With March right around the corner, teams that rank highly in the Top 25 are daring to dream of a magical run to the Final Four in Houston. Because the NCAA Tournament is so matchup-based, it makes sense that most observers can’t realistically pick their favorites until Selection Sunday. But what if there were data that allowed us to eliminate a few upstarts before we even saw the brackets? Well, there are ways to do that. One oddity of college basketball is how important and even predictive the preseason rankings are. Nate Silver uses them as one of the tools in his formula for picking winners, a fact that should nearly legitimize them by itself. But the argument behind it makes sense: Preseason rankings are a good way of measuring the overall roster talent of a team (because what else are we going to rank teams on before we see them play?), so teams that were ranked in the preseason Top 25 should generally be accepted as talented teams. But how predictive are they when it comes to the Final Four?
Using data stretching back to the 2003 Final Four, we looked at the average Preseason and Pre-Tournament rankings of every Final Four team. For the purposes of getting an actual number for an average, we changed “not ranked” to “35,” which was somewhat arbitrary but seemed about right given that several teams just missed out in the “also receiving votes” category, while others came totally out of nowhere (Note: if we had used a higher number, the averages and standard deviations would have both been slightly higher, but not much would have otherwise changed). Here are the results:
Shockingly, preseason rankings appear to be just as predictive of which teams will make the Final Four as pre-Tournament rankings. This makes us wonder why we even bother with that pesky regular season! But are preseason rankings truly destiny? Iowa, Xavier, Oregon and West Virginia all figure to be in and around the top 10 for the rest of the season, but none of that group were in the Top 25 when the season started. Could teams like these make the Final Four? The short answer, according to historical trends, is probably not. Since 2002, only four teams have ever made the Final Four after being unranked in the preseason but ranked after the regular season (this distinction is important, as it eliminates such Cinderellas like George Mason and VCU), and only one team has done it since 2006. In fact, since the 2012 NCAA Tournament, there have been just seven teams to achieve the rare feat of being unranked in the preseason but in the top 10 at Tournament time — interestingly, none of those teams has made it past the Sweet Sixteen.
The takeaway is to not count on upstarts like those four teams to make the Final Four. History is not in their favor. Every rule has its exception, though, and the one notable exception to this rule is represented in a 2011 Connecticut team which won the National Championship. We have learned that it’s probably safe to count on regular season’s surprises, but does that mean we should pencil in teams that were ranked before the season started but will be unranked at the end of the regular season? Recent history suggests those teams could be forces to reckon with. Since 2011 alone, Final Four teams Michigan State (2015), Kentucky (2014), Louisville (2012) and Butler (2011) all were ranked in the preseason poll but unranked going into the Tournament. The well dries up a bit when you consider teams before 2011, but that trend is strong enough to consider taking a look at the teams that are currently underperforming.
Unranked teams now that were ranked when the season started include: Gonzaga (#9), California (#14), Wisconsin (#17), Vanderbilt (#18), Connecticut (#20), LSU (#21), Butler (#24), and Michigan (#25). Not all of these teams are locks to make the NCAA Tournament, and the average preseason rank of 11.6 for Final Four teams makes Gonzaga and Cal the most realistic threats, but history suggests all of these should be monitored. Perhaps most importantly, though, when you are filling out your bracket in a few weeks, be sure to cast a glance back to the preseason rankings as well as the current ones. Over the years, they’ve proven equally important.
View Comments (2)
Hi, I appreciate the effort you make here to add an extra predictive layer to our bracket-filling decisions, but I won't be looking at the early November polls to make my mid-March decisions. The reason? The traditional powers who seemingly always manage to make the Final Four, even in supposedly off years (the usual suspects, we all know who they are) are also the teams that, almost by default, get ranked and very over-ranked in the preseason polls. So there's a correlation yes, in that the traditional powerhouses remain so for a reason--because they almost always comprise the teams competing in the Final 4. And therefore, they are blindly ranked by mostly guessing pollsters the following November. You have confused the cause and the effect here, and it makes a difference. The Final 4 predicts the preseason polls and not vice versa. Using this as a predictive tool in filling out brackets would be a poor substitute for watching games, understanding basketball, knowing who the usual suspect powerhouses are, and comprehending how the vicissitudes of the season in question inform those prior criteria.
Put another way: there is absolute no predictive magic lurking in the preseason polls. They merely recite the very conventional wisdom hanging over from the results of the prior season that exists at the beginning of the next. That the Final Four participants so very often reflect this conventional wisdom does not retroactively imbue the polls that merely restate it with any kind of predictive power because they are an effect of this cycle, not a cause of it.
So, coincidentally, I have already written an article this year that proved that the Preseason polls are not perfect because they overvalue the success of the previous year.
http://rushthecourt.net/2015/11/20/overvaluing-previous-years-success-in-preseason-rankings/
That being said, the picture is complicated. The Preseason polls may contain some fallacies and also be predictive of success to a certain degree, and the data suggests that they definitely are.