Analysis: What is determining performance?

A couple of earlier posts had looked at Ken Pomeroy's work correlating Offensive and Defensive Efficiencies with pace and/or the "four factors," and determining how well Georgetown performed in each game versus the expected result, based on opposition.

I thought I'd try to bring these two ideas together, to see which component(s) of Georgetown's offense or defense are determining how well the Hoyas actually play.

I'm going to call the point difference between actual points scored and expected (via Pomeroy methods) performance. So, if I plot performance over the course of the season:



This looks like a nice improvement as the season has progressed (ignore that leveling off from the last two games), although that early season win at Vanderbilt is a bit anomalous. Now I don't want to get into residual analysis here, but, for shits and grins, let's pull the Vandy game out and re-run the plot:



The slope of the line increases about 20%, and the R^2 is up 40%. Hopefully, someone videotaped that Vandy game and will write up what the team did right.

Anyway, now that we can see how much the team has improved, I thought I'd try to find a correlation between performance and any of the underlying statistics that Pomeroy provides.

First, let's re-post KenPom's table or correlations between Off. Eff. / Def. Eff. and components, now updated through 2/21:

                  Correlations
                 to OE       to DE
         Pace:  -0.40       -0.18

         eFG%:  +0.71*      -0.09
          OR%:  +0.35       -0.22
          TO%:  -0.60*      -0.44
          FTR:  +0.13       -0.09

     Opp eFG%:  +0.20       +0.77*
      Opp OR%:  -0.25       +0.55*
      Opp TO%:  -0.17       -0.37
      Opp FTR:  -0.32       +0.28 

These work quite well, but the question is whether these same correlations hold when we're looking for correlations to performance rather than efficiency. If it's simply a question of subtracting Off. Eff. from Def. Eff., you'd expect that eFG%, TO% on offense, and eFG% and OR% on defense would be the keys.

Here's what I get:




These are some interesting results. Overall, Off. Efficiency, and less so Def. Efficiency (or opponent's Off. Eff. in this case), are well correlated, as we'd expect. The especially strong correlation between the offense and performance shouldn't be a surprise. Georgetown has the most efficient offense in the country, so you'd expect that to be the primary weapon.

What's more interesting is what some of the contributing underlying factors are: namely eFG%, which we expected (thank you Roy Hibbert and Jon Wallace), but OR% rather than TO%. So it looks as if the team is able to overcome the high turnover rate, while good rebounding effort is still important. Defensively, only eFG% has any sort of strong correlation, which is a real surprise to me - I expected the defensive rebounding (opponent's OR%) to be very important. Hmmm.

The only other note is that, if we ignore the Vandy game, forcing the opponent to play slower than they like like (pace% = game pace / opp. normal pace) seems to help.