Countdown to Super Bowl XLVIII – Can Football Statistics Tell Us More?

I can’t help myself, I analyze everything: technology, exercise, electric cars, marketing of every ilk, airline disservice, and like many others, professional sports. This should be a really exciting Super Bowl – an irresistible force meets an immovable object! Given so much focus on the Super Bowl, this post won’t be about the next generation of technology but rather what I believe are suggested improvements in statistical measurement of a quarterback’s effectiveness. Since I asked several colleagues at Azure, including Dan Park (who helps edit every post), to offer opinions I will often say we throughout this writing.

In baseball, a game devoted to statistics, a hitter’s effectiveness was once measured by the combination of batting average, home runs and RBIs. But now there are several more sophisticated statistics like on-base average and slugging average that many feel better capture effectiveness. The traditional statistics in a football game have not been updated in the same way. They still primarily consist of passing yards, passing yards/attempt, sacks, interceptions, fumbles and passing touchdowns. To me, none fully capture quarterback (or offensive) effectiveness. For example, is it better to throw long passes and complete them less frequently (yielding higher yards/pass) or short passes that have a high completion percentage?  Should the offense favor more runs or passes? It seems to us that these types of questions show that traditional statistics miss the point of capturing effectiveness. Isn’t it a much more accurate measure of effectiveness to calculate how well a quarterback drives the team’s offense up the field? Of course, total yards for an offense is certainly a first step in measuring this. But it is insufficient for the same reason that total hits would be the wrong way to compare the effectiveness of two baseball players when one had 100 more at bats than the other (e.g. hits vs. batting average).

So we’d like to take a step forward by proposing that yards per possession is a much better measurement than total yards or yards per pass. What it captures is how far, on average, the team marches up the field. It doesn’t matter whether they run more, or throw more long passes or short passes if they can march up the field. Since most quarterbacks decide whether the play should be a run or a pass through audibles at the line of scrimmage, including both runs and passes in our measurement of a quarterback’s effectiveness takes into account the ability to read the defense and adjust plays (something Peyton Manning is known for). Let’s examine the 2 games played for the AFC and NFC championship.

 

While Manning is superior in every statistic, his yards/possession is the most striking number – his Denver offense moved the ball 63.4 yards/possession. It’s almost impossible to conceive of anyone achieving a much higher number, especially when one considers the fact that on the last possession he was killing the clock and thus potentially could have had more yards. Notice that Wilson had a lower average yards/play than either Kaepernick or Brady (and tied Kaepernick in yards per possession). Does this mean he performed worse than both?

This leads us to our suggested adjustments. Somehow, one must account for turnovers to better measure the quarterback’s effectiveness. If a team gets to punt the ball they usually move the other team back about 40 yards (2013 NFL regular season net punting average was 40.9) . One thing a turnover does is eliminate the ability to punt (or if they are close to the goal line, to score). Therefore, it’s as if the offense lost those 40 yards. Since a quarterback should not be held responsible for fumbles by running backs, we will make an adjustment just for turnovers directly by the quarterback – his fumbles and interceptions. We also feel it’s appropriate to re-calibrate the number of possessions. At the end of the first half, New England had a possession where they just took a knee (killed the clock).  We will subtract such possessions to calculate adjusted yards/possession.  Additionally, at the end of the game Denver ran out the clock (getting multiple first downs) and then took a knee. Since they did gain 38 yards before killing the clock, it seems accurate to count this as ½ a possession. In the table below, we have made similar adjustments for San Francisco and Seattle. The results are below.

The second new statistic that we believe measures a quarterback’s effectiveness is offensive points/possession. For this, we will only count possessions where the team was trying to score (7 for Denver and New England and 10 for San Francisco and Seattle). We don’t want to give credit to the quarterback if he failed to move the ball after getting field position that essentially guaranteed a field goal. Therefore, a field goal will not be considered offensive points if the offense moves the ball less than 20 yards and settles for a field goal. While our choice of 20 yards is somewhat arbitrary, it means that achieving 0 or 1 first down is usually a situation where the field goal would have been scored without the advancement. This means that two Seattle field goals and one San Francisco field goal do not count as offensive scores. This yields the following results.

The second new statistic that we believe measures a quarterback’s effectiveness is offensive points/possession. For this, we will only count possessions where the team was trying to score (7 for Denver and New England and 10 for San Francisco and Seattle). We don’t want to give credit to the quarterback if he failed to move the ball after getting field position that essentially guaranteed a field goal. Therefore, a field goal will not be considered offensive points if the offense moves the ball less than 20 yards and settles for a field goal. While our choice of 20 yards is somewhat arbitrary, it means that achieving 0 or 1 first down is usually a situation where the field goal would have been scored without the advancement. This means that two Seattle field goals and one San Francisco field goal do not count as offensive scores. This yields the following results.

What conclusions do we draw from this?  Too much credit is given to quarterbacks who perform at a mediocre level but have great defenses that allow their team to still win the game. Looking at the data, it seems apparent that Tom Brady played considerably better than either Kaepernick or Wilson. He lost the game because Denver scored 26 points and averaged an enormous number of yards per possession, thereby preventing New England from having more chances to score.  If one looks at the adjusted yards/possession, Manning is nearly 4X Kaepernick, yet the press often indicates that Kaepernick should be viewed as one of the elite quarterbacks in the league. It seems to us that both Kaepernick and Wilson benefit from their respective teams superior defenses, and that their teams win many games where they perform at a modest level.  Of course, one could also argue that Kaepernick and Wilson under-performed because they were playing against superior defenses.  Also, both of them are young and have potential to improve greatly going forward.

To gain more perspective, we decided to calculate adjusted yards/possession and points per possession for each of these quarterbacks in the divisional playoff games (which they all won).

We have not done the work, but we’re guessing the average points per possession for the league is about 2.6.  In the divisional games, Manning and Brady beat this by about 30%, Kaepernick achieved the average and Wilson fell short.  What this analysis suggests (but remember its based only on two games) is that both San Francisco and Seattle might be as successful with many of the quarterbacks around the league.  However, Denver and New England are much more dependent on their quarterbacks for success.  Of course we would need to extend this analysis to the entire season to have more certainty of such conclusions.
It will be interesting if Manning can retain this excellence in the Super Bowl going against the tough Seattle defense.  Based on these statistics, I would have to predict a Denver win.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s