Tempo & TOP per Drive in the 2012 Season

I have wanted to explore this topic more in-depth. Part of the impetus for this post comes from Dave Hudson's outstanding post on this topic at Football Study Hall. Another portion of my interest in this topic can be attributed to basketball and how Dean Oliver has made it possible to measure tempo by estimating the number of possessions played.

Drive-based statistics provide essential insight into team performance. Football cannot accurately measure tempo simply by accounting for the number of drives or non-offensive possessions played during the course of a game. FYI, I distinguish between drives and possessions. By my definition, the offense must run at least one play to classify a possession as a drive. All drives are possessions. Not all possessions are drives.

Why can't tempo be accurately measured by using the number of possessions in a game?

The answer to this question is surprisingly simple. A game played by two slow-paced with multiple three and outs will have a high number of possessions. Conversely, a game played with two up-tempo teams and very few explosive plays may not have as many possessions because each offense needs to run many plays before they can score.

Here is the formula I'm using to measure tempo:

• Tempo = (Plays / Total Seconds of Offensive Possession) * 1800

• I'm multiplying this by 1800 because it equals thirty minutes. Thirty minutes is a somewhat arbitrary number. I'm using it due to the fact that thirty minutes represents half of an entire game. I also like how it conveys the difference in tempo over tempo over an extended period of time.
• This only accounts for plays on statistically significant drives. 
• Time of possession is not measured in overtime. The statistics accumulated during overtime must be discarded.

I refuse to state that this metric is perfect. Based on what my observations, college football scorekeepers do not always accurately measure time of possession. 

The following chart provides tempo rankings for each FBS team in the 2012 season:

Here are what these numbers look like when grouping together teams by conference affiliation:

Revised QB Comparison Statistics

I released drive-based quarterback prospect statistics in February that provided drive-based statistics for each quarterback that seemed most likely to be taken in the NFL Draft. Matt Scott and Tyler Bray wound up going undrafted. Sean Renfree was selected late in the seventh round by the Atlanta Falcons. This revised version of the 2013 class and the comparison between the last three quarterback draft classes now include these changes.

However, the files containing drive-based analysis for Matt Scott and Tyler Bray will remain in the Dropbox folder.

Here are the revised versions of the aforementioned tables:

QB Prospect Comparison (2011-2013)

QB Draft Class of 2013

The Top 50 Offenses & Defenses Since 2009

These rankings have been compiled by converting the adjusted offensive and defensive efficiencies to z-scores. This means that some teams will finish higher than others in these rankings despite posting a superior adjusted efficiency because a z-score measures the number of standard deviations a data point is from the population's mean. For example, 2009 Georgia Tech had an adjusted offensive efficiency of 38.0 points per ten drives while 2011 Baylor's adjusted offensive efficiency 40.6. This discrepancy can be attributed to the fact that the average adjusted efficiency in 2009 was slightly lower than in 2011.

Some readers may notice that three offenses had a z-score that was more than three standard deviations above the mean while no defense was three standard deviations below the mean. The primary reason this happens is due to the fact that defenses perform much closer to their statistical ceiling than offenses. (The average defense is closer to allowing 0 points per drive than the average offense is to scoring 7 points per drive.) I'm going to use Stanford's offense in 2010 to help further illustrate why this occurred. The average adjusted offensive efficiency in 2010 was 21.9 points per ten drives. Stanford's adjusted offensive efficiency was 48.4 points per ten drives. That's 26.5 points better than the average. It's impossible for a defense to be that much better than the average because they can't allow negative points. 

The rankings for the Pythagorean expectation, adjusted offensive and defensive efficiencies, and opponent offense and defense reference the season in which the listed team played.

The Value of Starting Field Position

First, I have to link Football Study Hall. Bill Connelly added some extremely talented writers and shared boatloads of quality information. (Link) They also have the most thorough and statistically advanced 2013 team previews available. I couldn't possibly give this site a stronger recommendation.

The following five graphs illustrate the impact of starting field position on yards per play, percentage of available yards gained, percentage of drives ending in a score, percentage of drives ending in a touchdown, and points per drive. Field position statistics are available in the linked Dropbox folder. This data comes from nearly 60,000 drives during the 2009, 2010, 2011, and 2012 seasons.

The interesting thing about this graph is how rapidly yards per play declines when starting field position begins less than twenty yards away from the end zone. I personally think that yards per play provides more insight than other traditional metrics used in football. (Yards per game unquestionably has to be considered football's most useless statistic.) Starting field position should be mentioned more often when providing yards per play statistics because the condensed field leads to fewer explosive plays.

For whatever reason, the percentage of available yards was considerably higher for drives starting at the opponent's five-yard line than other comparable starting distances from the end zone. Here are a few reasons that may explain this somewhat counterintuitive finding:

1) Five yards is close enough to the end zone that the short distance makes it easier to score. However, it's relative distance from the end zone may help offenses by creating more space.

2) Too many teams still insist on running the ball up the middle in short distance situations, and this makes it easier to defend teams who prefer to jam the ball down the defense's throat when starting less than five yards away from the end zone. This strategy unquestionably works with the correct personnel. The reality is that most teams don't possess the offensive line of 2012 Alabama, 2011 Wisconsin, or 1995 Nebraska. In an area of limited space, spreading the field and making the opposition defend every square inch of turf may be the better strategy for teams without an offensive line capable of pushing the defensive line backwards.

3) Most likely, this can be explained by a limited sample size. Only fifty-two statistically significant drives started from this distance in the past four seasons. In other words, this result is nothing more than an aberration.

Scoring percentage obviously didn't illustrate a linear trend, but it came pretty close when drives began eighty to fifty yards away from the end zone. The fluctuation inside fifty yards may be due to the wide performance discrepancy among place kickers.

When a team gets possession of the ball within five yards of the end zone, they can realistically expect to score a touchdown about 84% of the time.

The average difference between one yard of field position is 0.05 points per drive. The following chart below illustrates how much net field position influences points scored per one, ten, and one hundred drives:

Keep in mind, the best net field position over the past four seasons was Boise State in 2011. They started 15.9 yards closer to the end zone than their opponents. I should probably remind readers that a negative number in net field position is considered a good thing because all net statistics on this site are calculated by subtracting defensive statistics from offensive statistics.

Four-Year Rankings

The Four-Year Program Rankings have been updated to include the most recent revisions.

This post contains Adjusted Efficiency Rankings for the entire FBS and each realigned conference. Conference USA, the MAC, and the Sun Belt could not be included in the conference rankings because those leagues have added former FCS members to their conference membership. Massachusetts, South Alabama, Texas State, and Texas-San Antonio were not included in the rankings because they have each been FCS members for one year.

• The ACC's conference rankings include Louisville, Pitt, and Syracuse. 
• The Big East's conference rankings include Central Florida, East Carolina, Houston, Memphis, Navy, SMU, and Tulane.
• The Big Ten's conference rankings include Rutgers and Maryland.
• The Mountain West's conference rankings include San Jose State and Utah State.
• The memberships of the Big 12, Pac-12, and SEC remain the same.