(04-28-2021 09:59 PM)slhNavy91 Wrote:  

(04-28-2021 09:24 PM)jedclampett Wrote:  

(04-28-2021 09:11 PM)slhNavy91 Wrote:  

(04-28-2021 08:58 PM)jedclampett Wrote:  

(04-28-2021 05:05 PM)slhNavy91 Wrote:  Thanks for calling my attention to it.
I show final 2020 Massey Composite (with 56 rankings compiled) conference average ranking:
ACC 59.53
AAC 68.98
SBC 71.97
AAC is 9.45 behind ACC and 2.99 ahead of SBC.
Reconstructing what I did two weeks ago, I was going from a post from late January. An earlier post in that thread ("AAC Still #1 non-contract-bowl conference?") cited Massey Composite with 55 rankings compiled - so just that single additional ranking compiled changed the numbers by 0.13 and 0.05 average ranking spots.

That offers NO explanation for messing up 2014!
Looks like I inserted the ACC MC conference average ranking there, not the difference between that and the AAC...so thanks!
ACC 48.73
mwc 76.73
CUSA 82.28
AAC 83.93
You are spot on: 35.2 behind the #5 and 7.2 behind #6.
I'll be correcting the original post in this thread, plus the January post. And here's an updated graph - still looks pretty similar.

The graph (reprinted below) is really a nice contribution to the discussion. Kudos for that!

It could further the discussion to share our thoughts about what the findings show. Here's one AAC fan's take on what the data show:

I. The Trendlines:

--The trendlines indicate that the average AAC rankings have improved by about 4 rankings "notches" (e.g., a "5 ranks" advantage in 2015 and a "9 ranks" advantage in 2020), relative to the average G4 rankings.

--The trendlines indicate that the average AAC rankings have improved much more markedly - -by about 17 rankings notches (e.g., a "~23 ranks" gap in 2015, dropping down to a "~6 ranks" gap in 2020), relative to the average P5 rankings.

--Taken together, the two trendlines suggest that the AAC's rankings may have improved somewhat more relative to the P5 rankings (by narrowing the gap with the P5 teams) than they have relative to the average G4 rankings.

II. The Year-by-Year Data in the Graph:

--A. Year-by-year comparisons with the P5 rankings:

----There appears to be some evidence of a step-wise function:

----Phase 1 (2014): P5 rankings were ~35 ranks ahead of AAC rankings.
----Phase 2 (2015-2018): P5 rankings were only ~15-25 ranks ahead.
----Phase 3 (2019-2020): P5 rankings were only ~5 ranks ahead.

--B. Year-by-year comparisons with the G4 rankings:

----The AAC rankings have oscillated within a slightly narrower range (with an upward trend) relative to the rankings of the G4 rankings.


III. Given how atypical (and "outlier-like") the 2015 AAC rankings were,

--is it possible that the association that we're examining might be "curvilinear," rather than "linear?"

There appears to be some hint of curvilinearity in the 2014-2020 data.

--what would the trendlines be if the graph were to zoom in on the years 2015-2020,

AAC-P5 trendline: The improvement vis a vis P5 rankings remains evident between 2015 and 2020, but it would be less marked in magnitude (closer to a "10 rankings" reduction in the AAC-P5 rankings gap than to a "17-rankings" reduction in the size of the rankings gap between 2014 and 2020).

--Nevertheless, this "10 rankings" reduction in the rankings gap would still be considered very significant.

AAC-G4 trendline: Compared to the 2014-2020 trendline, the 2015-2020 AAC-G4 trendline would flatten out considerably with an average gap of ~ "8 rankings" in favor of the AAC.

--The AAC hasn't moved much farther ahead of the G4 from the standpoint of their average rankings, but it has [i]maintained a fairly sizable rankings lead over the G4 since 2015.[/i]

.



.

Here's another look - at the conference average ranking in Massey Composite for all 10 conferences (plus the independents are in there) from 2013 to 2020. This is average ranking, so lower is better.

- AAC trendline is improvement, and looks like a greater improvement than any of the G4s.
- SunBelt and MAC show similar slope on the trendline, from a much worse starting point...but those two trendlines greatly benefit from their 2020 outliers - they were a lot flatter going 2013-2019.
- mwc relatively flat
-CUSA trending the wrong direction.
- in the contract-bowl conferences, B12 improving, B10 pretty flat, SEC coming back to the rest of that five-team group, and ACC & PAC12 coming back to the AAC.

That's composited average rankings - AAC benefits from being consistently stronger than the G4s top-to-bottom. But relevant to a couple posts ago, in the brand/perception arena, the top of the conference will still be important, so there is also still room to discuss top-25 rankings like AP/Coaches' Polls and CFP rankings.

It's a solid point that the 2020 data should perhaps be considered to be either "outliers" or even somewhat "unrepresentative" in some respects.

However, the 2014 data are also discrepant enough with the data between 2015 and 2019 to either be considered outliers or to suggest that some of the relationships might be curvilinear in nature (in that case, not only would linear regression lines, but also curvilinear trajectories (e.g., "data smoothing" lines) would be of interest.

It would thus be interesting to see what the graph would look like if the outlier years 2014 and 2020 were excluded from the graph.

Regarding 2020 - I suppose there is a discussion to be had about "unreliability" smaller samples, less inter-conference data, etc. But 2020 wasn't really an outlier for the AAC: looking at 2013-2020 or 2015-2020 it's close to the trendline. 2018 was an outlier to the negative, 2019 slightly less an outlier to the positive for the AAC.
2020 really changes things for the SunBelt and MAC. And we've discussed elsewhere how variably the five rating systems dictating CFP payout performance shares treated the MAC.
(I think I was adding my "other thoughts" edit while you were drafting your reply)

But here are 2015-2019 and 2015-2020. AAC does benefit from omitting 2014 - trendline flattens. It seems to me that this is a case where more data is better, though. Just a five year sample is less representative than the seven or eight, right? Anyway, here are two more graphs, with all 11 trendlines flattening.

That was quick! Very interesting. Thanks for posting.

I'm attaching the 2015-2020 version (below) so others can see it. The 2015-2019 version is so similar that the differences would be barely noticeable to most readers.

It shows that, even with the 2014 "outliers" trimmed out:

1) the AAC was a clear-cut "tweener" from 2015-2020

2) The AAC's rankings improved relative to four of the P5 conferences from between 2015 and 2020.

3) The rankings of the MWC, MAC, and SBC all improved to some extent from 2015 through 2020. The rankings of the SBC showed the most improvement, but this was partially offset by the fact that they started out dead last.

Very interesting. Plenty of details to ponder.