If you’ve been following along here for a while, you know I’m a big fan of the week six AP poll. Over the last 20 years, week six has outperformed previous AP rankings in predicting which team will cut down the nets in April. In fact, every national champion since 2004 has been ranked in the top 12 of that season’s week six AP poll.
So, without further ado, welcome to week six:
1. Louisville 2. Kansas 3. Ohio State 4. Maryland 5. Michigan 6. Gonzaga 7. Duke 8. Kentucky 9. Virginia 10. Oregon 11. Baylor 12. Auburn
Conversely, every subsequent poll from here to Selection Sunday is roughly equal to week six (or even worse) in terms of identifying which team will win it all. Each year in early December, it seems, we stand astride something of an evaluative continental divide.
The current season gives us a perfect illustration of why a poll that comes out now has a big advantage over preseason rankings. Consider the top 12 from the AP preseason poll:
1. Michigan State 2. Kentucky 3. Kansas 4. Duke 5. Louisville 6. Florida 7. Maryland 8. Gonzaga 9. North Carolina 10. Villanova 11. Virginia 12. Seton Hall
That’s not a bad top 12, all in all. The Spartans happen to be 1-2 in close games and Duke walloped MSU in East Lansing, but, obviously, it would be foolish to write off Tom Izzo in December. Meanwhile, Florida’s 6-3 and currently ranked down in the 20s or even 30s by reputable laptops in the possession of Ken Pomeroy and Bart Torvik. But those same laptops were high on the Gators in the preseason, just like the pollsters. The future’s tricky that way.
At least by week six, though, we have more information than we did in October. Specifically, we now know that, regardless of their eventual tournament fates, Michigan and, especially, Ohio State have far more potential than we realized in the preseason. This epiphany happens most years by week six with regard to an exceptionally small but highly visible population of teams.
Still, if “more information” were all there were to spotting title favorites, we’d expect later polls to outperform what we’re looking at this week. In theory, we should be adding more information and achieving a richer understanding of these teams with every game that’s played. There’s still plenty of basketball left to dissect after week six.
True enough in theory, but we now come to the other side of the so-called continental divide as it relates to the AP poll specifically. By this point in the season, ranking the teams inhabiting the highest reaches of the top 25 has become largely though not entirely an exercise in rewarding teams that stay undefeated the longest. (Repeat, not entirely. Kansas, I salute you.)
“Staying undefeated the longest” does indeed share some overlap with “teams that win it all.” See for example Virginia last year (the Cavaliers and Michigan were the last undefeated teams, and both lost on the same day) or Villanova in 2018 (the Wildcats shared “last undefeated” status with TCU and Arizona State, and all three lost the day before New Year’s Eve).
But, those two examples notwithstanding (pay no mind to that recency effect behind the curtain!), this particular two-fer had been recorded just four times in the previous 40 years: Florida 2006, UConn 1999, Duke 1992, and Kentucky 1978. Not to mention fixating on the last undefeated team can, once in a great while, lead us badly astray. Ask Clemson sometime what its seed was in the 2007 NCAA tournament.
It turns out that, given the notorious vagaries of November and December schedules, an even better predictive trait than a perfect first couple months is being really good at basketball. We may know more about that trait by early December than we give ourselves credit for.
This point was brought home last week when aforementioned friend Ken felt obligated to point out not once but on two separate occasions on Twitter that, yes, early-December KenPom ratings do indeed correlate pretty well with final KenPom ratings, the final AP poll, you name it. Ken’s point can be exported profitably to less analytic climes.
For instance, the week six AP poll translates well in terms of picking national champions. Specifically, week six beats the pants off of previous polls.
Average AP ranking of eventual national champion by week
Rank Preseason 9.0 Week 2 9.0 Week 3 8.5 Week 4 7.1 Week 5 7.1 Week 6 6.1
After week six, every subsequent poll from week seven to the end of the regular season ranks the eventual champion somewhere in between 6.0 and 6.9. In effect, defining the population of likely national championship contenders happens early and doesn’t take long.
I’ve taken the liberty of illustrating this process using the very latest in data visualization software….
Syracuse 2003 broke our handy little top-12 streak (the Orange were nowhere near being ranked in week six) because Jim Boeheim’s team made its “bottom 300-something” to “top 12” migration late in the season. Carmelo Anthony’s team didn’t even crack the top 25 until mid-January. That’s not necessarily unusual: Michigan charted an eerily similar late migration in 2018. What is rare, however, is for such a late-moving team to win it all. Usually we’ve spotted these upwardly mobile title contenders already by week six, the canonical examples still being Florida in 2006 and UConn in 2011. Both were unranked in the preseason but had vaulted into the top seven by week six.
Nevertheless, our tendency is to doubt if not belittle any intimations of early-season knowledge, with good reason. I don’t know about you, but my team is currently in the bottom 341. Maybe they can make it into the top 12 if they will just — finally! — change these two key things that I keep yelling about every time I watch them. Alternately, even if our team’s in the top 12, as fans of the sport at large we tend to buy into narratives of dispositive change. Surely, Virginia and North Carolina will figure things out on offense, right? Florida is going to get things sorted out eventually, no?
My team really could get better as the season goes on, and so, too, could the Hoos, Heels, and Gators. Other teams are learning and correcting and improving too. The question is how all of the above teams will fare in a zero-sum competition with opponents that are also adapting. For what it’s worth, history suggests we already have a pretty good answer to said question by early December.