So I've wondered this for a while, but it seems that every site that does RPI calculations for basketball ends up with different rankings/values for every team. I thought that the RPI formula was a well-defined calculation? I could possibly understand a little variance, especially early in the season: initial ratings bias, whether or not all games played count (some teams might play games outside of D-I, which may or may not be counted), etc. But for instance currently ESPN RPI has Ole Miss at 33, and another site that I check, rpiforecast.com has them around 121.
Is it just early season inaccuracy that will eventually level out, or are there additional factors that each statistician can play with for their own rankings? I feel like I've seen sites disagree by more than 10 spots late into the season in the past, which is significant considering how much of a factor RPI can be on bubble teams and NCAA tourney seeding.
Wikipedia link regarding RPI for reference.
Comments
S-E-C S-E-C S-E-C
Honestly, if you want what I consider the most reliable rankings, check out the Kenpom rankings at kenpom.com
They are updated with inputs from every game daily and do the best job in terms of overall evaluation.
One big impact on RPI can be how often the site updates the rankings, even if the data is up to date. I also believe RPIforecast attempts to show you where they will rank at years end rather than where they are now. Look at alot of their data points, especially in terms of expected records. Their version of RPI currently is RPILive
http://www.rpiforecast.com/live-rpi.html
This version has Mississippi at 28th currently
It probably already was in all of our minds, but your statement above just made the Mississippi game a must win OOC contest.
KenPom though, which I consider much more reliable currently has Mississippi at 100. They have losses to Creighton and Middle Tennessee. Their highest quality win so far is against St Joes. Early season issues with RPI show themselves in how highly rated RPI has Middle Tennessee. They are at #9, with a 7-1 record. Creighton is #2 with an 8-0 record. Shows you how highly it rates wins more than who they are over.
Tech is 33rd currently in KenPom.
These rankings make it even more critical to win the game since they are a borderline Top 100 opponent.
Whats your thoughts on MTSU?
I wouldn't sleep on them. They have a very good coach, Kermit Davis, and were the giant killer in the tourney last year. They lost 3 seniors which accounted for 29 ppg. A lot, but they kept a lot of guys with serious playing time.
In the past, this team gets to the cusp of the tourney and falls short most of the time, except for last year.
I am not knocking MTSU as much as pointing out Ole Miss really doesnt deserve a top 30 RPI. I was surprised more than anything to see that they took a loss to Tennessee State. That's like Tech losing to UNC Asheville, although it appears Tennessee State is no pushover this season either, with their only loss coming to Vanderbilt.
We will know alot more about the Blue Raiders in the upcoming weeks as they have games against Vanderbilt and VCU. After that though they break out in conference play where they will likely once again be close to undefeated in conference. They went 14-2 last season in the Sun Belt. I could see this team surpassing last seasons 25 wins, which was a school record.
Thanks for the reply.
FYI, MTSU is now Conference USA.
RPIforecast emphasizes each team's "Expected RPI", which is based off of running 10,000 simulations of the season using Jeff Sagarin's PREDICTOR rankings and a given standard deviation (calculated using some measurement of the "quality of fit" of the PREDICTOR rankings, at least that's what it was 3-4 years ago when I contacted him out of curiosity) and calculating each team's RPI for each simulation, and then calculating an average RPI for each team. For this, RPIforecast has Ole Miss at an "Expected RPI" of 121.3.
On the other hand, they also show their "Current RPI", which is only recalculated on this site once each day. Right now, that shows up for me as being 35. This discrepancy between them and ESPN (35 vs. 33) may be due to rounding errors or sorting teams differently in case of a tie (which almost certainly won't be the case later in the season).