Sunday, November 28, 2010

Checking in on the week 13 polls; a few oddities in the AP poll

I analyzed last week polls looking for inconsistencies and any interesting observations and was able to go into more depth on the AP poll since they publish the pollsters votes, and it is time to do the same this week.

First, the Coaches Poll.  There were five first place votes up for grabs with Boise State losing, and they all went to Auburn and TCU, plus Oregon lost two going to those teams.  Auburn was the big winner picking up a total of six and narrowing the gap from Oregon from 61 to 40.  Interestingly, Oregon's total is exactly the same even with the loss of two first place votes so they still benefited by moving up on some ballots due to Boise's loss.  Was Auburn's one point win over a then 2-loss Alabama that impressive that they should gain first place votes?

With Boise losing 472 points and LSU 379, there were plenty of points available for others to move up and they did.  Wisconsin gained 71, Stanford 121, Ohio State 97, and Michigan State a whopping 156.  Arkansas' win gave them a big boost too, 228 points.  Nothing terribly odd here.

Like before, lets look at the total for the undefeateds though.  With 4,248 points available from the voters top-3 picks and the top-3 teams only totaling 4,221, there are some voters that have a one or two loss team in their top-3.  You can certainly make a case for doing that (my computer has Stanford #2), but I'm somewhat surprised there are that many.  It will be interesting to see if this changes at all next week when the votes will be public, and if not, what if any bias there is from the coaches.  With Wisconsin's 1,282 total pretty close to a perfect (for 4th place) 1,298, it is likely that they are the primary beneficiaries of probably TCU being 4th or lower on some ballots.  Fair, perhaps as TCU's schedule being #82 according to my computer, but Wisconsin's isn't much better at only #72.  Stay tuned next week for analysis once we know the votes.

The Harris Poll, Boise 14 first place votes had to go somewhere and Oregon picked up 2, Auburn 11, and TCU 1, all gaining a good number of points.  The top-3 here have 8,194 of a perfect 8,208 points, so a few voters have likely TCU 4th or lower, but a far smaller percentage than the coaches.  Similar changes to the coaches in the rest of the top-10.  It will be interesting to see specifics on the voting next week.

The AP is where we can get a lot more detailed though.  Here, there were 10 first place votes from Boise to dole out and like the coaches poll, Oregon actually got fewer dropping 1.  Auburn picked up 10 and TCU 1 but all gained points as you would expect.  What exactly happened here?

It wasn't just one voter that made a switch from voting Oregon #1.  Surprisingly, 7 voters actually made Oregon their #1 but 8 that had them #1 swapped them out for another team.  I'm not sure the logic here as Oregon beat a good Oregon team by 19 while Auburn had to eke out a 1 point win, albeit on the road.  Was Auburn's win really so impressive as to overshadow Oregon's 19 point win?

Now, there were 2 voters that had Auburn #1 last week and dropped them which, to me at least, is more understandable.  The voters really seem to give a lot of credit for a 1 point win at Alabama.  Strange.

Oregon and Auburn are voted 1, 2, or 3 on every ballot, but TCU is voted #4 on 4 ballots and #5 on 1.  The beneficiaries are Stanford 4 times and Wisconsin twice.

How about the Wisconsin vs Ohio State debate?  Wisconsin has a clear overall lead, and in fact every single ballot has them ahead of Ohio State.  The AP voters are either A) impressed by Wisconsin's late season push or B) abide by head to head.  But if they really did that then Michigan State should be closer shouldn't they?

How about the Wisconsin vs Stanford debate?  Wisconsin is ahead of Stanford 33 times and Stanford is voted higher 27 times.  When Wisconsin is ahead, the average gap is 1.21 spots and when Stanford is ahead the average gap is 1.26.  This has narrowed a bit as last week Wisky was ahead on 35 vs 25 for Stanford.

As I did last week, I've created a chart that shows the high and low vote for each team so you can see the range.


Things are pretty close at the top, but then there is some awfully big variance, especially for Boise State.  But the voters aren't on the same page with Oklahoma, Nebraska, and Missouri too.  And someone thinks awfully highly of 3-loss and #18 South Carolina voting them #8 ahead of 1-loss Boise State and Nevada and a host of 2-loss teams in Arkansas, LSU, Oklahoma, Virginia Tech, Oklahoma State, Nebraska, Missouri, Utah, and Northern Illinois.  Ahead of a few of those, sure, but all of them?  Seems questionable, although they do have the 16th hardest schedule according to my computer.

If you have questions, leave a comment or send me a tweet @computerratings.

Today's BCS Bowl Analysis/Prediction; Who should and will be the BCS at large teams

There is much debate about who the best one-loss team is and who will be selected for the BCS bowls, and I'm not shy to weigh in so here is what my computer says about who should be in the bowls based on merit and who likely will be given the priorities of the bowls (money).

There are still a few games to play and with championship games, things could change, but as of now it appears the AQ teams will be Oregon, Auburn, Virginia Tech, Oklahoma/Nebraska (virtual tie in my ratings), Wisconsin/Ohio State (waiting on BCS, very close), and West Virginia (they'll win, UConn likely to lose).  TCU is in line for an automatic berth, and with Boise and LSU losing, Stanford should move to #4 in the BCS and get an auto-berth.  That leaves two spots and the following teams available:

  • Boise State
  • Ohio State/Wisconsin
  • Arkansas
  • Oklahoma/Nebraska loser
  • Missouri
Boise State should be seriously considered but won't be given the perception of their schedule.  But that is an incorrect perception as my computer has them with a 57th rated schedule which is ahead of Ohio State (62) and Wisconsin (72).  So on merit, they should be picked, but throw in the bowls desire to have hordes of people come to town for the game and that second strike means Boise State won't get a BCS berth.

