RPI vs reality

[broderickpresident]

Yesterday's win over Fordham is a good example of how worthless RPI is. Our performance could hardly have been more dominating and yet we dropped 5-10 spots in RPI because the simple act of playing someone as bad as Fordham is bad for your RPI (regardless of result). This is how power conferences have monopolized RPI for years. They have luxury of buying non-conference wins and being sure that the conference slate will help out their RPI (win or lose). Of course, this strategy can fail when your conference is a "power" in name only (like the ACC has been for much of the last decade).

Conversely, a smarter formula (KenPom) moved the Explorers up about 10 spots based on the dominant result yesterday. Our 45 in the KenPom rankings is about as high as its been all season.

The only blessing for us is that we seem to have successfully gamed the RPI thus far. It didn't make much sense to me why our RPI was so strong before the back-to-back wins over VCU/Butler. Hopefully we can stay in the top 35 heading into the A-10 tourney, but with 4 of our 7 remaining games against teams outside the RPI top 100, some slippage is probably inevitable regardless of result.

Of the other games, it looks like RPI oversells Temple a bit, but really undersells Saint Louis and GW

We also really need VCU to keep its RPI in the top 50 (currently about 39). They close their A-10 schedule with a brutal stretch, which provides lots of opportunities for big wins, but also for plenty of losses (at least they can't sustain many RPI-damaging losses, I suppose)

Check out their remaining slate:

Feb. 14 vs 50 Umass
Feb. 16 vs 118 GW
Feb. 19 at 56 SLU
Feb. 23 at 79 Xavier
Mar. 2 vs 14 Butler
Mar. 6 vs 76 Richmond
Mar. 10 at 40 Temple

[JoeFedorowicz]

Being a firm believer that RPI is the best final measure of a team's season, we have to talk about how it is different than KenPom or the (in my opinion) more bloated BPI.

RPI measures a team's success until that point, being that it it measures Winning %, Opponents Winning % and Opponents Opponents Winning %. KenPom (and others) rank teams from 1-344 based upon statistics. In the end, those two numbers should be close to the same. The problem, in my opinion, is that people believe these numbers are the same when they are not.

EDIT: And to add, I personally don't feel that blowouts in basketball is usable. Some teams have off nights. Others do not. College football, sure, because there are only 11 games or so to measure a lot of teams that don't play each other.

Also, including a measure based upon whether a player is injured (or suspended) or not is toeing a line. It doesn't take into account a lot of factors. For mid-season predictions, sure it is useful, but I really hope the selection committee is using RPI and RPI alone, and not KenPom or BPI (ESPN's admitted rip off of KenPom).

~Joe Fedorowicz
Philahoops.com

[MisterD]

RPI is just another reason for majors to avoid mid-majors in non-conference.

[vasalos]

If we keep winning we wont have to worry about rpi

[mookie]

Feb 10, 2013 12:54:24 GMT -5 vasalos said:

If we keep winning we wont have to worry about rpi

sustained!

[broderickpresident]

If not for the NIT committee finally using some smarter metrics last season (BPI, KenPom), La Salle never would have made the NIT last season

The RPI is, as some like to call it, a "blunt instrument" and can easily be gamed. If everyone had equal ability to schedule games they want, I'd have less of a problem with it.

Here's an interesting breakdown of how the sausage is made for some of the most common metrics:
espn.go.com/mens-college-basketball/story/_/id/7561413/bpi-college-basketball-power-index-explained

The RPI is easily the least sophisticated and offers the least value. I like that BPI and Sagarin cap the impact of blowouts (at a certain point there's no difference between a 25 or 50-point win. La Salle's win over Fordham being a perfect example), but it's mathematically irresponsible to not take margin of victory into account in a ranking metric.

Part of the reason other metrics liked us more than RPI last season was that no one definitively defeated us last season (I think we, along with Murray St, Cuse and Kentucky were only teams to not suffer double-digit losses last season). A 36-point loss is not the equal of a 1-point loss and vice versa for wins.