Behind the Numbers: Our Robot Overlords Think We Give UNC Too Much CreditPosted by KCarpenter on November 17th, 2011
Kellen Carpenter is an ACC microsite staffer and an RTC columnist. Behind the Numbers will publish weekly throughout the season.
Preseason rankings are a funny thing and once real basketball begins, these guesses about the season grow increasingly meaningless. People forget who was ranked where in the preseason by January, if they haven’t already forgotten by December. This, for what it’s worth, is probably a good thing. Preseason polls can turn out to be pretty embarrassing, highlighting how little journalists and coaches actually know about how the basketball season is going to turn out. Remember Kansas State and Michigan State last season? They weren’t exactly huge factors in the postseason despite being almost unanimously ranked in the top five at the beginning.
Gary Parrish’s “Poll Attacks” column is devoted to critiquing what he feels aren’t well-considered ballots. It’s an interesting idea and is usually very fascinating, but since Parrish can’t see the future, the columns have a tendency to not age all that well. What seems like understandable ridicule in November can seem less well-founded when the championship game is being played in April. Here is a particularly infamous passage from last year’s pre-season column:
Same dude voted Connecticut 18th.
This would’ve made sense three years ago, but it makes no sense now given that the Huskies have a coach who can’t seem to stay healthy, and a roster that looks nothing like your typical UConn roster. That’s why Big East coaches picked UConn 10th in the league, and why somebody needs to shoot me Ron Morris’ email address. With little effort, I can get him added to the official Big East email list, at which point he’ll start receiving announcements, and then this sort of stuff can probably be avoided. I don’t mean to be pushy.
Yes you do, Gary, but that’s okay. Connecticut struggled in Big East play and their national championship run was, in fact, pretty surprising. Who could have seen that coming? Well it turns out that poor Ron Morris did, but that doesn’t save him from Parrish’s scorn. That’s the dangerous and somewhat insidious side of the “Poll Attacks”: It’s an instrument used to shame those who deviate too far from the conventional wisdom, conventional wisdom that doesn’t necessarily have a great track record. The Internet is a wonderful thing, but the connectedness of journalists and the power of Twitter makes “groupthink” a very real danger. The echo chamber of ”conventional wisdom” dominates and influences voters, which turns out an often dubious preseason rankings list that has a powerful effect in shaping the narratives that guide the coverage of the season, which in turn strengthens the conventional wisdom even further. Before we start veering too strongly into media theory, let me make my point: Polls, particularly preseason polls, are flawed because they are influenced too strongly by the opinions of others.
So without the wisdom of the coaching or writing crowd to rely on, where do we turn? Why, to our robot overlords, of course. Ken Pomeroy’s preseason rankings, which he dubbed “My Crazy Uncle” in last year’s College Basketball Prospectus is a good place to start. Pomeroy’s regular season ratings are the gold standard for tempo-free statistics, but if you want to look at Jeff Sagarin’s rankings or Team Ranking’s wide variety of different power ratings, that’s all fine too. They roughly tell the same story: While humans have anointed North Carolina the best team in the country, our robot overlords are more skeptical, generally favoring Kentucky and Ohio State.
Why should you trust Ken’s Crazy Uncle? Well, as he notes in this year’s Prospectus, Pomeroy’s system did correctly predict the top seeds for last year’s NCAA Tournament with his preseason rankings where every single AP poll voter failed to get the top four right. This year, in order, the top four in his preseason rankings are projected as Kentucky, Ohio State, North Carolina, and Duke. This doesn’t seem that unreasonable, but considering the near-unanimous poll selection of North Carolina as the best team in the country, it’s odd to see that Pomeroy’s system (as well as just about any other computer-driven ranking system you can find) thinks that there are other teams that might be better.
Frankly, I think the machines are on to something. Let’s first talk about Ohio State, a team that was much loved by the computers (and me) during last year’s season. Is there some indication that this team has suddenly gotten worse? While David Lighty was certainly an integral part of the team, Jon Diebler and Dallas Lauderdale got very limited (though effective) touches on last year’s team. The Buckeyes bring back the best post player in the country in Jared Sullinger and his talented fellow sophomores Aaron Craft and DeShaun Thomas, not to mention a senior starter in William Buford. Thad Matta also brought in a talented freshmen class. This team was the best team in the country last year, and it’s hard to see them falling off much, if any.
For Kentucky, it’s much the same story. While Kentucky has had more roster turnover than Ohio State, they make up for it by bringing in a ridiculous surfeit of talent. The Wildcats lost a few key pieces in Brandon Knight, DeAndre Liggins, and Josh Harrellson, but John Calipari had no problem bringing in the best recruiting class in the country, while also convincing Terrence Jones, likely a top three pick in the NBA lottery, to come back for another season. A great team adds more (and maybe better) great players than it lost. Why should Kentucky be worse this year than last year?
Of course, you could make the same argument for North Carolina, saying that a good team kept its roster intact and added more pieces. The difference is starting position. Last year, Ohio State and Kentucky were both farther along than the Tar Heels. To put UNC ahead of these two teams, you have to project that UNC will improve at a much greater rate than either of these two teams. Why should that happen? What is it about North Carolina that says faster growth?
Oddly, a terrible beginning of the season may be the answer. Roy Williams‘ team looked rough at the beginning of the season, but the replacement of Larry Drew with Kendall Marshall and the simple process of Harrison Barnes‘ maturation led to an amazingly strong finish and a surprising Elite Eight finish. Now, if you were to graph the growth of the 2010-11 team and assume that the team would continue to improve at the rate it did last season, you might have a case for assuming North Carolina improves faster then others. For Ohio State and Kentucky, who were more consistently excellent from the beginning of the season, their apparent “growth rate” seems less dramatic.
Here’s a thought experiment: Imagine that North Carolina started last year’s season as good as they were at the end of the season. Why would people think that team would be better than an Ohio State or Kentucky team that consistently played better throughout last year? Sure UNC retains more of it’s core, but for the most part, that core didn’t perform as well as the players remaining for the other teams. North Carolina has benefited from a guiding narrative that, in the echo chamber of college basketball, managed to grow unchecked. The narrative was simple and often repeated: “Look how fast North Carolina is improving!” With that thought taken as a given, the inference of continued rapid improvement is easy to make. Suddenly, Roy William’s team was the near-consensus favorite to win the title.
Computer models don’t know about media narratives. They don’t have to listen to pundits talk endlessly about how many NBA-caliber players are returning to Chapel Hill. Pomeroy’s model looks at performance from the past, projects future performance given a new distribution of minutes and the impact of newcomers (specifically Top 50 recruits). This model, without hearing any spin, without the ability to get it’s feelings hurt by Gary Parrish, likes Kentucky and Ohio State better than North Carolina. Considering how clearly nearly every computer model slots North Carolina below the top spot, it’s maybe time that we all consider why we think the Tar Heels are better than Ohio State and Kentucky.