Since the Big-Televen doesn't have a championship game, their second best team won't incur another loss and is pretty much a lock to get a berth in a BCS bowl and that is probably fair with my computer having Ohio State at #6 and Wisconsin at #11.  You can right this one in in ink.

Assuming Auburn wins the SEC championship, the Sugar Bowl will be looking for another SEC team and would have gladly taken LSU, but with their loss, Arkansas is the most likely candidate and is worthy of it as my computer has them at #8.  I'd say you can right this one in ink too.

What about the OU/NU loser?  Well, that right there is the problem.  The loser is going to incur another loss and go to 10-3 while Arkansas gets to stand pat on their 10-2 record.  My computer has OU and NU at #9 and #10 respectively so they are worthy, but I expect they'll lose out.  What about Missouri?  They are really outside the being worthy rank at #13 in my computer but could move up a bit if Nebraska wins, but given that it is likely the Sugar Bowl making the pick, they'll lose out to Arkansas.

That would make the bowl games:
  • NC - Oregon vs Auburn
  • Rose - Wisconsin/Ohio State vs TCU
  • Sugar - Arkansas vs Ohio State/Wisconsin
  • Orange - Virginia Tech vs West Virginia
  • Fiesta - Oklahoma/Nebraska vs Stanford
The NC and Rose are fixed, the Sugar gets to pick next and would take Arkansas and the Big-Televen team to get the big crowd.  The Orange would stay away from asking Stanford alum to travel across the country and take the neighbor state battle leaving Stanford in the Fiesta.

Now, what happens if something unexpected occurs in the championship games?  Really the only one that would impact anything is if Auburn loses to South Carolina.  In this scenario it all depends on how far Auburn falls and there may be resistance to elevate TCU to the championship game, but that is what should happen and I expect it would.  If Auburn does manage to stay #2, the above stays the same, just replace Arkansas with South Carolina.

If Auburn falls to #3 or below, then things juggle around and I'd expect:
  • NC - Oregon vs TCU
  • Rose - Wisconsin/Ohio State vs Stanford
  • Sugar - South Carolina vs Ohio State/Wisconsin
  • Orange - Virginia Tech vs Auburn
  • Fiesta - Oklahoma/Nebraska vs West Virginia
The Rose would be alleviated of the requirement to take TCU and would jump at Stanford to keep the traditional matchup.  The Sugar would probably rather have Auburn but I believe would be obligated to the SEC champ meaning Auburn falls to the Orange to select and West Virginia ends up in the Fiesta.

Now, if Stanford is somehow passed for the #4 spot in the BCS by Wisconsin or Ohio State, that leaves the door open for a Big-12 team, the most likely would be Missouri as again, I think it would be hard to take a three-loss OU/NU loser.

Thoughts?

2010 Week 13 College Football Ratings and Rankings; Oregon #1, Stanford #2, TCU #3

After an entertaining Friday, and ok a few good games Saturday too, of college football, here are my computer's latest ratings and rankings.

The top two teams stay the same in Oregon and Stanford, both winning, and Stanford actually making up a bit of ground in the ratings on Oregon, but Oregon remains far and away the #1 team.  They could likely endure a loss and still be #1 in my computer at this point they've built such a strong resume of dominating their opponents.

Behind them, Boise falls from the #3 spot allowing TCU to move up to #3, but interestingly Boise drops only to #4 (see below).  Auburn's rating improves a bit but they stay at #5 and Ohio State moves to #6 ahead of Alabama who dropped a bit after losing to Auburn.  Arkansas beat LSU as expected and stays at #8 and OU's big win over OSU moves them up 3 spots to #9 ahead of their likely Big-12 championship foe, Nebraska at #10.

Many of you will be curious, stunned, perhaps even outraged that my computer could keep Boise #4 given their loss.  To understand that, it helps to understand a bit how my computer and other similar power ratings systems work.

First, my computer is tuned to predict points spreads well, not reflect who has won or lost or tell you who the most "deserving" team is.  My computer could care less who wins or loses, it only cares about the score and how it relates to what it "should be" or "should have been".  If a result, win or lose, is close to what it should have been, there will be little to no movement in a teams rating.  Correspondingly if they do better or worse than expected, there will be movement in that direction.  This means a team can win and have their rating drop, or lose and having it rise.

In this case, Boise was favored by Vegas by 15.5 but my computer had it much closer (respecting Nevada) favoring them by 6.2.  When Boise lost by 3, that was a poor result and their rating suffered dropping a healthy 1.8 points.  By comparison, Auburn's rating rose nearly 0.4 points with their win and Alabama's dropped over 0.5 points, so Boise's result was 3-4 times more significant as far as a ratings change.  So their rating definitely dropped.

Second, and to explain why their ranking only dropped one spot, my computer doesn't look at a teams ranking at all.  The ranking is just the order after you sort on the rating.  So a teams rise or fall in the rankings is due in part to their performance and its affect on their rating, but just as much or more on how close other teams are and what they do.  In this case, Boise had TCU close behind but a 3 point gap back to Auburn.  So unlike the polls where there is a discrete ranking and folks feel compelled to raise or drop a team X spots after a win or loss, my computer simply fairly rates a team, then lets the ranking just be a side effect.  Correspondingly, a ranking can move quite a bit with less of a ratings change, as Oklahoma State fell 1.3 ratings points but a full 5 spots.

Similar to Boise, my computer likes Stanford.  One might argue they don't "deserve" to be ranked as high as they are, but their results indicate that they would be favored and win over the majority of teams.  The bowl matchups and points spreads will be interesting to see if the polls are right or if computers like mine are right.  And for the record, my computer picked the Auburn/Alabama, Boise/Nevada, Arizona/Oregon, and LSU/Arkansas games all correct against the spread.

Enjoy!

