2014 College Football Preview: Experiments with DSR

Well, pointy football teams not named Rough Riders (or Roughriders, or Eskimos, or Blue Bombers) are starting soon. Indeed, the uncompensated labor force edition of the sport started last night, with Georgia State eeking out a win in a mostly barren Georgia Dome against Wheeler High School. My own beloved Georgia Tech opens up Saturday hosting Central Gwinnett High school.

Of course – this year comes the College Football Playoff where the final four will be determined by a select committee. While a Final Four is an improvement over the Bowl Championship Series – going to a committee is still relatively weak sauce. Given the general snail-like pace of college football games, I am not sure how people with real jobs (even if the real job is uttering platitudes for five figures per reading) and families have that sort of time. In any case, the data analysis exercise that was the BCS rankings is still interesting, and honestly that this system does not include some sort of computer ranking is a failing – even to identify the 8 teams which will be considered for the Final 4.

Needless to say, the data analysis exercise still interests me.  We’ve experimented with Pairwise Rankings (like the college hockey system) and applying the Analytic Hierarchical Process.  This year, we go to Football Outsiders for some inspiration, and the notion of Drive Success Rate (DSR).  What is interesting about DSR is that it is a bit of a spiritual cousin to offensive efficiency in the NBA (which I have covered before).  On a very basic level, what is the object of football?  Like soccer or field hockey, the goal is to get the ball from one end of the field to the other.  So we want to know how successful a team is at doing that.  However, football has a couple of wrinkles.  First, the teams take turns with the ball – and you have 4 downs to move the ball 10 yards.  So the goal of traversing 100 yards is more manageable.  What DSR does is measure how successfully you turn 1st and 10 into a first down or a touchdown.  That’s it.  This does not mean that field goals are not important – they are, but except in very specific circumstances, most of the time a field goal is a failure.

So how is DSR calculated?  It is easy enough to be calculated intutively:

  • Numerator: Every first down a team gets is a successful series of downs.  Every touchdown is also a success.  So success = first down + offensive TD
  • Denominator: Start with the number of drives – add first downs (since each first down begets another).  attempt = drives + first downs – kneeldowns.  We remove kneel downs since there the offense is not attempting to convert a first down.  Of course since I am doing a lot of games and a lot of drive charts, I am not guaranteeing perfection here, but that is the goal.

Since we have a game on record, we can show it as an example:

Abilene Christian: 26 first downs, 4 passing TDs, 0 rushing TDs = 30 … 26 first downs + 13 drives = 39 … drive success rate = 76.9%

Georgia State: 33 first downs, 4 passing TDs, 1 rushing TD = 38 first downs … 36 first downs + 13 drives = 49 … drive success rate = 77.6%

Needless to say, this was not a defensive struggle.  In one analysis of NFL results, it seemed that teams with above average DSRs for both offense and defense (the latter meaning low percentage) tended to win a LOT.  So this year’s experiment will use net DSR as a metric to determine team strength.

This inspires an obvious questions – aren’t special teams important?  Of course they are!  But good special teams results will make drives harder, and score points.  That comes out in the winning.  Turnovers will be baked into the winning too – and there is considerable debate as to whether recovering fumbles are an actual skill or not.

So we will use Net DSR (Offensive DSR – Defensive DSR) normalized by schedule (the opponents offensive and defense DSR against other teams) to identify team strength, and use it to normalize win/loss record.  By week 4 we’ll have some ideas on how it will look.

Advertisements

One thought on “2014 College Football Preview: Experiments with DSR

  1. Great stuff – and I just looked at differential of points per drive differential (points per drive scored minus points per drive allowed) from last year…and no surprise who the top 2 teams were: #1 Broncos, #2 Seahawks.

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