A realistic look at predictions

Shurna Drives

John Shura and the Wildcats have fans dreaming of the NCAA Tournament.

It has become too much! I know that people love predictions because getting into the NCAA Tournament is really the only thing that matters. I know that today's record doesn't matter as much as tomorrow's most certainly will. Still, there are some crazy predictions being thrown around on some very reputable blogs and message boards for Northwestern, Loyola and UIC (maybe even DePaul too, I haven't checked) about where the Wildcats, Ramblers and Flames might finish thanks to their RPIs.

Most of this has to do with RealTimeRPI.com's GAMER projections. These projections, quite honestly, are worth about as much as the advertising dollars the site is picking up from the Google ads in the middle. By which I mean they are worthless as a predictive tool right now.
While some advanced metrics - such as adjusted offensive and defensive efficiency - have erased much of their noise via sample size, that's not the case in terms of RPI. Sure, that 51 RPI looks pretty for Northwestern on the RealTimeRPI, but when you boil it down, it's composed of two factors - wins and losses. There's a lot that goes into those two things, including Strength of Schedule, Margin of Victory or Defeat, etc. This shouldn't be a simple calculation. Yet, as far as I can tell, that's exactly what RealTimeRPI is doing.

Creating numbers using just two variables creates a huge amount of variability. Every win and loss changes the RPI by a large magnitude. If Northwestern had lost to Iowa State, we might not even be having this conversation right now.

Using the Wildcats schedule as an example here is what RealTimeRPI says about the Big Ten season versus the much more stable Ken Pomeroy rankings:

  1. at Illinois - RPI: 66-61 loss; Pomeroy: 72-68 loss
  2. Michigan State - RPI: 57-54 win; P: 70-68 loss
  3. at Michigan - RPI: 69-65 win; P: 64-61 win
  4. Wisconsin - RPI: 63-55 win; P: 64-61 loss
  5. Purdue - RPI: 52-47 win; P: 66-61 loss
  6. at Ohio State - RPI: 53-45 loss; P: 73-62 loss
  7. Illinois - RPI: 67-60 win; P: 72-68 win
  8. at Minnesota - RPI: 65-62 loss; P: 72-59 loss
  9. at Michigan State - RPI: 50-41 loss; P: 74-65 loss
  10. Michigan - RPI: 75-59 win; P: 68-58 win
  11. Indiana - RPI: 74-59 win; P: 77-64 win
  12. at Iowa - RPI: 68-65 win; P: 66-60 win
  13. Minnesota - RPI: 68-59 win; P: 68-63 loss
  14. Penn St. - RPI: 74-59 win; P: 65-68 win
  15. at Wisconsin - RPI: 61-57 loss; P: 68-68 loss
  16. Iowa - RPI: 74-59 win; P: 70-56 win
  17. at Penn St. - RPI: 68-65 win; P: 62-61 win
  18. at Indiana - RPI: 68-65 win; P: 73-67 win

Add it all up and RealTimeRPI (with two non-conference wins thrown in) predicts a 25-6 overall record and a 13-5 Big Ten record. Hello NCAA Tournament. But not so fast, Pomeroy predicts a 21-10 overall record on a 9-9 Big Ten record. There might not be dancing in the streets in Evanston, but that's still definitely a bubble team.

Other examples of RealTimeRPI versus Pomeroy:

  •  Loyola's RPI projected record is 21-5, 14-4 in the Horizon League. Pomeroy predicts: 18-10, 9-9.
  • UIC's RPI projected record is 5-22, 2-16 in the Horizon League. Pomeroy predicts: 8-20, 4-14. (See, sometimes RPI is the pessimistic one!)
  • DePaul's RPI projected record is 11-19, 4-14 in the Big East. Pomeroy predicts: 11-19, 3-15. (And sometimes they agree!)

Just keep these in mind before booking NCAA Tournament tickets or planning that viewing party. While predictions are fun, they're probably going to be wrong. Still, properly measured hope sure is better than no hope at all.

Leave a comment