RankTeamRatingRecordScheduleChange
1Oregon91.24511-068.971+0, +0.021
2Stanford86.99411-172.454+0, +0.785
3TCU84.45712-063.149+1, -0.224
4Boise St84.06010-165.839-1, -1.849
5Auburn83.18712-068.902+0, +0.376
6Ohio State82.07711-165.213+1, +0.257
7Alabama82.0649-368.382-1, -0.554
8Arkansas81.44810-268.480+0, +0.052
9Oklahoma81.07110-270.828+3, +1.362
10Nebraska81.04210-267.610-1, +0.443
11Wisconsin80.25811-164.097+0, +0.395
12South Carolina80.2039-369.429+2, +0.731
13Missouri79.88010-269.064+0, +0.337
14Virginia Tech79.36810-265.739+2, +0.338
15Oklahoma St79.32410-267.738-5, -1.254
16Texas A&M79.2019-370.382-1, +0.069
17Nevada78.07511-162.706+4, +1.379
18Florida St77.7089-367.969+7, +2.234
19LSU77.47610-268.735-2, -0.334
20Arizona St77.0325-672.094+2, +0.445
21Arizona76.8217-471.562-3, -0.687
22Michigan St76.14511-165.218+1, +0.342
23Iowa75.6617-566.721-4, -1.836
24Southern Cal75.6327-573.988-4, -1.246
25Oregon St74.6835-675.497-1, -1.105
26West Virginia74.4278-364.315+14, +2.928
27Notre Dame73.9937-569.537+8, +1.298
28California73.8825-773.321-2, -1.236
29North Carolina St73.8768-466.792+0, -0.298
30Florida73.6217-569.844-3, -1.320

Thursday, November 25, 2010

All the stats you never wanted to know about each conference; SEC not clearly the strongest conference, Big-Televen only #4 conference

There has been a lot of debate lately about schedule strengths and if a non-AQ conference team deserves to be considered for a BCS bowl game.  In light of that, here are a bunch of status and ratings for all the conferences, many of which you probably never knew you cared about but now do.

Note some of this was triggered by seeing @DufresneLATimes retweet @socalspud saying the SEC has no road wins over BCS teams.  I'm not sure that's true (see below), but do follow me on Twitter as @computerratings to see other snippets of information and data like this analysis.

First, here is the average, high, and low rating for each conference using my computer's latest ratings.

RankConferenceAverageHighLow
1Pac-1076.47391.22465.286
2SEC73.96282.81159.807
3Big-1272.90080.59959.948
4Big-Televen70.18681.82059.281
5ACC68.73679.03057.141
6Independent68.37272.69563.198
7Big-East66.25772.97560.017
8WAC65.15585.90947.955
9Mountain-West64.56584.68147.408
10C-USA60.45969.78047.824
11MAC58.09571.08745.308
12Sun-Belt53.58260.10148.552

This goes against the conventional wisdom that the SEC is the strongest conference with the Pac-10 being a 2.5 point #1.  The SEC is followed closely by the Big-12 and there is a bit of a gap back to the Big-Televen.  Clearly, the Big-East is down this year finding themselves just ahead of the WAC and Mountain West.  It is interesting that Boise is moving to the Mountain West next year but right now the WAC has a higher average.

Now, this is skewed a bit by Oregon having such a high rating so if we throw out the high and low from each conference we get:

RankConferenceAverageHighLow
1Pac-1076.02786.20969.737
2SEC74.49382.61864.074
3Big-1273.42580.57867.830
4Big-Televen70.10579.86360.113
5Independent69.22469.22469.224
6ACC68.86675.47460.996
7Big-East66.17771.49964.682
8WAC64.64876.69652.446
9Mountain-West64.14273.70455.457
10C-USA60.79068.43254.223
11MAC58.07766.61049.563
12Sun-Belt53.37057.22050.660

The gap from #1 to #2 is a bit smaller, but still over 1.5 points, so this seems to say the Pac-10 is indeed the toughest conference.

Another way to look at it though is to look at just games a conference plays against non-conference opponents.  Treating each conference as a single team in my algorithm, this results in:

RankConferenceAverageRecordSchedule
1SEC73.25339-5-058.260
2Pac-1070.86421-9-063.054
3Big-1270.08838-8-060.638
4Big-Televen68.66035-8-058.950
5Independent67.33018-11-062.271
6ACC67.12427-15-060.288
7Big-East66.89423-15-060.950
8WAC65.08621-17-063.218
9Mountain-West63.70517-19-065.050
10C-USA61.71920-27-064.725
11MAC59.04315-37-064.805
12Sun-Belt56.2254-31-068.467
13FCS52.3907-83-069.713

The SEC, Pac-10, Big-12, and Big-Televen are again the top-four, but the SEC has moved ahead of the Pac-10 looking at it this way.  Their non-conference record is better (39-5 vs 21-9) but it is against a much weaker schedule.  This is likely explained by the SEC playing only 8 conference games, leaving 4 non-conference games for each of the 12 teams vs the Pac-10 playing 9 conference games and only 3 non-conference for each of the 10 teams.  The SEC usually uses the fourth non-conference game to play a very weak opponent dragging their schedule strength down.

But the Big-Televen also plays 8 conference games leaving 4 non-conference and similarly has a weak schedule but not as weak as the SEC's.  The Big-12 does the same but manages to have a stronger schedule than the SEC or Big-Televen.

