I think this is a good thing
obviously rpi has flaws
I like Kenpom

http://www.ncaa.com/news/basketball-men/article/2017-01-12/college-basketball-ncaa-tournament-selection-process-involves

HoopDreams;842792272 said:

---

I think this is a good thing.

---

I second with enthusiasm. I've always found RPI an embarrassment. The factors included were legitimate, but the combination was arbitrary and led to anomalous results. I'm thrilled that the NCAA is looking at moving onward.

HoopDreams;842792272 said:

---

I like Kenpom

http://www.ncaa.com/news/basketball-men/article/2017-01-12/college-basketball-ncaa-tournament-selection-process-involves

---

I agree. If you compare Kenpom Pac12 rankings vs RPI rankings, it makes a lot more sense. My blue colored glasses also like that kenpom ranks Cal 45 vs RPI ranking of 59.

Sagarin has Cal roughly where Kenpom does, in the 40s: [URL="http://sagarin.com/sports/cbsend.htm"][U]http://sagarin.com/sports/cbsend.htm[/U][/URL]

I actually do not think the factors included in the RPI rankings are legitimate. Or rather, it is not legitimate to exclude the factors excluded by the RPI. Margin of victory, opponents' margin of victory, opponents' opponents' margin of victory are valuable pieces of information in aggregate and should not be thrown away as RPI does.

OneKeg;842793166 said:

---

..... I actually do not think the factors included in the RPI rankings are legitimate. Or rather, it is not legitimate to exclude the factors excluded by the RPI. Margin of victory, opponents' margin of victory, opponents' opponents' margin of victory are valuable pieces of information in aggregate and should not be thrown away as RPI does.

---

Margin of victory (MOV) was deliberately excluded from RPI to remove the incentive for good teams to run up the score. Unfortunately, MOV is an important predictor in statistical models, so RPI completely misses a valuable indicator of a team's strength.

I suspect that the components in the RPI formula are present in Pomeroy's and Sagarin's models. I.e., they are legitimate metrics. But they are weighted arbitrarily in the RPI formula.