In this article, I'll be showing you the preliminary compu-picks 2012 top 25, comparing
it to the public rankings, and highlighting a few key results and caveats.
First, though, is a table showing what the actual public top 30 is.
I've created this table as a composite of a number of publicly available
rankings, including ESPN Schlabach, ESPN CF Live, CFN (multiple),
CBS Dodds, SI Staples (multiple), F/+ (early version), and others.
| 1 || LSU || 16 || Ohio St |
| 2 || USC || 17 || Kansas St |
| 3 || Bama || 18 || OK St |
| 4 || Oregon || 19 || Stanford |
| 5 || Oklahoma || 20 || Nebraska |
| 6 || Georgia || 21 || VA Tech |
| 7 || South Carolina || 22 || Clemson |
| 8 || Arkansas || 23 || Florida |
| 9 || Michigan || 24 || Notre Dame |
| 10 || Florida St || 25 || Boise St |
| 11 || West Virginia || 26 || Auburn |
| 12 || Michigan St || 27 || Louisville |
| 13 || Wisconsin || 28 || Utah |
| 14 || Texas || 29 || Baylor |
| 15 || TCU || 30 || Georgia Tech |
And now below is the compu-picks top 25, as well as three notable omissions:
| Rank || Team || Rating || Public Rank |
| 1 || Oklahoma || 0.83 || 5 |
| 2 || Alabama || 0.76 || 3 |
| 3 || Texas || 0.70 || 14 |
| 4 || Oregon || 0.55 || 4 |
| 5 || Oklahoma State || 0.52 || 18 |
| 6 || Texas Christian || 0.49 || 15 |
| 7 || Southern California || 0.48 || 2 |
| 8 || Florida State || 0.48 || 10 |
| 9 || Georgia || 0.45 || 6 |
| 10 || Clemson || 0.44 || 22 |
| 11 || Michigan || 0.44 || 9 |
| 12 || Florida || 0.43 || 23 |
| 13 || Louisiana State || 0.43 || 1 |
| 14 || Auburn || 0.43 || 26 |
| 15 || Stanford || 0.43 || 19 |
| 16 || Ohio State || 0.41 || 16 |
| 17 || Notre Dame || 0.38 || 24 |
| 18 || South Carolina || 0.36 || 7 |
| 19 || Brigham Young || 0.35 || NR |
| 20 || Vanderbilt || 0.32 || NR |
| 21 || West Virginia || 0.31 || 11 |
| 22 || Arkansas || 0.31 || 8 |
| 23 || Utah || 0.29 || 28 |
| 24 || Washington || 0.29 || NR |
| 25 || Nebraska || 0.27 || 20 |
| 27 || Michigan State || 0.26 || 12 |
| 31 || Kansas State || 0.21 || 17 |
| 41 || Wisconsin || 0.10 || 13 |
1) One of the most obvious contrarian moves the system has done is to rank #1 not LSU, not Alabama, and not USC, but
instead the Oklahoma Sooners, last year's consensus #1 team.
The linked article explains
why the model is higher on the Sooners than it was last year (when the system had them #5 instead
of #1 like many others).
2) Two other obvious contrarian projections are for Texas and LSU. Most people have LSU ranked far above Texas; Compu-Picks,
on the other hand, has Texas ranked far higher. I have attempted to explain some of its logic in two articles,
the first one
discussing the huge (and very likely unsustainable) turnover margin difference in 2011 (and 2010),
and the second one discussing talent inflows and outflows.
Of course, one key positive for LSU that the model didn't consider is the apparent upgrade at quarterback.
If they can in fact get solid play behind center, that would be a huge advantage compared to the model's projections.
Still, the model is not at all convinced that they're a true national title contender. Their resume
going into 2012 has far too many red flags and far too few elite indicators other than that they were a fantastic team last year.
Of course, the model was skeptical (though not this skeptical) about the Tigers last year, so feel free to take this clearly
controversial projection with a big grain of salt.
