clock menu more-arrow no yes mobile

Filed under:

NET? More like NERD

Why the NCAA’s shift away from RPI will just lead to more confusion come March

NCAA Basketball: Northwestern at Iowa Jeffrey Becker-USA TODAY Sports

“THE NCAA GOT RID OF THE RPI” is a flashy headline. There’s no denying that as it was universally hated. The more I dug into the new NCAA Evaluation Tool (NET), however, the more I realized we will just face the same conversations as we do every March. Here’s why it will be worse than those we faced with the RPI.

It is a black box system

This, to me, is a complete red flag. Simply put, they did not reveal any of the inner workings of the NET. The RPI, to its credit (and detriment) was a clean formula leveraging a teams win percentage, their opponents’ win percentage, and their opponents’ opponents’ win percentage weighted 25%, 50%, and 25% respectively. Road wins were given more value - 1.4 vs. 0.6 for a home win - and home losses were considered worse by the same amount. But it was easy.

Where Iowa fans’ got frustrated was with Fran McCaffery and Gary Barta’s insistence on scheduling sub-300 cupcakes. With 75% associated to an opponent, they wreaked havoc on Iowa’s out-of-conference RPI.

In the NCAA’s press release of the new and improved system we know the inputs: “game results, strength of schedule, game location, scoring margin, net offensive and defensive efficiency, and the quality of wins and losses.” That’s where the simplicity ends. A “Team Value Index” has yet to be defined, as Jerry Palm detailed. That seems fine, since it’s only the piece which will receive the highest weight.

Some other concerns I have:

  • Margin of victory is capped at 10 points per game. 10 point games are not created equal. Pace of play impacts it, as well as when that margin was obtained. A wire-to-wire 10 point win in a low possession game is now given the weight as a high-paced back-and-forth game where free throws extend the lead to 10 in the final minute.
  • Weighing each game equally. This is similar to the RPI (as is weighting road/home games differently) in that a win in November is the same as one in March. In Joe Lunardi’s analysis, he specifically called out Oklahoma as a team which was buoyed by early season performance. There is no change for recency bias, though he argues otherwise.
  • Too complicated to understand. As Jerry Palm put it: “If the algorithms (a fancy word for formula) are too complex for people to understand – and let’s be clear, that includes the committee itself – then they are too complex, period.”
  • No release of post-hoc analysis. If it truly is a step forward, a look backwards* at the predictive abilities of NET throughout a tournament would have been a good step towards the transparency the NCAA hopes to provide.

(Ed. Note: Jerry Palm emailed to point out the previously written “1) teams which would have been selected in lieu of tourney teams, 2) the seeding of past tournaments on NET’s merits” conflicts with the subjective nature of team selection so the above bullet has been amended)

It isn’t really any different

It will be utilized in nearly the exact same way as the RPI was: sorting. As NCAA senior vice president for basketball Dan Gavitt explained, “I don’t believe the committee will necessarily rely on the NET more. It will be used as the tool to sort the data on the team sheet.”

Then what is the point? Seriously, why overcomplicate a system for sorting teams if that is its primary use?

Further, they’re carrying over the quadrant wins from last year’s incarnation of tournament selection which boosted the aforementioned Sooners’ tournament bid.

This leads me to my third concern...

Subjectivity will still reign supreme

Since this is a sorting tool with “predictive elements,” there will still be occasions where teams are under-seeded (or over-seeded) or left out of the tournament entirely even though the NCAA’s proprietary algorithm says otherwise.

I suppose this is fine, since bubble talk has kept ESPN and other sports media entities afloat throughout February and March but I go back to my question of before: why? Why spend all of this money - and time and energy - on an algorithm which won’t be used to the T?

Much like the NCAA’s announcement around the draft earlier this month, I do believe them to be well-meaning in this endeavor. However, the release and implied execution of this directive leave a lot to be desired.