If we take a deeper look at the non-conference schedules we see some interesting items in comparing the Pac-10 and SEC:

  • The Pac-10 played 7 FCS schools, the SEC 11, in both cases all at the Pac-10/SEC school.
  • The SEC lost against an FCS school (Jacksonville State at Ole Miss 49-48) and the Pac-10 had one close call against an FCS school (Montana St at WSU 22-23)
  • The Pac-10 and SEC played head to head in one game where Oregon walloped Tennessee in Tennessee 48-13
  • Of the Pac-10's other 23 non-conference games, 13 (57%) were against AQ conference schools, 6 on the road, going 3-3 on the road and 6-1 at home, for a total record of 9-4.
  • Of the SEC's other 33 non-conference games, only 12 (36%) were against AQ conference schools, 4 on the road, 2 at a neutral site, going 2-2 on the road, 2-0 neutral, and 4-2 at home, for a total record of 8-4.
So, these are very similar profiles, the head to head win not being a big factor as it was the best Pac-10 team against a lower tier SEC team.  The knock against the SEC though is that they pad their schedule with an extra weak non-conference opponent while the Pac-10 plays an extra conference game, on average the Pac-10 opponent being stronger than the extra SEC non-conference opponent.

In any case, the perception that the SEC has the strongest schedule is clearly not an open and shut case, in fact one could argue the Pac-10's schedule is tougher.  And all the noise coming from the Big-Televen seems completely unfounded as they aren't even that close a #4 to the top-three conferences.

Thoughts or questions on the subject?  Leave a comment or message me on Twitter.

Monday, November 22, 2010

More checking in on the week 12 polls; AP has LSU as high as #2? Stanford as low as #12?

I wrote earlier about a quick analysis of the polls to see if there was any obvious manipulation or biased voting going on.  I couldn't help but dig deeper and thankfully the Associated Press publishes the votes of all their pollsters so we can take a deeper look.

First, I just looked at the votes and a few things jumped out, but as is often the case, visualizing data like this is usually a lot easier.  With that, I unveil a new chart I'll continue to update showing the top-25 teams, their current ranking on the nicely sloped line in the middle, and then the highest and lowest vote each team received.  Here it is for the most recent AP top-25 poll.


There is no surprise really at the top, Oregon and Auburn both fall within the range of #1 thru #4, but as I mentioned in my earlier post, both Boise and TCU are outside of the top-four on some ballots and TCU is no longer receiving any first place votes.  You can see this nicely on this chart and see that Boise's low is #5 and TCU's #6.  Who has them outside the top-four and who do they have in their place?
  • Craig James of ESPN has Boise #5 and TCU #6 with Stanford and Wisconsin slotting in at #3 and #4.
  • Ferd Lewis of the Honolulu Star has TCU #5 with LSU at #4.
  • Tom Murphy of the Arkansas Democratic Gazette has TCU #5 with LSU #4.
  • Jon Wilner of the San Jose Mercury News has TCU #5 with LSU at #2!
It is this last one that is a little bizarre and you can't accuse him of regional bias.  He seems to be using the logic that Auburn is #1 and LSU lost to them but would beat anyone else so must be #2.  I can't argue that a one-loss team can be rated ahead of an undefeated (my computer has Stanford #2), but given how unimpressive LSU's wins have been this one is a little hard to understand.

You might expect a fairly consistent "window" around the actual ranking for each team, but as can be seen above it actually is quite jagged and jumps around a bit.  A few other interesting observations:
  • Wisconsin's best and worst are both below LSU's, but they come out ahead.  This is due to the skewing of Jon Wilner and Wisconsin still being ahead shows the general consensus is that LSU is not close to #2.
  • Stanford's only loss is to #1 Oregon but Mark Anderson from Las Vegas has them #12!  He has them behind two-loss Missouri (10) and Oklahoma (11) as well as Alabama (6).
  • Oklahoma State has a low of #16 but manages to stay ahead of Michigan State and Arkansas that both have lows of #14.
  • Oklahoma is another like LSU that is getting a few pollsters rating them overly highly as they are behind a Virginia Tech that has a high and low both behind Oklahoma's.
  • Utah, at #23 is still getting support with one pollster having them as high as #16!
Wouldn't it be great if we had this transparency into the Coaches and Harris polls?

What do you think?  Leave a comment or send me a message on Twitter at @computerratings.  If you have questions or want more details on a particular team or pollster, let me know too.

I'll do this again next week and include a look into how voters change their votes from week to week.

Enjoy!

Sunday, November 21, 2010

2010 Week 12 BCS Bowl Game Predictions; Oregon/BSU, Wisconsin/Stanford, Auburn/tOSU, VaTech/TCU, Nebraska/Pitt

Time has passed and it is time to update my prior bowl game predictions.

Using my latest projections, there would be three undefeated teams; Oregon, Boise State, and TCU.  My computer currently projects Alabama to beat Auburn this week although it has Auburn winning the SEC championship game.

While Boise State is currently behind TCU, I expect them to move ahead after they beat Nevada and I don't expect Auburn to get back to #2 after the loss, so Oregon vs Boise State would be the BCS NC game.

The Rose Bowl would have the Big-Televen champ against their pick of teams since they lost Oregon to the NC game.  But since BSU would be the non-AQ conference AQ, they would not be obligated to take TCU.  An 11-1 Stanford would be very attractive to preserve the Big-Televen vs Pac-10 tradition, and my computer says Stanford is the better team so I'd like to think that is the direction they go, but if you believe the talking heads, they are talking about taking TCU to fulfill the non-AQ obligation now when they can get an undefeated TCU.  I doubt it happens though.  And I expect Wisconsin to keep the slim margin over Ohio State, even though the computers could narrow it a bit.

The Sugar would have Auburn as the SEC champ and they pick next.  If TCU is available, they could go that direction, but figure they'll go after the money/attendance and whomever the #2 Big-Televen team would fit that bill.  And should the Rose take TCU, I doubt the Sugar would pick Stanford.

The Orange goes next and should already have Virginia Tech and would likely go with TCU as the opponent.  If TCU does go to the Rose Bowl, they'd likely take Stanford just because any other choice would be a two-loss team as I predict LSU to lose to Arkansas.

