I was intrigued by one of Ken Pomeroy’s recent blogposts at KenPom.com. You can read the whole thing here, but this is the part that caught my attention:

*Every year, someone will talk about how much parity there is college basketball and most analytically-minded people will roll their eyes. I can’t speak for others, but in my case it’s because there really isn’t much change in parity from season to season. It just seems like a boilerplate statement used to generate a feel-good story, and it’s never hard to find a coach willing to support that line of thinking. “Our league is so hard to win this season. There is just so much depth in our league,” said coach of every team favored to win its league.*

*This season, though, might actually be worth talking about in this regard. There are a few things computers do better than humans in the college hoops world and measuring parity is one of them. And there is less of a difference between the second best team in college basketball and the 20th than we normally see.*

*Let’s look at the difference between the second-best and 20th-best teams (based on efficiency margin multiplied by 67/100):*

*2014, January 23 – 4.69**2003-13 Average – 6.21*

*This analysis ignores the top-ranked team because Arizona has a large cushion on the rest of the field right now. I’d guess that the number of times that #1 changes hands in the AP poll is largely what leads to pieces about parity. We haven’t had much of that so far and we’re getting deep enough into the season that if Arizona does lose they’ll still get some consideration to stay at #1, especially if Syracuse and Wichita State have lost by then.*

What struck me about this article is the fact that the question of parity has even drawn Ken Pomeroy’s attention. As he admitted himself, Ken is not one to adopt these sorts of blanket positions lightly. It got me to thinking: I’ve been focused all season on the extraordinary weakness of the top teams. Is there a way to quantify both the relative quality and parity of this year’s so-called elite?

Here’s what I did: for the last ten tourneys (every year that I have KenPom data), I took the Pythag average of the top four teams and the teams ranked 17^{th} through 20^{th}. This would be theoretically equivalent to the top seeds and the five seeds. Then I plotted them on this chart:

As you can see, the average Pythag of the top four teams in 2014 is markedly lower than that of any year since KenPom data was available. The weakest top four before this season was in 2009. But in that year, the fifth-seed grouping was markedly weaker than the top teams—by a wider gap than this year.

In fact, the gap between the one-seed and fifth-seed groupings is tighter this season than is has been since 2007. That year, however, the top seeds were the second strongest of the 10 years, while the fifth seeds were the strongest. Curiously, this was the year that there were only three upsets…and the Madometer recorded its remarkably low reading of 4.1% madness. I guess that when you have unusually strong top seeds, a tight parity gap doesn’t lead to tourney craziness.

But what happens when the top teams are historically bad AND the middling seeds aren’t that much worse? I suppose that 2009 is the closest corollary—and it was surprisingly the second chalkiest dance in 29 years. Actually, in the last three tourneys, which were all remarkably crazy, the gap between the best teams and the fifth seeds has been the widest of the decade. It makes sense, I suppose, if you consider that weak middling seeds are primed for first-round upsets.

So maybe, with relatively strong four to six seeds, we won’t see as many 4v13, 5v12 and 6v11 upsets in round one. Maybe, the craziness will get delayed a round or two—and the top seeds will fall to some of these tough, underappreciated squads. Right now, the teams ranked 17 through 20 in Pythag are Oklahoma State, Ohio State, San Diego State and Iowa State. Can you see them surprising a one seed? I can.

Pete, you keep saying that this year’s top teams’ pythags are historically weak, but I don’t think that’s the case. It really caught my attention when you were comparing this year to 2011; however when I went back and looked, every team in the top 10 this year was higher than in 2011 except one. There’s only one thing I can think of, unless I’m completely reading the data wrong. Did you update your data after Ken adjusted his formula to lesson the impact of blowouts? That changed all his previous data as well, so the pythag scores are quite a bit lower than they were under the old formula.

I updated after last year…because the adjustment was a calculation. But I’m not sure that I can apply the formula retroactively to the time when the teams were entering the tourney. I have to admit, I’m just learning the impact of Ken’s adjustments…and I’m not sure what to do about it. The fact is, in past years, all we had to go by was the numbers at the time. We can’t go back and act as if we had different numbers. If that were the case, I’d also have to adjust all the RPI numbers in my database, because those calculations have changed as well. Let me ask this: did Ken provide a simple way to take the Pythag numbers from the past and apply a formula to change them? Or are we just left with the readjusted results–which of course reflect the tourney results and therefore aren’t helpful in pre-tourney bracket pondering.

I don’t know if he published the formula. I haven’t been able to find it anywhere. That’s too bad, because at first glance, this would actually be one of the stronger years, everything being equal.

Of course, we also have to factor in the impact of the new “offensive-friendly” reffing policies. Somehow, it’s harder for me to see this year’s elite as among the best of the last ten years. I could be wrong.

Pete,

Are the pythag numbers lower for both offense and defense individually vs the overall efficiency rating? Wondering if the offense rule changes have increased offense but decresaed defense, or some other combination.

Not sure if I am getting across what I am thinking.

Gary

I get what you’re asking, Gary. I would assume that the offensive-friendly refereeing would make offenses more efficient and defenses less efficient. The fact that only 12 teams meet the past champion threshold of 92.2 points allowed per 100 possessions tells me defense is rating out as less efficient. (Last year, 28 teams made the grade.)

But the larger issue is, which Byron has touched on, is that KenPom changed his formula this year to discount blowouts against cream puffs…and that means this year’s numbers don’t jibe with the past. Yes, Ken went back and redid his numbers from the previous years…but all those numbers are POST tournament. We have no idea what they were PRE tourney. And those are the only numbers that matter to to bracket ponderers. So we’re a little in the dark here.

Pete,

Thanks for the reply. Would 1-2 games in the NCAA tourney make that much of an impact on the season numbers? I am thinking that small a sample size (for the majority of teams) would not change the number much. Maybe looking at a 1-2 game sample now to see how much of an impact it can have after 20+ games might shed some light on the subject.

Gary

Gary – The outcome of the tourney games has a big impact on the final Pythag numbers and ratings. Heck, one game does. Look at Michigan last night. They beat Purdue by eight…but projections were for them to win by 17. They were there with 3 minutes left, but just coasted in. That performance–a win against a decent Purdue squad–dropped them them six to 10 in the rankings. So…what the Pythag raw and rankings are pre-tourney are quite different from where they end up after the dance.