Every now and then, Purple Reign will grab a handful of box scores, a few cups of coffee and a calculator. Then, after hours locked in a room, the hope is that patterns will emerge.This happened a few weeks ago when PR took a look at Mick McCall's two-year tenure as an offensive coordinator – first at Bowling Green in 2007, then last season at Northwestern. The numbers at first didn't reveal much. Both the '07 Falcons and '08 Wildcats scored a bunch of points, both had pretty productive quarterbacks, both ran the spread.
But eventually, after going down statistical dead ends for a few hours, the numbers began making sense – at least a little. For example, Bowling Green QB Tyler Sheehan threw the ball 36.6 times per game in 2007 under McCall. In 2008, NU quarterback C.J. Bacher threw it 37 times per game. Interesting…What's more, it turns out those two quarterbacks had wildly different passing numbers with different coordinators. There were other similarities, and you are more than welcome to take a gander here.
Well, this week the coffee pot and calculator were again in overdrive at Purple Reign. The goal: To better understand Northwestern's third down numbers. This became an area of interest when PR looked at the numbers and realized that third down has been a good down for the Wildcats. This season, yes, but also in 2007 and 2008.Right now, NU ranks seventh in the nation in third down conversion percentage. The Cats are converting third downs at a 57.45 clip, 27-of-47. Nos. 2 through 6 are all between 60.61 percent and 57.50. Thus, the Cats are about three percentage points from being the second-best third down team in the nation. In other words, if they had converted 29-of-47 – instead of 27 – they'd be No. 2.
The No. 1 team, Florida, is completing two-thirds of their third downs, and the Gators are way out in front of everyone else.*
* We're only one-quarter the way though the season, and obviously three games is a small sample size. And obviously having Towson – poor, poor Towson – as one of those three skews things a bit for NU. But a lot of teams have had cupcakes on the schedule – Florida, for one, has played Charleston Southern and Troy. So everyone's sample size is small and somewhat skewed. But because Northwestern did this last year, and the year before that, PR didn't think it too early in the season to talk about this.As was mentioned, this is old hat for the Cats.
In 2007, NU ranked 19th in the nation in third down efficiency at 45.8 percent; that was No. 13 among BCS schools. The No. 1 team – once again Florida – was at 53.4 percent; No. 10 Hawaii was at 47.5 percent.In 2008, the Cats were 21st – 10th among BCS conference teams – converting 46.41 percent of their third downs. The No. 1 team, Tulsa, completed 56.99 percent, and No. 10, Rice, completed 51.1 percent.*
* It's interesting that the curve for third down efficiency went up by a not-negligible 3.59 percent. Maybe the proliferation of the spread helped that. Don't know. But 53.4, which led the nation in 2007, would have been good for just fifth in 2008.** Surely teams' current 2009 numbers will go down as conference play starts – and that includes Northwestern. But it will be interesting to see if the trend of higher conversion rates continues.Northwestern's back-to-back top 21 finishes in third-down efficiency – coupled with the hot start this season on third down – is all the more impressive when you consider who else appears that high on the list in 2008 and 2007. The other teams that were No. 21 or higher both seasons:
NavyThat accounts for five Heisman finalist in 2008 and 2007 – OU's Sam Bradford, MU's Chase Daniel, UT's Colt McCoy and UF's Tim Tebow (twice). Oh, and every one of those teams – besides Navy – was ranked No. 1 in the nation or receiving first-place votes at some point the last two seasons.
This of course means that there were a lot of good teams that weren't on there – USC, Ohio State, LSU, Alabama, etc.Moreover, there is only a two-team list that was 21 or above in 2007 and 2008 and in the top seven this season: NU and Florida.
So, NU is good on third down.With that as the premise, Purple Reign looked more closely at third down numbers from 2008 and 2009. It was '08 and '09 because: Those are the two most recent seasons, those are the two seasons with McCall as O-coordinator, doing a third season might have taken me until Week 7. (There is a bit about '07 below, just not quite as much. No one is trying to remember '07 too much anyway, right?)
