Long before I had a fascination with advanced hoops stats, I wanted to build an objective ratings system. My initial experimentation involved making a system that looked good and conformed to public opinion in some way. That’s what I used for a while. The problem for me was that the basis of that system was subjective. Eventually I decided I wanted a system with a clear goal, to predict the scores of upcoming games. This throws the notion of “looking good” out the window. How the output looks is irrelevant. Changes can only be made to the system if they make it a better predictor.
When I shut down my original ratings system and switched to pure efficiency for the ‘05-‘06 season, I was concerned about the impact of margin of victory. However any experiments to limit margin of victory or to limit the effects of blowouts over weak competition resulted in a system that was less accurate on predicting future scores. Thus, I am left with a fairly simple algorithm: Pomeroy Ratings are essentially a pace-adjusted, schedule-adjusted list of average scoring margin, with further allowance for when the game was played.
(For those who might get the idea that this article is going to be self-promotion for my own ratings, please substitute “Sagarin Predictor” for “Pomeroy Ratings” from here on out. There are few substantive differences between the rankings of the two. Jeff’s a better marketer than I am and prominently displays his basic ratings on his web site, with the predictor version hidden on the right-hand side. The basic Sagarin Ratings conform to public opinion a little better. Indeed, I could make changes to my system to make it look better and not lose significant predictive value, but as you’ll see, I think something is lost in doing that.)
Which brings me to Kentucky. It has been brought to my attention that the Wildcats are the top-ranked team in the Associated Press poll this week. I wouldn’t know since I haven’t checked out the AP poll in years, but apparently every single one of the writers with a vote believes that the Wildcats are the best team in the land. (Honestly, I can’t believe that if you gave some of these writers truth serum, a few of them wouldn’t admit to believing Kansas is better. But then again, maybe that’s not what folks in the AP poll are voting on. Perhaps that’s why I don’t pay attention to it--I have no idea what the poll is trying to tell me.)
The top ranking in the polls is an obvious contrast to UK’s position of number 9 in the Pomeroy Ratings as of this morning. Many have remarked that this seems a tad bit low, and you might wonder why the Wildcats are treated so poorly by my system. In fact, the occasional e-mail comes my way asking just that. But really, I’d like to turn the question around to our analytical-minded audience.
Why would a rating system that relies on margin of victory, and generally does a good job portraying the relative strength of teams around hoops nation, fail with Kentucky?
I think there are three possible responses here. First, you can accept that Kentucky is overrated by the media. Early reputation-building wins over North Carolina and UConn don’t look quite as impressive as they did in early December. And those are conceivably the best wins that the Wildcats have to date. SEC fans can look to Arkansas last season and Vanderbilt two years ago as teams that racked up a lot of early wins over lesser foes but suffered what seemed to be surprisingly cruel fates in conference play. To defend your stance, though, you have to ignore the fact that scouts think three of Kentucky’s starters are going to have an impact in the NBA--in a little more than nine months.
A second option is to avoid the question altogether. Clearly, the system is out to lunch on Kentucky. Therefore, something must be wrong with the system. End of discussion. Something like this (from last February).
Personally, I prefer a third option. Accept the premise of the question for a moment and address it directly. In order to do that we need to understand why margin of victory is an important element of a good prediction scheme. While a team plays to win the game, its chances of winning are much greater when it leads by 20 with five minutes to go than if the game is tied at that point. There is a significant incentive to building a comfortable lead rather than just trying to stay a point ahead of the opposition all game long. You never know when your opponent is going to go all Chandler Parsons on you. For this reason teams capable of building big leads typically build them, and past results indicate that those teams are headed for good things in the future.
This year’s edition of Kentucky has a history of not putting away weaker opponents. Normally that's a sign of bad things to come, and in this case it's the direct cause of UK's lower-than-expected rating. In recent contests against both Georgia and Auburn, the game was tied with eight minutes to go. A week ago the Cats were tied with Florida with five minutes left. Essentially, the Wildcats have been willing to throw away the first 75 to 85 percent of the game. Playing what amounts to a five- to eight-minute game against a less-talented opponent is normally a dangerous thing when done regularly--it gives the more-talented team less time to prove its superiority.
So far, Kentucky hasn’t been burned by this pattern of play. They’ve had some close calls, sure, but they’re still unbeaten. The question is whether UK is headed for a reality check like other teams that have repeated close calls against weak competition, or whether they are truly different. I am open to either possibility at the moment, but I would submit that they could be different. To illustrate why, let’s look at the only team from the tempo-free era that I feel certain was different.
In 2006 Gonzaga had arguably the best offense in the country. And Gonzaga knew they had the best offense in the country. If points were needed, Adam Morrison usually found a way to get them. Any number of offensive moves combined with excellent perimeter shooting allowed him to post a sick 120 offensive rating while using 32 percent of the Bulldogs’ possessions. And when all else failed, a wild flailing of the arms would guarantee Morrison a trip to the free throw line if the ‘Zags needed it. (In addition, J.P Batista wasn’t a bad second option.)
This kind of effectiveness led to an interesting behavior: Gonzaga only played defense when they absolutely had to. Against most WCC foes, this meant not having to play defense at all. But against UCLA in the Sweet 16, this meant having to play defense a lot. It wasn’t until Morrison walked off the court crying after the tournament collapse against the Bruins that I was convinced a team could violate the laws of hoops physics and be stellar without beating up on unspectacular competition.
Normally, I’m not one to let one game sway my opinion. But were it not for some seriously bad luck in the final minute on that Thursday night almost four years ago in Oakland, Gonzaga would have been in the Elite Eight with only Memphis between them and the Final Four. This despite a Pomeroy rating that languished in the 40s most of the season. And that explained to me exactly what happened to the Zags in the previous 32 games. Their defense wasn’t great, but it wasn’t as god-awful as it graded out during the season, either. They rarely felt like they had to play defense.
However, I don’t think there’s been another team quite like that Gonzaga squad. There’s a case that last season’s Oklahoma team showed similar behavior at times. It’s easy to see the resemblance between the two--both had a big-time scorer and free throw machine, and reserved defensive effort for special occasions. However, whatever mojo the Sooners had in close games was gone in March, even when Blake Griffin returned from a concussion.
So is Kentucky more like the ’06 Zags or the myriad other teams that are ultimately exposed after a lucky run against weaker competition? Of course, it’s impossible to say with certainty. To speculate though, I think one can concoct an explanation for why this year’s Wildcats might be like Gonzaga. You have a collection of talent led by guys who’ve been told they are great for nearly their entire lives. Thus they may well believe they have enough talent to win five-minute games. Heck, they may be so bored with the weaker competition that they want to play five-minute games just to challenge themselves. In my mind that is plausible, although I’m not convinced of it either.
We’ll get another data point in this interesting experiment with tonight’s game at South Carolina. And the Wildcats should provide a more definitive response to this issue in the back end of an SEC schedule that contains much tougher road tests than the one in Columbia. In the meantime, I hope the Prospectus audience will keep this issue in mind as they watch Kentucky over the next few weeks. The Wildcats aren’t likely to go away and neither are questions about why margin-of-victory based systems don’t like them as much as humans do. However, the possible answers to those questions can provide additional insight on what might be a unique team.
Ken Pomeroy is an author of Basketball Prospectus.
You can contact Ken by clicking here or click here to see Ken's other articles.