There's a lot of general misunderstandings about the RPI. I'm not going to get into all of them, but I can explain to you, mathematically, the advantages of scheduling division 2 opponents over very bad division 1 opponents.
Our variables (some courtesy of live-rpi)-
Tennessee's RPI- 0.5688 (59th in country)
Tennessee's SOS- 0.5544
Tennessee's AWP- 0.6119
Tennessee's Adj record- 16.4 wins 10.4 losses
Kennesaw State's SOS- 0.4565
Kennesaw State's AWP- 0.0942
Presbyterian SOS- 0.4372
Presbyterian AWP- 0.1570
Our formulas:
RPI = (AWP * 0.25) + (SOS * 0.75)
SOS impact = (AWP * 0.66) + (OWP * 0.33
SOS = (OWP *0.66) + (OOWP * 0.33)
In this case, the OWP is actually our opponent's AWP, and the OOWP is their OWP.
Assumptions-
We don't have our opponents' OWP. We have their SOS, which uses the OWP as a heavy weight, but also contains the OOWP. To get the OWP would require nested for loops parsing ~900 data points per team. Since this is by hand, that ain't happening, so let's assume OWP and OOWP are roughly equal and make our SOS stats for those teams their OWP. This means our final answers won't be entirely accurate, but it will only affect it by a spot or two in the rankings and not necessarily against UT's favor. A further note, contrary to popular belief, the venue does NOT affect the opponent's weight on the schedule. Neither does winning or losing. SOS is taken by adding up all your opponent's adjusted records and all their opponent's adjusted records to produce OWP and OOWP. The weighting for venue affects these adjusted records, but there is no further weighting.
We also don't have the opponent's weight on our SOS. Instead, we're going to multiply UT's SOS by 32 and assume each opponent is equal weight. This will actually negatively affect UT by a few spots because of the way OWP is calculated. As most of you know, wins/losses are of different values depending upon venue. Because of this, teams that lose a great deal, such as our opponents above, have a heavier weight due to all those home losses. On an unrelated note, the converse is true for dominant teams due to their success on the road.
Calculations:
So now with variables and assumptions out of the way, let's assume we're getting rid of Kennesaw State and Presbyterian and replacing with division 2 opponents who don't count towards RPI. First, we have to remove those adjusted wins from our adj. win column. Since they were both home wins, we need to remove 1.2
New Adj. record= (16.4-1.2) wins 10.4 losses
= 15.2 wins 10.4 wins
New AWP- 15.2/(15.2+10.4) = .5938
Note: Adjusted record was calculated from UT's schedule using a for loop by hand.
Now let's get our SOS. As stated above, we'll multiply our current SOS by 32(#games) to get a number we can subtract from.
SOS * #games = 0.5544 * 32 = 17.7408
OK, now let's get the SOS impact for both K-state and Presbyterian.
K state impact = ((0.0942 * 0.66) + (.4565 * 0.33)
= .2128
Pres impact = (.157 * 0.66) + (0.4372 * 0.33)
= .2479
Now let's subtract these impacts from our number above
New SOS * #games = 17.7408-.2128 - .2479 = 17.2801
Now, we divide this number over our new number of games, 30.
New SOS = 17.2801/30= .5760
Finally, with our New SOS and AWP, let's calculate RPI.
New RPI = (New SOS * 0.75) + (New AWP * 0.25)
RPI = (.5760 * 0.75) + (.5938 * 0.25) = .5805
So, what would have been the impact of playing d-II opponents instead of those two games? Well, with that number, Tennessee would have finished 52nd, a good 7 spots higher and in front of Cal and Nova. Given our assumptions potentially negative impact on UT, the high 40's is definitely within our error. Regardless of whether we get in or not with those numbers, its clear that in terms of RPI it's smarter to schedule division 2 teams over bottom feeder d-1 teams.
