30 January 2014

Dusting off the CCR Statistics

You know that we at Carmody Court love us some efficiency statistics as much as the next geeky NU sports fanatic.  But while efficiency statistics are insightful and an improvement over some of the traditional statistics cited by coaches and talking heads alike (i.e., the dreaded rebound margin) the efficiency statistics aren't the statistical holy grail in our view.  For we know not all 1.00 PPP defensive efficiencies are alike.  Nor are all 1.00 PPP offensive efficiencies for that matter.  That is because these statistics don't take into account some very relevant factors such as opponent strength and where the game is played.

We first dug deeper into the efficiency statistics during the 2010-11 season and continued with the analysis for the 2011-12 season.  We shelved in our 2012-13 because it was rather meaningless given the trainwreck of a season that resulted from all the personnel losses.

With the 2013-14 season taking a turn upward and the NIT now a topic of discussion among NU fans we figured it would be a good time to dust off the analysis to see what we can glean about this NU team

We labeled this in-depth efficiency analysis the Carmody Court Ratio (or "CCR").  Essentially the CCR takes NU's PPP efficiency data from each game and adjusts them for (1) opponent strength and (2) home court advantage.  For the former we use the opponent's PPP data within B1G play and for the latter we convert the Sagarin home court advantage data into a PPP and allocate it 50/50 to offense/defense.   In so doing the hope is to put into better context what, say, the NU offense's 0.956 PPP at Wisconsin on Wednesday really says about how NU's offense performed in that game.


The first thing to notice about the NU defense's CCR statistic within B1G play (the "B1G CCR-D") is that the dotted trend line is significantly positive with a slope of 0.0424.  What that means is that over the course of B1G play NU's defense -- as adjusted for opponent strength and game location -- has improved from game-to-game by 0.0424 PPP.

For instance the trend line is about -0.1 PPP for the first game against Wisconsin.  That says when B1G play began that NU's defense was about -0.1 PPP worse than an average B1G team.  At the other end of the chart is the game at Wisconsin by which time the trendline is about +0.2 PPP.  That is over the course of 9 B1G games NU's defense has improved by 0.3 PPP.

The stat heads will be please to know that the analysis passes the F-test with flying colors and the t-statistic is significant at a 90% confidence level.

Of course this trendline can not continue.  If it were then NU's defense would be outperforming the average B1G defense by 0.6 PPP by the end of the season.  This is preposterous so at some point soon the trendline will flatten.  All that being said this analysis given tangible support for what we all know: NU's defense has turned a corner and is for real.

Since this is the first CCR analysis of this season we feel compelled to highlight some of the individual game statistics.  The first thing to jump out at us is that the narrative of NU defense suddenly turned on a switch starting with the Illini game is not so clean cut.  During the "pre-switch" days NU's defense did stink against Wisconsin and at Iowa, but it did perform like an average B1G defense at Michigan (+0.005 PPP).  Post-switch NU's defense has performed remarkably well, but there is the hiccup against Iowa in which NU's defense underperformed relative to the average B1G defense by 0.086 PPP.

So in summary it is true that NU's defense has improved significantly but it's not like we can expect there to not be any defensive hiccups the rest of the season (e.g., Minnesota's press may turn into easy Gopher points and wreak havoc on NU's defensive stats).


The NU offense's CCR statistic within B1G play (the "B1G CCR-O") is also noteworthy.  The nearly flat slope and statistical insignificance of the dotted trend line says that not much has happened with NU's offense during B1G play.  And the fact that that the solid line has been below zero for each of the nine games confirms what we already know: that NU's offense stinks.

NU's best offensive performance came at Iowa when it "only" underperformed an average B1G offense by 0.028 PPP.  The next "best" performance came on Wednesday at Wisconsin when NU underperformed by 0.069 PPP.  Other than those two games NU's offense has underperformed the average B1G offense by at least 0.135 PPP.  That's really sad.

 Overall B1G CCR

The story with NU's combined offense and defense CCR statistic within B1G play (the "B1G CCR") is similar to that of the B1G CCR-D.  What he see is a significantly positive and statistically significant trend line. That is heartening.

The biggest difference between the B1G CCR and B1G CCR-D is that the former's statistics are weighed down by the offense to such an extent that NU overall has played better than the average B1G team in just two games: @Indiana and @Wisconsin.  That "feels" right to us because while the wins over the Illini and Boilers were nice they just didn't jazz us up like the upsets at Indiana and even more so at Wisconsin. 

What's also interesting to us about this analysis is that the game against Iowa was the second worst of the year.  Only the opening home loss to Wisconsin -- of which we should never speak of again -- was worse.  That confirms why we were so downtrodden after both losses.

Lastly we note that the trend line has passed into positive territory.  This is consistent with NU's slightly above average sixth spot in the B1G standings.  Yes it's just a silly coincidence but we still like the narrative nonetheless.