Don't know exactly where to start with the numbers that PR crunched, so let's just dive in…In 2008, the Cats had five games where they converted at least 50 percent of their third downs. Pretty good, right? Well, what's interesting – and utterly counterintuitive – is that four of those five games were losses.
Against Michigan State – a 37-20 Week 6 defeat – NU converted an outlandish 62.5 percent of its thirds, the best single-game mark of the season. However, that number is a bit deceiving. While the third down conversion rate was high, the overall third down efficiency – not so much. C.J. Bacher threw two interceptions on third down.So while 62.5 percent is incredible – 10 for 19 – nearly one-third of the unconverted third downs resulted in turnovers.
Another good third down game that resulted in a loss was against Indiana. NU converted at a 52.6 percent clip, the fourth-best single-game number on the season, but still lost to an utterly inferior opponent. Once again, there were too many third down turnovers – a pick and two fumbles, both lost.Yet another good-conversion, bad-result game came in Week 10 against Ohio State. The Buckeyes let NU convert 53.3 percent of its thirds, but still blew out NU in Evanston.
One common denominator about the Michigan State and Ohio State games, where NU's third-down success belies the results, is that NU trailed big early. Northwestern was in a 24-7 halftime hole of both games, so maybe OSU and MSU took a "keep everything in front of us" approach, which allowed NU to rack up first downs but prevented them from scoring. Just a thought. (No such thoughts on the Indiana game. That was just a mess.)What's also interesting about the MSU and OSU games is that NU's average to-go distance on third down was not good, yet the Cats still converted well over half the time. Against MSU, NU's average to-go number was 6.29 yards, the eighth-longest average on the year. And against OSU, the number was 7.1, the 11th-worst average.
And the "real" to-go numbers* weren't all that good either: an even six yards against MSU, 6.4 against OSU. Still, in both games the Cats converted more than 53 percent.
* Thoroughly going overboard with this third down stuff, I calculated first for "average third down distance." That's simple enough – the average distance-to-first-down number for the entire game.
But the "real" to-go distance is a little different, and in my opinion, it's a better stat. The "real" number is the average distance-to-first, minus the longest third down situation. Then, of course, you divide not by the total of third downs, but by that total minus one.
The philosophy behind this is that one really long third down can have a drastic effect on the average to-go number, but a negligible effect on the conversion percentage.The idea for a separate "real" distance stat was borne of the Indiana game. You see, NU faced 19 third downs, but one of them was a 25-yarder. Thus, that 25-yard third down attempt had a significant ballooning effect on the average third down distance, but an insignificant impact the conversion percentage. It moved the conversion percentage from 55.5 to 52.6 – a five percent difference. However, the sum of the third down yardage goes from 77 to 102 – a 33 percent difference.
NU didn't get a first down on the third-and-25 play, which counted for 1/19th toward the conversion rate. But those 25 yards were much more significant to the average yards stat – way more than 1/19th. Thus, against IU, NU's average to-go number on third down was 5.37 yards. But its average "real" to-go number – with that inflating 25-yarder taken out – was 4.28, a full yard-plus less. Make sense?
Basically, PR figures the "real" number is more instructive for the team's true average to-go distance because it helps eliminate the effect of huge outliers like third-and-25. In most cases it didn't make too big of a difference. Anyway, that's why there is a separate stat for it. (Henceforth, the average to-go number and average "real" to-go numbers will be presented together like so: Avg to-go/Avg "real" to-go.)Moving on…the fifth-best third down game, against Missouri, was also a loss. NU converted 52.4 percent of its thirds and had a season best to-go distance of 5.29/4.55.
PR was pretty stunned that four of the top five third down conversion games were losses. And the one that was a win, a 56.3 percent effort, came against an overmatched Southern Illinois team.
Maybe this would look better is graph form, yeah?
|Opponent||3rd down %||3rd down % rank||Result|
Or we could put it in order, from worst first down game to best.
|Opponent||3rd down %||3rd down % rank||Result|
What's also interesting about those 2008 third downs…
* The best single-game conversion rate, against MSU, came in a game with the eighth-highest to-go (and eight-highest "real" to-go) average. And of course that season-best rate also came in the team's first loss. Indeed, the Spartans game is a baffling one when trying to decode the significance of third down efficiency: On average, NU had its longest to-go numbers, yet still converted more of them. (And still lost.)
