RPI Education

[BeantownHoya]

Anyone willing to give me a brief RPI(Sagarin) lesson.

Clemson 9-3 SOS: 215 #44

Georgetown 7-3 SOS: 206 #104

I don't get it.

My belief was that a lot of these ranking were based on SOS and "Quality Wins". Is that correct? If so, I am even more confused. Sagarin has us with a better SOS currently. Their biggest win was probably Ohio St. , Gtown's according to Sagarin would be Clemson.

Someone please explain what im missing. I have no problem with them being ranked higher than us, but do their credentials merit them being 60 pts higher?

[showcase]

Actually, Saragin has GU ranked at 105.

www.usatoday.com/sports/sagarin/bkt0405.htm

Details aside, I don't think I can explain the logic of the formula that lies behind the ranking themselves. Heck, I can't understand his claim that after a while the rankings aren't "biased" by the preseason rankings.

It's probably worth noting that although GU's separated from Klempson by 60 slots, it's only works out to a 4 or 5 point difference. Hopefully the next few weeks will see that spread decrease.

[SFHoya99]

I may be wrong, but I'm pretty sure both Pomeroy and Sagarin compare teams by a massive amount of transitive data.

For instance, both GU and Clemson played Norfolk State -- and Clemson won by more. So Clemson is better than GU by the difference.

Of course, GU beat Clemson by 15, so another data point is GU is better by fifteen.

I don't know how far transitive each goes. For example, Clemson beat Norfolk State by a lot (there's a cap on Pomeroy at 16), Rutgers beat Norfolk by six, and GU beat Norfolk by 0 (or 8?).

Also, Clemson - Norfolk - Rutgers - PSU - GU is another data point.

Again, I don't know how far they go, but essentially they take all those data points and combine them. I don't know if there is weighting. This helps to take a lot of the sample size issues away, but obviously not all.

I don't know the exact details, but that's the gist of it.

On the flip side, RPI is a straight formula -- Winning % *.25 + Opponents Winning % * .50 + Opponents' Opponents Winning % *.25, I think.

The winning % are now adjusted for home/away.

RPI annoys me greatly because while Pomeroy and Sagarin check their predictions against actual results (and I hope fine tune them), the RPI is not as good a predictor.

Basically it is a made-up stat that the NCAA is now manipulating to force teams to play more road games and mid-majors.

[nodak89]

with the info from a prior thread this is what I get...
hoyatalk2.proboards30.com/index.cgi?action=display&board=general&thread=1103546765&start=0

Using the modified winning percentage (home loss/road win x 1.4, and road loss/home win x 0.6, neutral site games x 1.0)

Clemson's 9-3 is
Home 7-0 = 4.2-0
Away 1-1 = 1.4-0.6
Neutral 1-2 = 1-2
total = 6.6-2.6
for a weighted winning % of .717

GU's 7-3 is
Home 5-2 = 3-2.8
Away 1-0 = 1.4-0
Neutral 2-1 = 2-1
total = 6.4-3.8
for a weighted winning % of .627

Our two home losses really hurt the weighted winning percentage. With roughly equal SOS, therein lies the gap.