Filed under:

Data Deep Dive: Is Alabama actually passing more? And does it impact time of possession?

This is a follow-up, or more properly an adjunct, to Brent’s piece two weeks ago about whether scoring too quickly actually impacts the defense negatively. If you’ve not read that, check it out here.

Today, we look at two related questions: Is Alabama actually passing more, and, if so, is its passing offense negatively impacting the time of possession?

NB: This would be easier for some to visualize with a scatter plot, but the learning curve for the software that Brent and Parker ordinarily use is fairly steep. So, I’ve tried to clarify this with text and have included the equations for you instead — in the comments, I’ll drop the raw data as well. -EJE

The first step is to assess whether, in a given year, time of possession was significantly correlated within itself. That is, just because a team may have 30:32 of average possession one year and finish 59th in such category, that does not mean that the same numbers the next year would put it roughly in the middle of the country. So, we have to see if there is general correlation between that time of possession and whether it is correlated between final rank. There are always outlier years, and, as the game has gotten faster, what may have been quick possessions in 2010 may not have remained so for 2017.

To little surprise, over the past ten years, there was a statistically significant correlation between brute-force TOP and overall ranking in TOP. That this correlation is not P= 1.000 should tell you that there definite quirks and outliers from year-to-year. But, generally, this is a strong negative regression — as TOP rises, the ordinal ranking drops, which we would say is an improvement — “raising” from a 19th to 2nd ranking, for instance.

TOP x TOP Rank: 2010-2019

Calculation

R = CoVariance / (XRa St. Dev. * YRa St. Dev.)

Result Details

X Ranks
Mean: 5.5
Standard Dev: 3.02

Y Ranks
Mean: 5.5
Standard Dev: 3.03

Combined
Covariance = -77 / 9 = -8.56
R = -8.56 / (3.02 * 3.03) = -0.936

Outcome:
rs = -0.93618, p (2-tailed) = 7E-05.

Alabama Passing % x Passing Rank: 2010 - 2019

With that answered, we turn to the next question: Does the percent of pass attempts correlate within itself as an ordinal category? Like the TOP above, is there a general trend where passing plays represent an increased percentage of all plays — is the nation becoming more reliant on throwing the danged ball, in other words? If it is, is Alabama actually throwing more now than it was a decade ago as a percentage of plays — with more snaps, we can’t just look at the number of attempts.

Let’s start by telling you that over the last decade, only once has Alabama thrown the ball for at least 50% of its plays — 2019 (50.63). Only two other Alabama teams even hit 45% — 2013 (45.76%) and 2010 (47.04%). So, for all the grousing about Alabama being reliant on the passing game, that’s not really quite true in sheer numerical terms — Alabama is, and will likely always be, a run-first program.

Calculation

R = CoVariance / (XRa St. Dev. * YRa St. Dev.)

Result Details

X Ranks
Mean: 5.5
Standard Dev: 3.03

Y Ranks
Mean: 5.5
Standard Dev: 3.03

Combined
Covariance = -65.5 / 9 = -7.28
R = -7.28 / (3.03 * 3.03) = -0.794

Outcome:
rs = -0.79394, p (2-tailed) = 0.0061.

The end result of this one shows a lot more variance in the relationship between passing attempts and where that percentage ranks at the end of the year. So, while the general trend is to throw the ball more, there are weird outliers here. Perhaps, to put the numbers in some perspective, look at some real-world examples.

• In 2019, Alabama threw 409 times. That accounted for 50.63% of its offensive snaps, and the Crimson Tide were 44th in the nation in such attempts.
• In 2018, Alabama threw even more — 438 times, which ranked 40th in the nation. But, in terms of Alabama’s total offense, that was only 43.85% of all snaps.

In other words. Teams #RTDB more in 2019 than 2018.

So, yes, there is a relationship but it is nowhere as strong as the TOP numbers.

Time to put it all together. Is Alabama’s percent of passing plays related to its TOP or not?

In the 10-year span at issue, Alabama’s TOP has never been less than 30:10 (2019), and it has never been greater than 34:12 (2015 — Derrick Henry, duh). It has never been ranked above 4th in TOP (2013) nor below 62nd (2019).

We see just as much variation in passing offense — Alabama’s percentage of passing plays has been as low as 36.71% (in 2017, when Jalen couldn’t hit the Pacific Ocean standing on the North Shore) and as high as 50.10% (when Sark’s PAC 12 showed way too much.)

This will be an interesting analysis. So, let’s run the data.

THE BIG QUESTION: Passing Offense x TOP

Calculation

R = CoVariance / (XRa St. Dev. * YRa St. Dev.)

Result Details

X Ranks
Mean: 5.5
Standard Dev: 3.02

Y Ranks
Mean: 5.5
Standard Dev: 3.03

Combined
Covariance = -32.5 / 9 = -3.61
R = -3.61 / (3.02 * 3.03) = -0.395

Outcome:
rs = -0.39514, p (2-tailed) = 0.25842.

Putting it all together:

Surprisingly, there is only a 74% relationship between passing offense rank and where a team ranks in time of possession. This is in no way statistically significant.

This seems counterintuitive to be sure. Most conventional wisdom holds that the more teams throw, and the quicker they score or turn the ball back over, the less the possession time.

But, like many pieces of received wisdom, that’s simply not the case.

What we see instead in this modern era is that passing offenses can and are used just as effectively to control the clock as we would expect to happen when teams #RTDB for 44 minutes a game. Thus, it is the type and quality of the throws, the deliberation of the offense — even with more and more snaps per game — that is deciding time of possession, not a passing offense, per se. You can get there either way, to be sure. And four-wide can kill just as much clock, and be just as dominating, as lining up in the full-house.

Final note: Look, don’t submit this to your thesis defense committee or anything. This is a very crude, hamfisted look at the associations here. There are assumptions baked into the analyses, sample size issues, dozens of unaccounted-for independent variables, and the like. Instead, it should be used as a thought exercise and as a discussion piece.

More pass attempts don’t necessarily mean that teams are actually passing more. Not in 2020, not with the offenses designed to maximize possessions and the numbers of them. The better metric is, as Brent and Parker have noted in their respective coverage, how efficient an offense is: points per possession. For instance, “conservative” Jim McElwain actually threw the ball far more than Lane Kiffin did, as a percentage of the offense. But Kiffin’s offense was more efficient. And, of course, simple analyses like this don’t take into account game management — the need to run the ball while nursing a lead. Or, a coach trying to shorten the game as a huge underdog. Or running it down at half etc.

But, as a launching point? Have it.

And, along the way, we may just want to apologize to Lane Kiffin about running the ball more — he actually did.