@DCHawker
If you want better results look at the original, Jeff Sagarin. You can also check Kenneth Massey, who many think has the better system and his web site also lists a summary of all 40 computer rankings so you can see for yourself how well they correlate to reality and each other.
The problem I have with some of the systems is that they were developed by people that might know a lot about statistical methods but are really clueless about basketball and try to explain everything with numbers, even when they really do not duplicate reality.
As I indicated before, I develop (engineering) models for a living and the fist test we run is what we call the "reasonableness test" in which you look at the results/predictions and match them against the real world results and see how close they match; the better the match the better the model. For example, if I develop a model to sort KU players by height based on various physical measurements, the logical order would be:
- Landen Lucas 6-10
- Hunter Mickelson 6-10
- Cliff Alexander 6-8
- Sviatoslav Mykhailiuk 6-8
- Jamari Traylor 6-8
- Perry Ellis 6-8
- Kelly Oubre Jr. 6-7
- Brannen Greene 6-7
- Wayne Selden Jr. 6-5
- Christian Garrett 6-4
- Josh Pollard 6-4
- Evan Manning 6-3
- Tyler Self 6-2
- Frank Mason III 5-11
Or something very close to that, based on numbers presented in the KU Official Roster.
Now, if the model has Lucas and Mickelson reversed or Alexander, Svi, Traylor , Perry and and Green in different order but grouped together it would also be acceptable since they are all roughly the same height. However, if the model had Tyler Self ahead of Ellis and Mason right behind Oubre and Traylor and Selden at the bottom, you would know the model has some serious flaws, right? It does not make a difference what numbers you use to justify, we know it is wrong because we can simply put them side by side, and unless you are 100% blind, you can see the difference clear as day, and again, regardless of the numbers, the difference is obvious, and it is even more convincing when 38 out of 40 models list the order as shown above or with slight variations. Of course the model creator can say the he used a "subjective factor" (Pomeroy's Luck) to arrive at the "true" height and justify the numbers, but the end result still does not pass the reasonableness test. I guess the justification would be in the words of Groucho Marx ...who are you going to believe, me or your own eyes?
I am not sure how closely you have followed Pomeroy's rankings throughout the season in relation to KU, a few of us in this forum have, and like I indicated, I documented the numbers and the discrepancy with the great majority of the other rankings. I believe this is as much as I can write on the subject, you can decide for yourself which predictor your trust the most...unfortunately there is no model to predict that...:(
In the end there are no right and wrong or blacks and white answers and its is really 50 shades of grey...minus all the kinky stuff...:)