That leaves the Fiesta Bowl which would have Nebraska or Oklahoma State (hard to pick between the two right now, I have them virtually tied) against the default of Pitt as the Big-East champ.

  • BCS NCG - Oregon 12-0 vs Boise State 12-0
  • Rose Bowl - Wisconsin 11-1 vs Stanford 11-1 (possibly TCU 12-0)
  • Sugar Bowl - Auburn 12-1 vs Ohio State 11-1
  • Orange Bowl - Virginia Tech 11-2 vs TCU 12-0 (possibly Stanford 11-1)
  • Fiesta Bowl - Nebraska 11-2 (or possibly Oklahoma State) vs Pitt 8-4
Hopefully I'm not biased, as I'm pleased that both TCU and Stanford get into BCS games.  This does depend on LSU losing to Arkansas as if they don't, there is a good chance they go to the Orange Bowl instead of TCU/Stanford.

Thoughts?

Checking in on the week 12 polls; Voter manipulation of the BCS?

As we get closer to the end of the regular season, the polls become increasingly important given their impact on the BCS rankings, so the skeptical amongst us may start to be suspicious about what voters might be doing to manipulate the rankings to get the result they want.  So, it is fun to do a little analysis to see how the votes change from week to week and what might be the reason.

In the AP poll, Oregon and TCU lost a first place vote each and Auburn and Boise picked them up.  Oregon lost two points though so at least one other that had them #2 or lower also dropped them a spot, or the person dropping Oregon dropped them multiple spots.  Similarly, Auburn picked up three points so either they were farther back on the poll that added a first place vote or they were moved up on some other ballots too.  But Boise is the big gainer, picking up 17 points, mostly at the expense of TCU which lost 21.

These changes don't seem severe, but given that only one of the teams played, what is going on here?  Boise's move up is certainly understandable given their 51-0 win over a good Fresno State, and the likely culprit for that would be TCU so that is fine.  But what about the swings in first place votes?  Boise getting one is understandable, but how does an idle Auburn pick one up from an idle Oregon to TCU?  Especially when my computer had the impact of other games improving TCU's rating higher than that of Auburn and Oregon's about the same?  It is only one vote, but something is slightly fishy.

Another check is to see if the four teams that are a clear #1 thru #4 are getting all of those votes.  The answer right now is no with a 9 point gap.  So there are either 9 ballots with one of the top-4 teams at #5, one of the top-4 teams is as low as #13, or something in between.  So clearly someone thinks a one-loss team is better than an undefeated team (likely Boise State or TCU).  My computer believes Stanford is better than Boise State and TCU so I can't argue with that.

But the AP doesn't matter as far as the BCS goes, so let's look at the Coaches Poll.  There, Boise and TCU hold their first place votes while Oregon picks up two from Auburn.  However, even while losing at least two points because of that, Auburn picked up a point in their total.  They didn't get those from Boise as they gained 12 points, instead they likely got them from TCU as they lost 18.

The Coaches poll however, has far more voters with the top-4 at #5 or lower on their ballots.  The gap here is a full 48 points meaning there are a lot of coaches that think a one-loss team is better than the undefeated teams.  Again, I can't argue with that, but that is sure a lot of them since there are only 59 voters.  Nearly ever voter has an undefeated team behind a one-loss team, or a few coaches have quite a few one-loss teams ahead of an undefeated.  Anyone interested in transparency here?

The Harris Poll is the other poll used by the BCS, and here Oregon gains 3 first place votes and 5 points, Auburn loses 4 first place votes and 6 points, Boise picks up 3 first place votes and a whopping 29 points, and TCU loses 2 first place votes and 26 points.  Boise's change is understandable, but it is interesting that Auburn loses ground here while they gained ground in the other two polls.

In the Harris poll, despite there being about twice as many voters (114), the top-4 all being #1 thru #4 gap is just 20 points.  This would seem to indicate these voters respect an undefeated season more than the coaches do.

So, is there a conspiracy?  One certainly can't say so from just analyzing the totals.  There are some oddities, and I suspect some voters making changes this week they wish they would have last week as the reason, and while that is unfortunate and may indicate they didn't do their homework before voting, it is better to get it right before it really matters.

It is interesting that a far higher percentage of the coaches are apt to vote a one-loss team ahead of an undefeated team though.  If you think the voting is stacked towards a particular conference though, the answer is no with the following representation:

  • ACC - 6
  • Big-Televen - 6
  • Big-12 - 6
  • Big-East - 4
  • C-USA - 6
  • Independent - 2
  • MAC - 6
  • Mtn-West - 4
  • Pac-10 - 5
  • SEC - 6
  • Sun-Belt - 4
  • WAC - 4
Basically, all conferences have roughly 50% of their coaches voting.  You could argue the Big-Televen and Independent's are over represented (6 of 11 and 2 of 3) and the MAC, Mountain West, Sun-Belt, and WAC are under (6 of 13, 4 of 9, 4 of 9, and 4 of 9) but it is pretty close.  It is probably safe to say that the coaches from the SEC and Big-Televen are most likely to have one of their one-loss teams ahead of one or both of Boise and TCU, but if it is limited to these conferences all 12 of those voters would have to have two of their one-loss teams ahead of both Boise and TCU.  Actually, that exact scenario would make the 48 point gap I mentioned above.  Might these coaches drop TCU and Boise further down if required to keep them out of the championship game?  Interesting to think about.

Wouldn't it be great if all the voting was transparent?

2010 BCS Computer Scenarios; What if Oregon and Auburn don't win out?

Since I have a BCS compliant algorithm that has results similar to many of the BCS computers, I can play around with scenarios to see what might happen.  In light of that and all the discussion about what happens if one of Oregon or Auburn lose, will Boise State pass TCU, and who will win the Big-Televen, here are some scenarios and what my computer says would happen in them.