3) One general observation about the model's projections is that it's generally higher than consensus for the Big 12 and lower than consensus for
the Big Ten. I haven't dug into this fully, but I believe that it's largely a function of how highly the model rated the Big 12 last year, and how
poorly the model rated the Big Ten last year. In fact, the model has been down on the Big Ten for each of the past two seasons. This is largely
a function of how much the league has struggled in non-conference games; in 2011 it went 10-12 against fellow AQ's, 4-6 in bowl games, and had a couple of
terrible regular season non-conference losses. In 2010, the league went 10-9 against fellow AQ leagues, 3-5 in bowl games, and also had some terrible OOC losses.
Two straight years, they've lost two games to the MAC or Sun Belt, the 1-A bottom-feeder leagues. In 2011, they tossed
in a 1-2 record against Conference USA as well. Until the league stops losing so many bad OOC games, or starts winning some elite
OOC games (Michigan against Alabama would be a good start), the model will continue to reflect these performances in its ratings,
which then impact the projections going forward.
4) The model also has a fairly interesting take on the SEC. Not especially pro or con, but rather just contrarian about
which specific teams to be high on. As noted above, it's much lower than consensus on LSU, and it also throws Arkansas and
South Carolina into that mix. Meanwhile, the model is especially high on Auburn, Florida and Vanderbilt (as far as I can tell,
it is the ONLY source that has the Dores in a projected top 25).
Florida is an especially interesting case. In a lot of ways, they resemble Texas,
and the model likes them for many of the same reasons. Like Texas, they have struggled the past two seasons, but before that they were
extremely elite (which still does affect the numbers). Also like Texas, they recruit at an extremely high level (though Texas's recruiting
trend is exceptional while Florida's is more holding steady), and they only suffered much smaller than usual draft losses.
One key edge that Florida has against Texas is that the Gators suffered a -12 turnover margin and -0.5 fumble luck, both better than what
the Longhorns enjoyed (which means that there's much more room for improvement).
Meanwhile, South Carolina, a team Compu-Picks is not high on, has a few key negative indicators. First, they suffered their highest
draft losses in at least a decade, and the last time their losses were even remotely close was 2002. Combine that with (slightly) slipping recruiting
and a 2011 campaign that was an outlier compared to most previous years (though 2010 was comparable), and you start to see some indicators that
point to decline rather than further improvement.
Some notes about the current projections:
1) There are a number of data fields that I don't have access to yet, such as injuries, detailed returning starter data, and a few others.
As I am able to access these sources of data, my projections will change as a result. Needless to say, I also do not yet have data summarizing
all the offseason personnel changes, injuries, suspensions, arrests, etc.
2) Similarly, I am still in the process of tweaking and attempting to improve the projection process. While I do not expect substantial rating changes due to
model alterations, I do expect some level of movement to take place as part of the improvement process.
There are a few important notes and caveats I need to make about this preseason projection model:
1) This model is primarily based on the main compu-picks model. Essentially, it attempts to predict how well a team will rate given its rating history,
as well as a number of other data points, such as returning starters, draft talent lost, turnovers, recruiting, etc.
2) There is a substantial amount of noise in these projections, which is to be expected given the large number of unknowns (who will have good and bad luck with injuries,
which young players will improve and which won't, how specific matchups will come into play, etc.). If I were to have all the data inputs (which I don't),
the standard error is a bit over 0.2.
It's important to look at the projections with this in mind to get a sense of how material the projected differences are. Given a standard error around 0.2,
it is fairly safe to project Alabama to be a much better team than Arkansas, but it is not safe to project Vanderbilt to be any better than Arkansas, much less a lot better.
3) Compu-Picks does not endorse implicitly or explicitly any form of illegal gambling.
Compu-Picks is intended to be used for entertainment purposes only.
4) No guarantee or warranty is offered or implied by Compu-Picks for any information provided and/or predictions made.
2012 Compu-Picks Blog
Questions, comments or suggestions? Email me at firstname.lastname@example.org