What a difference three weeks makes to the college hoops balance of power

Among the college hoops media, teams only seem to be as good or bad as their last couple games. (And, ironically, as I proofread this, Boston College shocks Syracuse.) Even fans paying close attention don’t tend to look much further in the past than the last AP top 20 ratings—at least until Selection Sunday.

If you think the balance of power in college basketball changes a lot from one week to the next, however, you might be surprised at how it shakes up over the course of three weeks. I sure was last night when I stumbled across an Excel sheet from January 28 ranking the KenPom top 20.  It made me wonder about two things:

  1. How have the efficiency numbers for today’s top 20 changed over the course of three weeks?
  2. What has happened to the individual teams that made up the top 20 on January 28?

Let’s answer the first question. Take a look at the chart below. The orange line represents the Pythag values of the 20 most efficient from three weeks ago. The red line represents today’s KenPom top 20—and the schools along the bottom occupy those 20 slots.


The first thing that jumps out is how much the overall efficiency of the nation’s best teams has dropped in just 22 days. Every single position is a notch lower—and teams ranked 12th through 20th are considerably lower than their past counterparts. Moreover, the second best team of January 28, Creighton, was playing more efficiently than the best team, Arizona, is today. In fact, on January 28, you had 15 teams, ranked from second to 16th, who were bunched together in the efficiency gap between today’s top 11 schools.

Let me put that another way. Three weeks ago, there wasn’t a sizeable drop off in team quality until the 16th ranked team. Today, the quality drops off at the 12th team. And, oh, by the way, the overall quality of the top teams has generally declined.

(NOTE: If I compared the quality curve of today’s KenPom top 20 to past years, as I’ve done in previous blogposts, these numbers would look shockingly low. And I would therefore conclude that we’re in for a crazy 2014 tourney, since there’s a positive correlation between the relative strength of the top 20 and the chalkiness of bracket outcomes. However, you’ll notice I’ve stopped doing those comparisons. The reason is that this year’s KenPom numbers are derived using a different formula than previous years. Ken acknowledged that he’s removed the bias of mismatches from his efficiency ratings—and that’s driven overall values lower. So the old pre-tourney Pythag values I’ve provided aren’t exactly comparable to this season’s Pythag numbers. There’s no way around it: we’re dealing with apples and, well, slightly different apples. The bottom line: it’s difficult to say whether this years best teams are weaker than their historical counterparts or not. Throw in the impact of the new offensive-friendly rules on efficiency numbers…and we’re essentially in uncharted statistical territory.) 

As surprised as I was to see how the KenPom top 20 had sagged in terms of efficiency over the last three weeks, I was even more stunned to see how the quality of those individual teams had changed since January 28.  Check out this next chart. The orange line represents the Pythag values of the AP Top 20 from three weeks ago. And those old efficiency leaders are listed along the bottom. The red line represents what their Pythag value is today.


Look at how dramatically the efficiency values of the erstwhile top 20 has changed:

  • Some teams, like Louisville, Virginia, Villanova and Kentucky, have seen virtually no change in their overall efficiency.
  • Other teams, like Arizona, Creighton, Syracuse, Wichita State and Kansas, have only seen a small degradation in their efficiency.
  • Another group has suffered a more significant drop-off, including Iowa, Wisconsin, Michigan and Ohio State (Hey, those are all Big Ten teams!)
  • And then there are the major decliners: Michigan State, Pittsburgh Oklahoma State (yikes!), Florida State (yikes! part two) and Iowa State.
  • Finally, you have Florida and Duke, the only two teams who’ve made substantial improvements in their overall efficiency

I have three takeaways from this analysis:

  1. A lot can change in the quality of a team over the course of five or six games.
  2. By downplaying mismatches in his new algorithm, KenPom’s Pythag values are based on less data and thus prone to wider swings. Ken admits as much. That’s at least part of the reason that the fate of some teams have changed so drastically in three weeks.
  3. There are 25 days until Selection Sunday. Today’s KenPom top 20 are liable to see a similar range of changes in their quality between now and then.
This entry was posted in Tourney Trends. Bookmark the permalink.

7 Responses to What a difference three weeks makes to the college hoops balance of power

  1. Greg says:

    Great post! It’s unfortunate that we can no longer rely on the absolute numbers of past champions, but the offensive and defensive rankings should still hold true, right?

    What the heck is going on in the B1G? I also was thinking about how quickly things change when I noticed that OK St. dropped out of the Top 25 this week.

  2. jbessa says:

    By season’s end the volatility should start to decrease in kenpom (as he has stated) and overall the ratings should be more accurate. For some teams, like Wichita St or SDSU that have played weaker schedules we could still see some movement after they play in the NCAA tournament.

    I like to look at kenpom’s luck ratings. With a high luck rating, kenpom is basically saying that he doesn’t think the team is as good as its record (see Cincinatti). The opposite for low luck rating (see Iowa). Surprising to see an undefeated team like Wichita St have a neutral luck rating with an undefeated season.

    Is this the year Alford breaks through with a deep tournament run? UCLA surging and defense playing much better.

    • Andy says:

      I absolutely agree with you regarding luck. Too often, the teams with either the most amount of luck or the teams with the least amount of luck always end up failing to meet expectations (though don’t tell that to Norfolk State, #1 in that category in 2012 when they stunned Missouri). Look at Florida from last year, all year they had a lockdown as the #1 team on kenpom.com, but they could never win a close game and as a result their luck was around -.100 or so. Their draw was an easy one featuring lots of teams that either did not play in the previous year’s dance, were reliable underachievers (like Georgetown), or were led by snake-bit coaches. So really, there was no Elite Eight alternative. But once they got there and faced someone halfway decent….yeah, you saw what happened.

      It’s that bad luck rating, combined with a few other factors, that has me not so high on not only Iowa, but also Louisville and Kentucky. Heck, for what it’s worth, if Tennessee gets in, their raw Pythag numbers suggest a possible Cinderella, until you see their awful luck.

      Meanwhile on the other side, I agree on Cincinnati being weak, but that is also due to their already weak numbers for a favorite coupled with a dismal strength of schedule. I’m sorry, but this team may very well be knocked out in the first round if they draw the right team to pull one off. Same with San Diego State. Villanova also has too much luck working for them. That said, some of the models suggest they are a Final Four contender….then again I can’t shake this feeling that they may be an underachiever when the brackets are released. Is it the fact that when they lose, they get blown out? I dunno.

      I guess my theory is this, teams either too lucky or too unlucky never make the Final Four. But overly lucky teams might be able to pull off an upset, maybe even two. But overly unlucky teams? Forget about it. (Incidentally, on another recent post I actually suggested Vermont as a possible Cinderella since they nearly won at Duke and have decent Pythag numbers for an underdog, but overlooked that they too have really bad luck. We’ll find out in a month I guess, assuming they get in first.)

  3. David says:

    Question for Pete or anyone who can answer… I ‘ve started to read the seed guide and other articles in the Tips+ section. For many of the qualifiers/disqualifiers it says a team that has won ‘X’ amount of games in their past ten pre-tourney games has a certain record. Does ‘pre-tourney’ mean the past ten games before March Madness or the past ten games before their respective conference tournament? Thanks

  4. Louie says:

    Is there a “grain of salt” number anywhere that reconciles margin of victory with strength of schedule? I’m thinking margin of victory is just a little more important than strength of schedule – but by how much? translation: how far off the pedestal should i take Wichita?

    • ptiernan says:

      KenPom possession-based data should shed a light here. It factors in strength of schedule. Wichita State rates out as the #5 team overall, with the eighth best offense and tenth best defense.

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>