Any time was a good time for college basketball to drop RPI. But the NCAA has to be feeling particularly good right now, ahead of its selection committee's top 16 release Saturday, that it doesn't have to contend with free-falling Kansas being ranked No. 1 by the old and misguided metric.
Instead, we'll be talking about NET: the committee's new evaluation tool that sorts each school's games into quadrants on the selection committee's team sheets. On Saturday we'll get our first look at how much the committee is relying on the new numbers. We'll get to that in a minute, but first let's take stock of where we stand with NET and what we've learned about it.
Of course, when it was introduced, it appeared as if the NCAA had replaced one problem with another. But since then the new metric has stabilized, falling more or less into line with other quantitative systems. Like coaches, we don't know the exact methodology behind NET, but our impression is that it captures some combination of predictive metrics like BPI and résumé metrics like strength of record (SOR), plus a dash of RPI still in there. A partial explanation of the factors that go into NET can be found here.
The quirky part about NET: If it were needed only as a method to place games into quadrants, it probably ought to be based solely on predictive metrics. If it were used only as a method to seed teams for the tournament, there's a case it should be based only on résumé metrics. Instead, it seems to be a conflation.
A consequence of that: Virginia Tech's Buzz Williams noted the other day he sought an otherwise meaningless basket at the end of a game to improve his team's margin of defeat.
On the whole, a quick-and-dirty approximation for NET is actually to simply average the ranks of BPI and SOR, and that often gets you fairly close (something that would not have been true in RPI).
So back to Saturday's rankings reveal. For the first time we'll get some insight into how heavily the committee will rely on its new system. We'll see that in how closely the committee's rankings match NET's, but in particular, these four teams should provide interesting test cases from which we can glean some answers.