Those of you who are like me and have a reflexive desire to check Ken Pomeroy's ratings after a statistical bonanza like Kentucky's game with the Eastern Michigan Eagles last night, you might have noticed that Kentucky jumped five spots, from 12 to 8, based on a win over EMU at home.
There are a couple of things at work here that you should be aware of. One, KenPom's rankings are influenced from his pre-season rankings, which were based on last year's results, and that influence will not disappear until late January. That is probably having an impact on the ratings, but maybe not as much as you might think. If you are wondering why Pomeroy uses this extra factor in his early ratings, it is explained in some depth here (Warning: Statistical stuff. If that makes your head hurt, skip it). Pomeroy uses this weighting to minimize the wild swings we can sometimes see in the ratings -- minimize, but not completely eliminate.
Kentucky has made this big jump due to one thing -- defense. In terms of unadjusted numbers, Kentucky's performance on defense is even better than Pomeroy's figures. Unfortunately, I cannot back out the influence of the pre-season ratings (since we are now into January) so I can't tell if the defensive pre-season rating is helping or hurting. I can tell you that the offensive rating is likely to be helping some, but probably not a lot.
I can hear some of you saying, "But Kentucky's defense hasn't been that good against [fill in the name of your most upsetting loss or losses here], and you're right, . Kentucky has a pretty anomalous defensive profile right now. Against most of the weaker teams they have played, the Wildcats' defense has put up some staggeringly efficient numbers, but much less so against teams like Duke and Louisville.
"Wait," you plead, "Aren't Pomeroy's numbers adjusted for competition?" Absolutely, and for good reason. If all games were treated equally, a great game against a good team would count the same as a great game against a cupcake, and that would make his ratings a lot less accurate. So in order to make them more useful, Pomeroy adjusts his efficiency ratings for the level of the competition, so that a great game against a weak team, like EMU last night, does not overwhelm losses to better teams. The problem is, once the numbers get obscene, like against EMU, they hit hard.
Kentucky right now is an interesting case, because against weak competition, the Wildcats have been otherworldly good defensively, but against stronger teams they have been average or slightly below average. UK has not given up any particularly high defensive numbers, like the 123.6 (or 1.24 points per possession) it gave up against Indiana in the NCAA Tournament last year. UK won that game by being even more efficient than IU at 140 (1.4 p/p, a really strong offensive game). Essentially, they just outscored the Hoosiers.
At about 100 DE, or 1.00 points per possession, defensive efficiency transitions to "good" from "just okay." Note that this distinction is somewhat arbitrary, but not completely so. For example the average, or mean DE for Division I is 99.3 right now, but if this is like almost every other year, it will wind up between 101 and 102 at the end of the season. Right now, UK is allowing only 0.84 points per possession, 4th in the entire nation. It has four losses, but all of those losses with the exception of Notre Dame and Baylor have been very close, and in none of those games did the opponent put up great offensive numbers -- instead, Kentucky simply could not score.
Against the Irish, for example, UK managed an offensive efficiency of only 87.5, or 0.88 points/possession, and although Notre Dame is a good team overall, they are not a strong defensive squad, allowing a DE of nearly 95, 11 points worse than UK per 100 possessions. Notre Dame didn't need to defend the Wildcats, though, they defended themselves by shooting so poorly. Against Baylor, it was even worse -- UK could have built a new arena out of the bricks they laid in that one: 77.3 OE, or 0.77 p/p.
Just by way of explanation, let's look at Kentucky's six lowest-ranked non-conference foes last year vs. this year:
|2012 – 6 Lowest-ranked Non-Conf. Opponents|
|2013 – 6 Lowest-ranked Non-Conf. Opponents|
As you can see, the six "easiest" games for UK last ear were not appreciably more difficult than this year, but the defensive efficiency difference is substantial -- over six points per 100 possessions on average. Now, I realize averaging numbers like DE introduces a small error, but it isn't enough to worry about in this case.
What all this means is that Kentucky is defending worse teams better than last year, which isn't really surprising considering the fact that Willie Cauley-Stein and Nerlens Noel both play together a lot, and that much front-line height really shows up against shorter, weaker teams. That is evidenced by the fact that Kentucky is 5th in the nation in shot-blocking percentage.
The truth is, it is Kentucky's offense that is hurting them this year, not the defense, and this is a flip-flop from earlier in the season. As we have seen with the Wisconsin Badgers over the years, really powerful defense has a tendency to disproportionately affect Pomeroy's ratings. For an explanation of why this happens, please read this blog post by Pomeroy explaining and disclaiming it.
I actually don't think Kentucky will be like Wisconsin for most of the season, but they might, and here's why: UK is playing in an offensively challenged SEC this year, and that's going to result in some ugly DE's when UK gets hold of them, especially in Rupp Arena. We could see some DE numbers in the 50's against SEC teams, which will have a larger impact on the efficiency stats than even the EMU game did.
Teams that do well against Kentucky this year are going to be predominantly teams that can shoot the ball from the outside. See JLeverenz' post on how opponents have scored against UK, and his follow-on here about the Louisville game, to get a better feel for why this is. Kentucky is a very tough team to score on in the paint, but not so much in transition or from the outside, and all the teams that have put up big numbers from three against UK have either beaten them or been more competitive than they should have been.
The bottom line is that like all systems, Pomeroy's is flawed. I know many of you believe more in your eyes than you do in statistics, and I don't have a problem with that as long as you recognize that your eyes are deceiving you. Of course, those who believe in statistics to the exclusion of all else, or comparative won-loss resumes, are deceived also. There is no magic bullet to predicting basketball. You can lay odds with predictive systems like Pomeroy's, but this article is not about defending or rejecting either his system or the eye test, but rather an attempt to explain how a 4-loss team winds up ranked 8th in any rational system.
Kentucky is not the 8th best team in the country right now, in my humble opinion, but they are better than unranked. When I compare Pomeroy's rankings and my eyes to what I have seen this year, there simply aren't 25 teams better than Kentucky. The pure resume rankers would naturally disagree, but they, too, are deceived by their own adherence to comparative results to the exclusion of everything else.
Remember that when somebody asks you, "Who has UK beaten?"