* An interesting '08 tidbit for this week: The worst single-game conversion rate came against Minnesota, which was a Northwestern win. That horrid 20 percent conversion game – two for 10 – was weird all around: Mike Kafka set that rushing record, Brendan Smith returned a pick at the buzzer (?), NU won despite not scoring a single second-half point on offense. What's more, NU also faced to-go averages of 10.5/10.0.
* And Minnesota wasn't the only "bad third down" game that NU won. At Michigan, the Cats had a 7.59/6.94 to-go average; both were the worst on the season. And The Cats' 35.3 percent conversion rate was pretty bad, too – the 10th worst game in that department.Yet NU still won by a touchdown. In Ann Arbor.
So these 2008 numbers are pretty odd. Consider: Of NU's 13 games, the eight worst in terms of third-down conversion percentage were all wins. And of the remaining five – the "top" five – four of them were losses.Thus, the percentages, it seems, lie. They lie like a rug, like a politician, like a kid trying to avoid getting grounded. But if the conversion rate doesn't tell us anything, what does?
How about the number of third downs? This, it turns out, is a far more instructive stat. (At least for 2008…see below.)OK, so in '08, NU lost pretty much all its uberefficient third down games in terms of efficiency. But look a little closer at those games, and you see this: Of the four losses, you'll find NU's three highest totals of third down attempts.
So the percentages were high, but the frequency of third downs was also high. And this – in '08 – was telling of a loss.Northwestern had 24 third down attempts against MSU, 21 against Missouri and 19 against Indiana. (Against Ohio State there were 15 attempts.)
While the stat in the box score always reads, Third Down Efficiency, a more telling number for the Cats in 2008* would have been, Third Down Frequency. The average number of third down attempts in NU's nine wins was 14.4. The average number of third down attempts in NU's four losses was 19.8. Percentages be damned.
* This stat doesn't at all hold true for 2007. In its losses, Northwestern had 15.2 third down attempts per game. In its six wins, Northwestern had 16.5 attempts. That's not only a somewhat negligible difference, but it's also backward from what we saw in 2008.The Cats' third down pattern so far this season has followed a more predictable trajectory. In the Cats' first game, the bloodbath against Towson, NU converted 70.6 percent of its third downs and faced a reasonable to-go average of 6.23/5.56. That was the team's most convincing win.
Against Eastern Michigan, another win – albeit far less convincing – NU converted 56.3 percent of its third downs and faced a 6.1/5.8 to-go average.Two 56-plus percent outings, two "real" to-go numbers of less than 5.9 yards, two wins. You can guess what happened against Syracuse: The Cats converted a more pedestrian 42.9 percent of its third downs, and their average to-go and "real" to-go numbers were through the roof – 8.1/6.85. Naturally, the Cats lost that one.
Basically, Purple Reign's study of the third down numbers has revealed that this statistic is incredibly deceiving. History shows that NU's better third down games didn't always result in wins. Although more recent history shows that NU is winning its better third down games.Obviously, you want to have a good third down conversion rate. You want to be in the top 21 in the nation in this category, up there with Oklahoma and Florida and Texas Tech and Texas and the rest of the them. This much is true, and the Cats are in elite company with these teams – No. 1-ranked company, Heisman-winning company.
Still, for all NU's success on third down, it's obviously not the key to success. There are other factors in play, like turnovers – which often come on third down – and defensive productivity and yards per carry and a whole bunch of other things.So, are third downs important? Yes. Does Northwestern's place in the pantheon of good third down teams bode well? Absolutely. But it's obvious that the Cats can't simply rely on their third down productivity to wins ball games. If they could, they'd be BCS-bound.
Not that being good on third down is a bad thing.
To reach David Vranicar, publisher of Purple Reign, please write to firstname.lastname@example.org.