Thursday, January 1, 2009

Imitation is the sincerest form . . .

Edited (see below).

I'm an avid reader of Yet Another Basketball Blog, since Dan Hanner is both a lucid analyst of college basketball, and a Georgetown season-ticket holder. A recent post there linked to a conference season-preview of the Big Ten by Spartan Weblog, by means of a tempo-free aerial pioneered (?) by Hawkeye Hoops. Also, Villanova by the Numbers posted a Big East conference review aerial for last season over the summer.

So, while we're just a few games into the Big East conference season, let's take a look at Big East aerial (click to enlarge). With most teams having played 10-15 games so far, the uncertainty of the stats is large, so any results taken from this exercise should be taken with an enormous grain of salt.

A quick explanation of the chart:
  • Adj. offensive and defensive efficiencies come from, and are through games played 31-Dec-08.
  • Note that the ordinate axis (defensive efficiency) is scaled backwards, from high to low, so that better defensive teams are at the top of the graph
  • I've color-coded the team markers, from black to yellow, indicating the nominal Pythagorean Rating, using Rating = AOEff^11.5 / (AOEff^11.5 + ADEff^11.5)
  • The solid grey line is a sort of 1:1 line, indicating where a team would be equally capable at offense and defense. Teams to the left or above the line are better defensively, while teams below or to the right are better offensively. Edit - I had this screwed up in the original plot - now corrected (I know, how hard is it to draw a 1:1 line).
  • The diagonal grey dotted lines are isopleths. Teams that are the approximately same distance from a line (e.g. Syracuse and Villanova) should be equally good, although they may do so by different means (here, Syracuse by offense, Villanova by defense). Edit - I realized after posting this piece that my isopleths were drawn incorrectly. I believe they are now correct, but because the isopleths have moved, I've amended the text below.

At this point early in the season, there appears to be several tiers of teams, but not many surprises:
  1. Georgetown currently is the class of the Big East, as KenPom has them ranked #1 overall in adj. offensive efficiency and #7 in adj. defensive efficiency.
  2. West Virginia (!), UConn and Pitt represent the three biggest threats to G'town's supremacy, with the Mountaineers slightly ahead of the other two. Pitt has a relatively balanced offense and defense, while UConn relies more on offense (as discussed previously here by SFHoya) and WVU is starting to take on Huggins' reputation for strong defense.
  3. Louisville (even after yesterday's loss to UNLV) is somewhere between the top four teams in the league and the next cluster afterwards. The Cardinals play the best defense in the league (and 2nd nationally) per, but struggle to score (12th). It turns out that being a great defensive team is better in Pomeroy's ratings than a great offensive team - I'll discuss this further below.
  4. Syracuse, Notre Dame, Villanova, Marquette and Cincinnati are grouped closely enough to be considered a third tier. The Orange and the Warriors/Golden Eagles are more dependent upon their offense, while the Wildcats and Bearcats rely more on their defense. In fact, you could say that Cinci is a poor man's Nova, and Marquette is a poor man's 'Cuse, but don't quote me on that. Notre Dame (like Louisville) represents a true extreme - the Irish have the league's 2nd best offense but a surprisingly porous defense (15th). I'd say that right now, the teams capable of making the NCAA tournament ends here.
  5. Finally, the gang of six that make up the have-nots in the Big East (Providence, Seton Hall, St. John's, Rutgers, DePaul and South Florida), with no extreme outliers in offensive/defensive balance. The Friars look to be the best of this bunch, but even they have a long way to go to make it up to the next tier.

With regards to the value of offense versus defense in KenPom's rating system, you may notice that as you head from the lower-left to the upper-right corner of the chart, the slopes of the isopleths changes from 45° (or -1, to be mathematical) to more horizontal slopes (or closer to 0). The contour line that G'town is next to has a slope = -2/3. This is a result of the Pythagorean Rating method - I won't go through the math here, but will be happy to explain in comments if someone is interested.

What this means is that, as a team moves into the upper end of the ratings, there is a larger return in improved defense rather than improved offense. You can see this if you pick a team, and then estimate how much it would have to improve on offense or defense to make it to the next isopleth. For example, for WVU to make it to G'town's line, they'd need to improve ~5 points in Adj. Off. Eff., or ~3.5 points in Adj. Def. Eff. Do keep in mind that I'm not talking about the real world here, just how to game's ratings.

One other point, that I hope the color-coding of the markers brings out, is that the difference between Georgetown (1st) and UConn (7th) in rating points is quite small (0.016), while the difference between UConn and Syracuse(30th) is larger (0.036), and the jump from Syracuse to Providence (79th) is much larger still (0.152).

As I said at the top, it's still early in the basketball season, and the underlying uncertainties for these stats are large - small sample size and differences in schedule quality foremost. I'd like to return to this chart in a month to see how all the teams have moved. If I forget, please prod me to do so.


Just an FYI that I've started posting HD Box scores for conference games. Right now, I've got 4 out of 5 posted, with the Syracuse / Seton Hall game a no-go since no substitution data was listed in the play-by-play. My first real test will be this Saturday (5 games) - I'm expecting the divorce papers on Monday.


  1. Nice--I hadn't seen this tempo-free aerial graph, but it's very informative.

    Maybe rotate it counterclockwise 45 degrees, with the 1:1 line running vertically up the middle? Then the isopleths would be closer to horizontal and Louisville would no longer look like the top team in the league. (Air out the y axis to make the isopleths parallel, maybe?)

    Anyway, what your chart captures--as kenpom suggests--is the amazing stength of West Virginia.