The games I'll look at are:

Auburn at Alabama
Boise State at Nevada
Arizona at Oregon
LSU vs Arkansas at Little Rock AR
TCU at New Mexico
Michigan at Ohio State
Michigan State at Penn State
Northwestern at Wisconsin
Utah State at Boise State
Oregon at Oregon State
Oklahoma at Oklahoma State
South Carolina vs Auburn
Nebraska vs Oklahoma State

First, let's assume that all the top teams win out (Oregon, Auburn, Boise State, TCU, Ohio State, Wisconsin, Michigan State, LSU, Oklahoma State).  The assumption is this would put Oregon and Auburn in the national championship game, but which of Boise or TCU gets the automatic BCS berth in the Rose Bowl?  And who wins the Big-Televen?  Here would be my BCS algorithm top-20 in that scenario:

RankTeamRatingRecordScheduleChange
1Auburn87.45913-074.283+1, +0.989
2Oregon87.05312-074.955-1, +0.313
3Oklahoma St85.31212-175.441+3, +1.496
4LSU84.96911-174.693+1, +0.922
5Stanford84.86810-176.187-2, -0.051
6Boise St83.98512-072.856+1, +0.526
7TCU83.80312-072.439-3, -0.748
8Missouri83.1919-276.416+0, -0.239
9Wisconsin83.05111-172.251+1, +0.193
10Ohio State82.74911-172.684+4, +0.334
11Michigan St82.67611-172.522+4, +0.361
12Texas A&M82.3618-376.750+0, -0.122
13Oklahoma82.3099-377.255-4, -0.684
14Nebraska82.1949-375.986-3, -0.634
15Arkansas81.6469-375.465-2, -0.811
16Alabama81.3479-375.078+0, -0.513
17Virginia Tech79.8099-272.690+2, +0.039
18Arizona79.7557-476.365+0, -0.281
19Southern Cal79.5567-477.190+1, -0.072
20Nevada79.49410-272.223-3, -0.717

So Auburn would move slightly ahead of Oregon but they'd remain #1 and #2 and face off in the national championship.  Boise State would also slip past TCU, so assuming they stay about even in the polls, they'd likely get the automatic bid and TCU would have to hope for an at-large berth.  And the Big-Televen title would be very close with less than 0.4 points separating the three teams but Wisconsin holding the edge.  Unless they lose favor in the polls, they'd likely hold on to get the Rose Bowl berth.
That scenario played out pretty much as expected, so let's look at a more interesting one.  What if everyone wins out but Oregon loses to Arizona?

RankTeamRatingRecordScheduleChange
1Auburn87.40313-074.241+1, +0.933
2Oklahoma St85.29512-175.422+4, +1.479
3LSU84.91411-174.641+2, +0.867
4Stanford84.60310-176.025-1, -0.316
5Oregon84.53811-174.978-4, -2.202
6Boise St83.97212-072.851+1, +0.513
7TCU83.78012-072.426-3, -0.771
8Missouri83.1879-276.413+0, -0.243
9Wisconsin83.09711-172.270+1, +0.239
10Ohio State82.81211-172.732+4, +0.397
11Michigan St82.73811-172.570+4, +0.423
12Texas A&M82.3468-376.736+0, -0.137
13Oklahoma82.2999-377.245-4, -0.694
14Nebraska82.1769-375.971-3, -0.652
15Arkansas81.5989-375.423-2, -0.859
16Arizona81.3808-376.081+2, +1.344
17Alabama81.2909-375.030-1, -0.570
18Virginia Tech79.7999-272.682+1, +0.029
19Nevada79.46810-272.202-2, -0.743
20Utah79.4379-273.114+1, -0.163

In this case Oregon falls all the way to #5 opening the door for someone else to move into the championship game.  My rankings would move Oklahoma State all the way to #2 but they are likely too far back in the polls  and my algorithm has them a bit higher than the other computers.  Simply because of the polls, it is likely that the highest of Boise/TCU would go and like above, Boise looks be edge TCU out.
Alright, let's get more interesting.  Let's give Oregon their win back, but what if Auburn loses to Alabama (which my computer predicts) but wins the SEC championship game?

RankTeamRatingRecordScheduleChange
1Oregon87.08612-074.984+0, +0.346
2Oklahoma St85.32012-175.451+4, +1.504
3Auburn84.93412-174.325-1, -1.536
4Stanford84.89310-176.208-1, -0.026
5LSU84.83011-174.627+0, +0.783
6Boise St84.00112-072.875+1, +0.542
7TCU83.81912-072.454-3, -0.732
8Missouri83.2059-276.431+0, -0.225
9Alabama83.18410-274.856+7, +1.324
10Wisconsin83.11411-172.302+0, +0.256
11Ohio State82.81611-172.743+3, +0.401
12Michigan St82.75311-172.583+3, +0.438
13Texas A&M82.3568-376.747-1, -0.127
14Oklahoma82.3179-377.263-5, -0.676
15Nebraska82.2059-375.999-4, -0.623
16Arkansas81.5129-375.350-3, -0.945
17Virginia Tech79.8169-272.703+2, +0.046
18Arizona79.7847-476.391+0, -0.252
19Southern Cal79.5817-477.215+1, -0.047
20Nevada79.51010-272.240-3, -0.701

Here, Auburn drops, but only to #3.  It is entirely possible that they do stay fairly high in the computers meaning it depends on what they do in the polls.  Given they'd have the SEC championship game to make a statement and the perception that the SEC is the best conference, there is a good chance they are close enough to Boise and TCU in the polls that the computers keeping them ahead of those teams could make it very close for them to still get to the BCS championship game.  As far as LSU's case to get there, they would be behind Auburn in my computer as above and would have lost head to head so likely wouldn't get enough consideration.

