, pass.

The fact that there is no correlation between AI and success as a head coach just goes to show that game management is not the most important part of a coach's job. A coach that gameplans well but knows nothing of game management will do better than the reverse. The bad part for the coaches is the game management decisions are the low-hanging fruit for critics because they're decisions that are tangible to us as fans.

Also, I just love that Art Shell is one of the coaches to never go for it on 4th and 2.

Points: 0

#7 by BuccoBruce (not verified) // May 20, 2013 - 11:42am

How does that correlate with win percentage?

Points: 0

#11 by horn // May 20, 2013 - 11:52am

Well, with < 4 mins left from the article, the Eagles maintained possession, then went TD pass, Vikes ball, Eagles recover fumble, Eagles run it in for TD, Vikes ball, Gannon Int, and then Ruzek FG. So, it's pretty safe to assume there was virtually no time left and the Vikes had timeouts, and Buddy decided to kick on 3rd, up 5, so that if there was a muffed snap they could kick on 4th.

Coaches kick all the time on 3rd down at the end of games, it's not uncommon. Obviously, Randall's quick kicks were far less common.

Points: 0

#13 by Travis // May 20, 2013 - 12:13pm

There was 1:56 left in the game (the first Eagles TD came at 4:00, and the ensuing drives you give took 55 and 40 seconds), and the Vikings were (likely) out of timeouts (the play-by-play does not list them, but the listed game times indicate at most one stoppage after that point). The two-minute warning had just happened.

Points: 0

#9 by horn // May 20, 2013 - 11:45am

Perhaps a longer project, but I'd like to see coaches with 5+ years experience correlated with their Defense rankings.

Reid was more conservative with Jim Johnson because you could punt from the 42-yd line and all those years da Iggles had a top 5 or top 2 Defense, it made more sense because the Cowgirls or Jints were going to have a much tougher time going 80-95 yds to score than 58 yards.

[Yes, he still should have been more aggressive.]

Points: 0

#15 by justanothersteve // May 20, 2013 - 12:49pm

I was wondering if the same was true of BB. His early years with the Pats, the defense was pretty dominant and he could count on it to hold the other team even at midfield. As inconsistant as they've been these last few years, BB probably doesn't trust the defense like he did 10 years ago.

Points: 0

#10 by Travis // May 20, 2013 - 11:48am

Wyche almost always took the long field goal, going for it just 7.9 percent of the time.

This will change when you get to the late-80s Bengals, as Wyche barely trusted Jim Breech from over 40.

Points: 0

#12 by Some Yahoo (not verified) // May 20, 2013 - 11:53am

From just a glance, it seems like coaches with a background in defense seem to be among the highly rated. Has a correlation been done with that?

Points: 0

#16 by Briguy // May 20, 2013 - 1:45pm

I noticed this as well. Specifically, some defensive coaches who had a reputation for being overly conservative are pretty high up there (Dungy at 1.28? Really?)

Points: 0

#29 by Scott C // May 20, 2013 - 7:38pm

Yeah, #11 -- Schottenheimer. Classically fans wanted to fire him for a less 'conservative' coach.

Perhaps the defensive minded or those who start their careers in situations with strong running games and defenses lead to a false perception by the fan base.

Perhaps the constant critique of being too 'conservative' leads them to be more aggressive?

Schottenheimer doesn't really surprise me. He called for LT to throw (8?) times on fourth down -- 7 times for TDs! Every year there was about one fake punt and two to three trick/gimmick fourth down specials -- and he did not shy away from going for it on fourth and goal rather than kick a field goal.

Points: 0

#33 by dmstorm22 // May 20, 2013 - 10:13pm

Dungy learned from his mistakes and became quite aggressive in his later years (at least this is from what I remember - don't know what his year-by-year AI was).

I know some Colts fans who think it was his mistake of punting on 4th and 1 near midfield on the 1st drive of the second half in the 2004 Divisional loss to New England (down 3-6) that was the trigger for him to become more aggressive.

Points: 0

#17 by bravehoptoad // May 20, 2013 - 2:01pm

It's so cute that the Harbaughs have such similar AIs.

Points: 0

#18 by DOL (not verified) // May 20, 2013 - 2:25pm

Frustrating to see this year after year. "Aggressiveness Index" might be better labeled "Playing the Game Correctly."

Points: 0

#20 by peterplaysbass // May 20, 2013 - 2:52pm

Ha ha! That's why Jacksonville is number 1, right?

Points: 0

#22 by Aaron Brooks G… // May 20, 2013 - 3:25pm

What with Kotite, Haslett, and Coslet in the top-6? But of course Knox, Gruden, Billick, and Fox are idiots.

Remember that mindless aggression gets you high on this list, too.

