How strong is the relationship between offensive line play and team strength?

In recent years, the Seattle Seahawks have had a notoriously poor offensive line. In the past 5 years, Pro Football Focus (PFF) has ranked their line as 20, 27, 19, 30, and 32 out of the 32 teams in the league. At the same time, the team finished #1 in Football Outsiders’ DVOA metric in 4 of those 5 years, and #9 in the other year.

On the other hand, there are the Cleveland Browns, whose line has been a consistent strength of the team.  The Browns have finished in the top half of the league in each of the last 5 years and 3 top 6 finishes, per PFF: 5, 12, 6, 5, and 16 in the 5 years since 2012.  However, consistent performance has not translated to success on the field, and the Browns have consistently missed the playoffs and picked inside the top 10 in the draft.

So does offensive line play not really matter? Let’s take a look at the data over the past 5 seasons:

There are a few patterns that jump out here. First, there is a correlation between o-line play (as measured by PFF) and team strength (as measured by DVOA). If every team’s strength were completely driven by o-line play and had the same rank in line play and DVOA, then every point would lie along a line from the lower left to the upper right. While obviously that isn’t the case, there is clustering along that line. Second, Seattle and New England are clear outliers in the good team – bad o-line group in the upper left region of the graph (along with the 2012 Bears and 2013 Cardinals, elite defensive teams). Third, the good o-line – bad team group in the lower right consists of teams with awful defenses (2015 Falcons and Saints) and teams with no quarterbacks (Cleveland, 2013 Vikings, 2015 Cowboys).

Here’s a sloppily annotated version of the above graph:

Teams with elite defenses or QBs have the possibility of being good regardless of o-line play, and teams with disastrous defenses or QBs will be bad regardless of o-line play. For everyone else, teams with good o-lines are generally and and teams with bad o-lines are generally bad.

Why might that be the case? Here are some possible explanations:

  1. PFF graders know how good teams are and subconsciously give better teams better grades, regardless of actual play. For example, graders might be more lenient on offensive linemen on plays that are successful due to the contributions of skill position players, even if the linemen aren’t directly responsible for the play’s success. Unfortunately, no objective measures of o-line play exist so there isn’t an alternative to PFF grades.
  2. The organizations that are able to build good offensive lines are also the organizations that are strong at putting together other aspects of the team.
  3. Offensive line play is very important to team success.

My guess is that it is some combination of each of these factors, but I would love to hear others’ thoughts.

  • Post Corner

    Where would Denver 2011 fall? Everyone ignored that they had the 30th ranked OLine. Press also tried to prop up the 24th ranked DEF that gave up 40+ points 5 times.

    • Ben B

      They were #24 in DVOA, so they’d fall under bad team – bad o-line

    • Adam

      This speaks to point #1. The reason Denver’s line looked so bad was because Tebow held the ball forever and had zero pocket awareness. The parade of sacks taken by Tebow undoubtedly influenced the opinions of PFF graders. Funny how in 2012 with Manning, that line suddenly looked a lot better.

  • shubox

    This is great info. 2 questions:
    1. does the .3 correlation include the outliers? If so, what does that number increase once the outliers are excluded?
    2. do you have any insight on the correlation between RB success (fantasy points, yards, PFF grade etc.) and OL?

    • Ben B

      Thanks!
      1. Yes, it includes everyone. The correlation increases when looking at correlation with offense only (I think it’s something like .36) and would increase further if removing weird circumstances like a starting QB being injured for a season (don’t have the numbers in front of me).
      2. Unfortunately not, but that is a good suggestion!

      • shubox

        got you thanks for the quick reply.

        btw looking at the graph, i think the correlation would be pretty significant if you were to exclude NE and SEA and CLE (who pretty much tank on purpose) guessing it would jump to over .4