I am not an RPI maven, but, I suspect that playing Charleston and UCDavis kinda suppresses one's RPI. Vandy, GB, Creighton, and (obviously) Stanford are good, but, at this early point, not enough to offset the aforementioned two. Frankly, given the weakness of the bottom half of the AAC, UConn may not crack the top handful of RPI-rated teams. Now, if you were to just measure out-of-conference schedules, UConn would hold up pretty well.
Sorry - gotta disagree. The RPI NEVER makes sense. It's built on a "sounds good" but flawed foundation. Been proven many times over that it's predictive ability of rating two teams against each other is very low. With current schedules, no matter how many games are played, RPI won't get the order of the teams right. (Digging back into the recesses of my brain, pulling out some of my graduate engineering math stuff, I think I might be able to argue that IF every team played every other team exactly once, (a 348 game season!!), then RPI might be be accurate - though I'd have to think on this for a bit)
If every team played every other team, everyone's SOS would be approximately the same and RPI would roughly equate to winning percentage. The whole point of RPI is that it compensates for strength of schedule and, to that end, it is a more reliable measure than winning percentage. But it obviously fails when compared with most knowledgeable actual analysis.
Now that the Cardinals' quarterback is out for the year, we don't think that in Arizona either
So Sagarin's UCONN-ND is ND favored by .81 point after home court is figured in.
A metric that isn't predictive isn't accurately ranking teams based on their performance. While no system would be 100% accurate, because there are some factors that influence the outcome of games beyond simply past performance (specific match-up advantages, recent injuries, etc.), the more predictive the metric is, the more accurately it is assessing past performance.
Yeah, but it's closer and the ratings are wildly inaccurate based on the few games so far.
No faith at all comparing teams where the ratings are that close. I think it does a good job of separating the cream from the next level, etc. And of pointing out that there is not one team with the "cream" all to itself this year. All that matters on Saturday is who plays better.
Well, put whatever restrictions you want on the data, that doesn't change the fact that, within a given framework, the more predictive model is more accurately assessing past performance.