Points: 0

#19 by Jeff S (not verified) // May 20, 2013 - 2:46pm

I think Brian Burke touches on this, but one reason we might see successful coaches go for it less is because better teams probably prefer "low variability" outcomes. Bad teams benefit disproportionately from "high variability" outcomes. The dichotomy faced by successful head coaches is interesting. On the one hand, since they have the better team, they are probably more likely to convert--swinging the "go for it" analysis even more strongly towards going for it. On the other hand, since they're the better team, they probably prefer "lower variance" outcomes--and simply win the game by being the better team--thus perhaps swinging the analysis towards conservatism. This is where a "team specific" 4th down calculator would help. But not practical due to the lack of data.

Points: 0

#21 by theslothook // May 20, 2013 - 3:00pm

This kind of echoes what some of the people have said above, but clearly this index isn't controlling for many of the potential factors that are likely affecting a coach's decision to be aggressive. I suspect a model that accounts for some lags in seasons past plus variables accounting for quality of opponent's offense & defense plus your own offense and defense. As it stands, its just too noisy imo.

Points: 0

#23 by Shattenjager // May 20, 2013 - 3:30pm

I think it would be interesting to combine this with Chase Stuart's game scripts and see if anything falls on your head out of it.

Points: 0

#25 by Karl Cuba // May 20, 2013 - 4:44pm

Does AI account for whether or not you're behind and in a rather desperate situation? It could explain why Jacksonville, NYJ and Oakland are all at the top. You stink ergo you go for it. Tu es enim ita foetent.

Points: 0

#27 by nat // May 20, 2013 - 5:12pm

The second paragraph outlines some excluded plays, obvious catch-up situations, for one.

Points: 0

#26 by Bjorn Nittmo // May 20, 2013 - 5:09pm

My quick read here is that the AI is based entirely on 4th down decisions, but I'd think it should also include end-of-half decisions, eg John Fox's indefensible kneel-down at the end of regulation in the Denver-Baltimore playoff game. Seems like the same measure of risk aversion as on 4th down, with conventional wisdom among coaches egregiously overweighting what could go wrong (turnover) at expense of taking advantage of favorable circumstances (your team has the ball).

Points: 0

#30 by herewegobrowni… (not verified) // May 20, 2013 - 7:53pm

As (nearly the only) resident Browns fan here, got to ask the obvious:

Shurmur 9th this past season?! Really?! Even his 2011 ranking is too high.

It's ironic that he is serving under Chip Kelly now, who is often implied to be the one to break the aggressiveness records (should he maintain a similar style of play-calling) and, incidentally, inspired some of Belichick's play-calling.

Points: 0

#34 by Jordan (not verified) // May 21, 2013 - 1:53am

When I looked at this I expected Schottenheimer to be in the bottom ten. I was too young to really remember his days as Chief's headcoach, but it seems "Martyball" might have been a little overstated. I'd love to see a breakdown of AI in playoffs as compared to the regular season. Was Marty more aggressive during the regular season as compared to the playoffs, and did that have an affect on his inability to make it in the postseason? Do some coaches become more emboldened during the playoffs as opposed to the regular season, and the other way around?

Interesting questions.

Points: 0

#42 by jonnyblazin // May 22, 2013 - 7:48pm

There are different types of aggressiveness though. Like if a team has a TD lead in the 4th quarter and the coach runs the ball up the gut three times and punts, that would not be aggressive. But it has nothing to do with this analysis.

Points: 0

#41 by jdawga (not verified) // May 22, 2013 - 2:51pm

Rex Ryan is higher than usual, but I think that's mainly due to Tebow - so Tebow did have an on field impact after all! Of the eight Jets 4th down attempts, I think at least 4 were fake punts with Tebow either running or throwing.

Points: 0

#47 by Curmudgeonland (not verified) // Jun 02, 2013 - 8:26am

Do you weight by clock time and score? I understand ruling out the plays in the last 10 seconds; but its not very agressive to go for it down two scores with 5 minutes left; that getting into desperation territory. Wiating until its too late to affect the outcome is not really being more agreesive. Ditto on the otherside of the equation; up late or with a big league, time may be more important than the possession. E.g., punting up two scores to give the other team 95 yards at the 2 minute warning vs. possibly giving them the ball at the (or 47 for a field goal) should not reflect on agressiveness.

Points: 0

#48 by nat // Jun 03, 2013 - 10:31pm

From the article:
The Aggressiveness Index excludes obvious catch-up situations: third quarter, trailing by 15 or more points; fourth quarter, trailing by nine or more points; and in the last five minutes of the game, trailing by any amount.
It's not weighting, but it deals with what you are thinking about.

Points: 0

Save 10%
& Support Aaron
Support Football Outsiders' independent media and Aaron Schatz. Use promo code SCHATZ to save 10% on any FO+ membership and give half the cost of your membership to tip Aaron.