One thing these scenarios have not taken into account is one of Boise's past opponents, Virginia Tech.  Since the computers "should" take past opponents into account (I question whether Billingsley does as a team that doesn't play doesn't change in his ratings) this could affect things.  So what if we add Virginia Tech beating Virginia and Florida State to the first scenario above?

RankTeamRatingRecordScheduleChange
1Auburn87.45113-074.272+1, +0.981
2Oregon87.04812-074.948-1, +0.308
3Oklahoma St85.28812-175.423+3, +1.472
4LSU84.96511-174.689+1, +0.918
5Stanford84.86110-176.178-2, -0.058
6Boise St84.08512-072.910+1, +0.626
7TCU83.78012-072.417-3, -0.771
8Missouri83.1669-276.395+0, -0.264
9Wisconsin83.04811-172.248+1, +0.190
10Ohio State82.74811-172.683+4, +0.333
11Michigan St82.67511-172.520+4, +0.360
12Texas A&M82.3378-376.730+0, -0.146
13Oklahoma82.2309-377.179-4, -0.763
14Nebraska82.1759-375.971-3, -0.653
15Arkansas81.6409-375.460-2, -0.817
16Alabama81.3459-375.076+0, -0.515
17Virginia Tech80.47711-272.603+2, +0.707
18Arizona79.7507-476.360+0, -0.286
19Southern Cal79.5517-477.184+1, -0.077
20Nevada79.49310-272.219-3, -0.718

This bumps Boise State up about 0.1 giving them a slightly larger edge over TCU but not enough to move farther up.  If Virginia Tech were to lose their last two, it would hurt Boise and put them about 0.3 behind TCU.

In the who knows what happens scenario, what if both Oregon and Auburn lose?  That will depend so much on the polls that it is hard to say.  Auburn it appears has the better shot at staying high enough, but at least one of Boise/TCU figure to make it.  But the pollsters could manipulate things which would be unfortunate.

Thoughts?

Schmidt Computer BCS Algorithm Update; Oregon #1, Auburn #2, Stanford #3

I created an algorithm that follows the BCS rules awhile ago and shared the current rankings last week, so figured I shared them again, particularly since this can give some insight into what the BCS computers might do.

There is no change in the top-3, Auburn actually getting slightly closer to Oregon, but behind them there is some shakeup.  TCU moves up two spots due to Nebraska's loss and also LSU playing a weak Ole Miss.  Boise's win over a decent Fresno State also helped them boosting them a couple of spots to #7.

In the competition for the highest ranking among Big-Televen teams, Wisconsin stays ahead and up 3 spots to #10 despite Ohio State gaining more ratings points but only moving up 1 to #14.

Even though this is a different algorithm than my preferred standard one, the analysis of Stanford being the best one-loss team holds up and they have nearly a full point ratings advantage over LSU.  Oklahoma State is very close behind them though.

RankTeamRatingRecordScheduleChange
RankTeamRatingRecordScheduleChange
1Oregon86.74010-074.266+0, -0.250
2Auburn86.47011-072.971+0, -0.296
3Stanford84.91910-176.247+0, -0.113
4TCU84.55111-072.916+2, +0.247
5LSU84.04710-174.052-1, -0.641
6Oklahoma St83.81610-174.567+1, -0.471
7Boise St83.45910-072.533+2, +0.547
8Missouri83.4309-276.607+0, +0.252
9Oklahoma82.9939-276.350+3, +0.829
10Wisconsin82.85810-172.326+3, +0.788
11Nebraska82.8289-275.097-6, -1.505
12Texas A&M82.4838-376.852+4, +1.234
13Arkansas82.4579-274.638+1, +0.795
14Ohio State82.41510-172.648+1, +1.074
15Michigan St82.31510-171.997-4, -0.005
16Alabama81.8609-274.315-6, -0.535
17Nevada80.21110-171.524+2, +0.256
18Arizona80.0367-375.080+0, -0.295
19Virginia Tech79.7709-272.673+5, +0.913
20Southern Cal79.6287-477.257-3, -1.492
21South Carolina79.6008-374.154-1, -0.307
21Utah79.6009-273.230+4, +0.885
23Iowa78.4937-474.344-2, -0.942
24Florida St78.4208-373.552+4, +0.570
25Mississippi St78.1867-473.938-3, -1.143

Who is the best one loss team in college football? Rational analysis says Stanford

There is much debate going on about who the best one-loss team is right now, and if they are good enough to move ahead of Boise and/or TCU to get into the BCS championship game.  The simple answer is Stanford, and yes (just look at my current ratings), but let's take a closer look at who the best one-loss team is.

First, who are the candidates?  Candidates include the aforementioned Stanford, Ohio State, Oklahoma State, Wisconsin, LSU, Nevada, and Michigan State.  Let's take a look at each one individually.

Stanford's sole loss came at #1 Oregon and they actually led pretty big early before Oregon came roaring back. As can be seen in their chart below, they have one of the toughest schedules in the nation, presently #8, and have played a bunch of meaningful games (larger green dots) and won impressively in a bunch of them, many on the road (five conference road games, ND on the road).  Stanford is also finishing strong after having their three worst games mid-season.  Other than the close games against USC and WSU, which they justifiably get dinged for (green dots well below their current rating), a pretty impressive performance.


Ohio State has some similarities to Stanford, but a key few differences.  First, their one loss is to another one-loss team.  Second, their overall schedule is not nearly what Stanford's is ranking only #55.  Third, they've had the advantage of playing at home more going on the road only four times compared to Stanford's six and their big non-conference game was at home.  They also benefit from their schedule not including Michigan State this year.  Put that all together and Stanford seems to be the more impressive team.


