Reviewing Points per Possession in the BCS Bowls

In the run up to bowl season, I previewed the BCS Bowls using points per possession measures. Points per possession and plays per possession seemed like theoretically attractive metrics because they are fairly intuitive, parsimonious, and seemed to capture parts of offensive and defensive efficiency, explosiveness, and field position in one convenient metric.

In particular, I was drawn to points per possession because of its parsimony. If variances in points per possession—just a single, easy to understand number—could explain (and hopefully predict using season averages) even half of a game’s outcome, then the metric would be valuable. Bill has already whittled down the top five most important stats for the college game, but it’s worth seeing how well points per possession performed by itself as well.

Points per possession on both offense and defense seemed to have decent initial explanatory power, as the BCS teams had not only great points per possession numbers, but were among the most formidable teams statistically (in terms of F/+ rankings) and put together some of the best seasons overall (in terms of winning percentages).

Points per possession is not opponent-adjusted (yet), only captures parts of efficiency, explosiveness, and field position, and there are plenty of other factors that determine whether a team with better points per possession numbers would actually win. However, it’s worth seeing if points per possession had any predictive value in the BCS Bowls. I wrote this earlier:

This two-way makes Florida State, Alabama, and Baylor look like lopsided favorites. Michigan State has a steep defensive edge over Stanford and is barely less efficient on offense (0.07 points per possession). The Ohio State-Clemson matchup is more of a toss-up, as Ohio State has a much higher average points per possession, but allows an average of 0.17 points per possession more than Clemson’s defense.

So two out of the three “lopsided favorites” were outright losers, while Florida State eked out a win over Auburn. The points per possession scatterplots were much more predictive for the Rose and Orange Bowls. Michigan State had a normally elite performance on defense despite missing the play of linebacker Max Bullough, while Ohio State-Clemson was ultimately decided based on turnovers.

Season averages for points per possession obviously weren’t perfect in predicting the winner of the BCS Bowls. That’s likely because, as mentioned before, the season averages were not opponent-adjusted, and numerous other variables factor in to a team winning a single game (turnovers and field position were particularly critical in the National Championship and Orange Bowl, for instance).

However, below are the points per possession scores for the BCS Bowl games alongside their offensive season averages:

Florida State, Ohio State, and Alabama all performed significantly worse than their average on offensive points per possession, while Oklahoma, UCF, and Clemson all outperformed their averages.