Are we headed over the quality cliff?

It’s been nearly a month since I checked how the KenPom Top 20 quality curve compared to the curves of the sanest tourney in 2007 and maddest one in 2011. For those who haven’t read the previous posts, there’s a connection between the quality of the top teams and tourney unpredictability. The year when the elite squads were more efficient than their historical counterparts correlated to chalky dances, while those years when they were less efficient aligned with upset-laden tourneys.

The 2011 tourney saw the worst top 20 teams in terms of KenPom efficiency since the data became available in 2004. Conversely, 2007 featured the highest quality top 20. And what happened? The 2011 dance tied a record for upsets (13). Meanwhile, 2007 broke records for predictability, with just three upsets.

Last time I did the quality curve comparison, Florida rated out as the best team of the last nine years. Schools ranked second through 15 were slightly above average, and those between 16 and 20 were below average. The curve has undergone a dramatic change since then. Basically, what we are staring at is a quality cliff.

Take a look at how the red line compares to the faded blue (average), orange (best field) and blue lines (worst field). I’ve built this chart in three phases to demonstrate the quality cliff:

quality_curve_2-23-13

The top three teams—Florida, Indiana and Louisville—are solidly better than average. In fact, the Gators and Hoosiers are among the best of the last nine years. Then, from the fourth best team (Gonzaga) to the 13th most efficient squad (Ohio State), the quality curve eerily follows the average curve.

And that’s when we fall off the cliff. The quality difference between Ohio State and Arizona is profound. The Buckeyes are actually slightly better than your average 13th best team. The Wildcats are much worst than even the worst 14th best team. And from there on, from Colorado State through to Virginia, the curve tracks along with the historically worst—and most unpredictable—tourney.

How big is this cliff? Consider these two points:

  • As the 14th best team, Arizona would figure to get a four seed. In most years, their Pythag would’ve been good enough for just a six seed.
  • As the 16th best team, Oklahoma State would also get a four seed. In 2007, their efficiency rating would’ve earned them an eight seed.

Let’s face it: if the teams rated worse than 13th by KenPom efficiency data don’t pick things up, we will be looking at some historically bad four and five seeds. Let’s assume that the seeding breaks down perfectly by KenPom data—an assumption that never comes to pass…but bear with me. If this were to happen, here’s the dynamic we’d be staring at in our brackets:

  • Two tops seeds would be historically strong (Florida and Indiana), one would be better than average (Louisville) and the last (Gonzaga) would be slightly weaker than average.
  • The second and third seeds would all right around average.
  • Ohio State would be a slightly better four seed than average—and the other three teams (Arizona, Colorado State and Oklahoma State) would be awful.
  • San Diego State, Minnesota, Georgetown and Virginia would all be among the worst five seeds of the last nine years.

So what would all this mean for building your bracket? My guess is that we’d see one or two 4v13 shockers and a couple of 5v12 upsets. But with the unusual strength of top seeds, you wouldn’t see many second or third seed upsets on their side of the bracket.

On the two seed side of the bracket, I’d predict that it would play close to historical seed match-up probabilities. Three two seeds would reach the Sweet 16 and a couple three seeds would face them there. And the winner would face a strong one seed in the Elite Eight.

The bottom line is that we could see a tourney where there are lots of upsets in round one, maybe just a couple in round two…and then relative chalkiness from there on in. Of course, the vulnerability of the three through six seeds depends on the quality of the teams seeded six through 13. That’s essentially the 21st through 52nd best teams in college ball. After Selection Sunday, I’ll do a curve of the actual teams at their actual seed positions and we’ll have a better picture of how the tourney may play out.

This entry was posted in Measuring Madness. Bookmark the permalink.

2 Responses to Are we headed over the quality cliff?

  1. Paul A says:

    According to KenPom, eFG% TO% OR% and FT% are the four key factors to picking winners. Have you ever run the numbers to see if there is any truth to this? Also, I was jotting numbers down on my computer (i am a geek. i stare at spreadsheets and look for correlations for a long time) and noticed that in the last couple of dances that a team that has a SOS, both O and D combined, to be greater than 5 pt difference more then their opponets, was a great predictor of success (this seemed to only occur with 1, 2, 3, and some 4 seeds). It got me thinking, and here is where you come in. Is there any correlation between a team having a significant advantage in one stat or a certain order of stats (maybe one of the big four) and success in that game? Example, Team A has a FT% of 4pts higher then Team B, is succesful X amount of times. Hope I made my point clear.

    • ptiernan says:

      I do think those factors are important, and I’ve seen some correlation between TO% (mainly on defense) and overachievement against expectations. But I have tracked enough years to draw any conclusions. Not sure I understand the calculation on SOS, O and D. Are you talking OE and DE rankings? So a team with an SOS of 40, an OE of 10 and a DE of 20 would score 70?

      If you’re an Insider, you can see how teams do in head-to-head games across 20 or so stats. It’s in “Toss-up Picks” under the Tip+ menu.

Leave a Reply

Your email address will not be published. Required fields are marked *

* Copy This Password *

* Type Or Paste Password Here *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>