(08-08-2022 11:50 PM)Milwaukee Wrote: (08-08-2022 06:20 PM)quo vadis Wrote: When we compare the final Massey Composite of 2021 with the final AP poll, we get:
Massey .... Cincy #6, Houston #21, ULL #22
AP .......... Cincy #4, ULL #16, Houston #17, BYU #19, Utah St #24, SDSU #25
That looks like a systematic over-ranking of G5, to me.
It might, if one were to view the Massey Composite as the "ACTUAL TRUTH," but it isn't.
The Massey Composite is just the average of dozens and dozens of pollsters and ranking systems, each of which is flawed in its own way.
There is the "garbage in, garbage out" problem with the Massey Composite.
A pretty strong case can be made that the sports writers and the coaches are probably more knowledgeable than the various and sundry folks that make all the ratings that are compiled into the Massey Composite.
Thus, I don't believe there is a strong case to be made that the Massey Composite rankings are any more valid than the AP or Coaches polls.
IMHO, the major value of the Massey Composite football rankings is that they rank every FBS program - - not that the Massey Composite top 25 is more valid.
If it were otherwise, the AP or Coaches polls might have been replaced by the Massey Composite, but they haven't been, and don't seem likely to be.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
NOTE: It might be helpful if there were something like a kenpom.com or barttorvik.com for college football (and maybe there is), since such indices provide quantitative (statistics-driven) alternatives to poll-type rankings systems.
However, even if we had a kenpom.com for college football, it wouldn't automatically become "the gold standard." Even in college basketball, which has a greater wealth of rankings data, there is no single rankings system that is considered "superior" to the others.
The benefit of having such a system is that having a wider range of very different ways of ranking teams could assist in developing a ranking methodology that could offset the limitations of the existing methodologies.
However, it could take years and considerable effort to develop a system significantly better than the (highly-flawed) NET used by the NCAA to rank D1 basketball teams.
About the bold - FWIW, I don't believe the MC is the "actual truth", there is no actual truth unless God wants to give us weekly rankings.
But I do think it is better - closer to the theoretical, unknowable "truth"- than the AP and Coaches polls. It includes those polls, plus dozens of what IIRC are mostly computers. To my knowledge, computers tend to be developed by "geeks" who don't have dogs in the hunt, IMO they tend to be more fascinated by the math and statistics of determining what makes a team "better" than rooting for the SEC or the ACC or the SBC. Yes, each computer reflects the beliefs of its developer about what makes a "good team" - computer A might emphasize SOS a bit more, computer B might emphasize MOV a bit more, etc. but then that's why you compile a bunch of them, because there's a better chance that these will balance out. That's the idea the BCS formula was based on. multiple computers for the computer component.
So - after around week six, which is when IIRC most computers have flushed out data from the prior year as placeholders - I trust the computers more than human voters who are reacting in real-time. Computers are less passionate, less emotional, more "objective", presumably, to me. Unless a computer geek has written code that says something like "If team is from 'SEC' add three points to its ranking", it is "objective" in a systematic sense. It might be wrong, but not prone to week to week subjectivities like a human voter might be.
IOWs, to me, most of these MC computers attempt to do what "KenPom" and others do with college hoops, provide an "objective" statistics-based appraisal of teams. For example, I just clicked on one of the computers used in the MC, ESPN's "football power index", which seems to be designed to do that kind of thing. FWIW, it had only one G5, Cincy at #10, in its final top 25 last year. It might be flawed, but it is trying to do the same thing, I imagine.
That said, I agree that I wouldn't trust any single computer as the "gold standard", which is why I like the MC - it compiles a bunch of them, and other ranking systems.
As for "replacing" the AP and Coaches polls, I'm not sure for what purpose? They aren't formally included in the CFP rankings that determine playoff spots. As for what gets publicized in the mass media, well, the AP is the mass media so there is an overlap there, LOL, but more generally I think the AP and Coaches polls have that prominence because of tradition, they've had the prominence for decades, not because they are better than computers.
So I am quite comfortable using the MC as a flawed, but better benchmark for what is "closer to the truth", and thus with my theory that the voters tend to be biased towards G-teams with a soft schedule/good record resume later in the year.
Just MO.