Analysis: Using KenPom.com's Game Plan page

When you look at the Game Plan page for Georgetown, below the table for individual game performance is a list of correlation coefficients for offensive and defensive efficiency versus various statistics, such as pace, effective field goal percentage, turnover percentage, etc.

Here's the correlation table, as of Feb 18, 2007:

                       Correlations
                  to OE       to DE
        Pace:  -0.41       -0.20

        eFG%:  +0.72*      -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%:  -0.26       +0.57*
     Opp TO%:  -0.17       -0.39
     Opp FTR:  -0.32       +0.29

Bold  values are significant with a 95% confidence
Bold* values are significant with a 99% confidence

For Georgetown, Off. Eff. is well correlated with Eff. FG%, TO %, and pace, in that order. Def. Eff. is well correlated with Opp. Eff. FG%, Opp. Off. Reb. % and TO%, in that order.

Let's look at one of these correlations in more detail (click on image to enlarge).

Here's a plot of Off. Eff. vs. Pace for Georgetown for the season so far. You'll note that the R^2 value in the plot is equal to the square of the correlation in Ken's table (-0.41 x -0.41 = 0.168). You'll also notice that there is a pace outlier for the dataset, namely the Vanderbilt game, with 73 possessions. For the season as a whole, Hoya games average 58.5 possessions, with a standard deviation of 4.6 [58.5 ± 4.6]. At ~95% CI (2 st. dev's.), the range is 49.4 - 67.7 possessions, so only the Vanderbilt game stands as an outlier.

If we remove it from the data set and re-run the numbers, the new plot looks like:


















Well, a couple of things jump out here. The correlation is much stronger (r = -0.547), and the slope of the line is much steeper (pace has a larger effect). If you remove the Vanderbilt game from the data set and re-run Pomeroy's entire table, you'll find that the OE vs. pace correlation is the only one that shows a significant change (i.e. other than pace, the game was typical for Georgetown):

                        Correlations
                  to OE       to DE
        Pace:  -0.55       -0.27

        eFG%:  +0.72*      -0.09
         OR%:  +0.36       -0.22
         TO%:  -0.61*      -0.44
         FTR:  +0.13       -0.10

    Opp eFG%:  +0.20       +0.77*
     Opp OR%:  -0.26       +0.58*
     Opp TO%:  -0.17       -0.39
     Opp FTR:  -0.34       +0.30

Bold  values are significant with a 95% confidence
Bold* values are significant with a 99% confidence

Well, this is all fine and good, but I'd like to take the analysis of pace effects a bit further, hopefully later this week.