Oklahoma State has a lot of similarities to Ohio State.  Their schedule is similar (#47), no road non-conference opponent, had the majority of their games at home (on the road five times), but their loss was at home and to a team that now has 2 losses.  They also didn't have a good non-conference opponent and benefited from the Big-12 South being down this year.  Last, they beat a 5-5 Troy at home by only 3.  A good performance this year, but not as good as Ohio State or Stanford.


Wisconsin primarily cares about being the best in the Big-Televen so let's look at how they compare to Ohio State.  Yes, they beat them head to head, but they had them at home, and they themselves lost to another one-loss Big-Televen team, so we have to look beyond that.  Overall, their schedule is a little weaker although they did go on the road for one game but against a very weak UNLV.  But their good non-conference win was just by a single point at home against a 4-6 Arizona State compared to Ohio State more easily handling Miami.  This seems to clearly indicate that Ohio State is the better one-loss team when the whole body of work is considered.


LSU is the team everyone is crowing about being the best one-loss team, primarily because they play in the SEC and particularly the SEC West.  They, like Stanford, have their one loss to an undefeated team, but their schedule is quite a bit lower at #40.  Their non-conference schedule included a 7-3 West Virginia they beat by only 6 at home and a 6-5 North Carolina decimated by suspension at a neutral site that they also won by only 6.  So they didn't play a non-conference road game and have only had three road games in total thus far, a big advantage.  Their only convincing (2 TD or more) wins are against 2-9 Vanderbilt, 7-4 Miss St., FCS McNeese State, and Louisiana Monroe, only the MSU game standing out.  Couple that with narrow wins at home over a 4-7 Ole Miss, Alabama, and a 5-6 Tennessee and their resume does not seem as impressive as many of the others.  And if you want to play connect the dots, Boise State just beat Fresno State 51-0, a Fresno State that scored 38 at Ole Miss, a team LSU just struggled to beat at home.


Nevada probably has the most up and down chart of all the teams in question, due in part to their best win being over a schizophrenic California.  Their only loss was on the road to a good Hawaii team, but their only other good opponent was Cal and they got them at home.  Their schedule is clearly weaker at #101 and some closer that expected wins against San Jose State and Utah State don't help their case.


Michigan State is perhaps surprisingly the lowest rated of the one-loss teams, even behind Nevada.  This is due in part to them amazingly having only three road games thus far, but also some mediocre results against not the best schedule (#62).  They beat 4-6 Florida Atlantic by only 13 in a virtual home game in Detroit, nearly lost to a 6-5 Notre Dame at home, and had a tough time with a 4-7 Purdue team this weekend.  Couple that with their loss being a 31 pointer to a now 7-4 Iowa, and you can see why they are clearly behind Ohio State and Wisconsin in the Big-Televen.


So there you have it, an analysis of the one-loss teams.  Stanford seems to clearly be the best according to the computer and also if you look at the results and aren't blinded by specific conference bias.  But here is a summary if the above didn't convince you.  The best in each column is highlighted in green, the worst in red.


TeamRoad WinsConvincing Wins1Narrow Wins2Schedule StrengthLoss Opponent Record3
Stanford522810-0
Ohio State3215510-1
Oklahoma St532479-2
Wisconsin4126610-1
LSU2164011-0
Nevada4111018-3
Michigan St222627-4

1 - Win by 14 or greater against FBS team with winning record
2 - Win by 7 or less
3 - Current record of team lost to

This also makes it pretty clear.  Yes, you can debate the schedule strength since it uses my ratings, but the rest is unrelated to my ratings and seems to clearly show Stanford is ranked at the top or well in each category.  LSU is best in one, but worst in three of them which is why they are justifiably lower ranked.

Thoughts?

2010 Week 12 College Football Ratings and Rankings; Oregon #1, Stanford #2, Boise St #3, TCU #4, Auburn #5

With the top-3 teams in the BCS not playing this week, it was all about jockeying for position behind them.  Most teams held serve, but there are a few changes in the ratings and rankings.

The top-6 teams stay the same in the rankings, but Boise State draws closer to Stanford for #2 and they both extend their margin over TCU.  Auburn and Alabama remain in a virtual tie, the winner next week having a chance to move up, but there is a pretty big margin to get to #2.

The changes begin at #7 with Ohio State moving up a spot swapping places with Arkansas.  And strangely (but see below), Nebraska moves up a spot despite their win dropping Oklahoma State to #10.  Wisconsin and Oklahoma are making late season bids for respect, although my computer says Ohio State is better than Wisconsin, and OU/OSU this week should be a very good game.

How, you ask did Nebraska move up even with the loss?  There are two factors.  First, my computer predicted them to lose, picking Texas A&M by 1 since the game was at College Station.  They lost by 3, pretty darn close to the prediction, so they are not penalized for having a result very close to what was supposed to happen.  But more importantly, a bunch of prior opponents had good weeks.  Washington had a big win and moved up, Missouri won by more than expected, Oklahoma State won big, Texas won big, and Idaho won big as well.  The combination of playing pretty much as expected and prior opponents improving results in a slightly improved rating and moving up a spot.  Weird, but that is the way an unbiased system works.

Despite picking the Texas A&M upset, the computer had a poor week going 23-27-1.  But after 5 consecutive winning weeks, it was bound to happen.  See the prediction performance for more details.

Here is the top-30.  See the web-site for the full list.

Rank
Team
Rating
Record
Schedule
Change
Rank
Team
Rating
Record
Schedule
Change
1Oregon91.22410-068.559+0, +0.596
2Stanford86.20910-172.371+0, +2.086
3Boise St85.90910-064.274+0, +2.347
4TCU84.68111-063.878+0, +1.157
5Auburn82.81111-067.609+0, +0.618
6Alabama82.6189-267.356+0, +0.700
7Ohio State81.82010-165.429+1, +1.741
8Arkansas81.3969-268.542-1, +1.279
9Nebraska80.5999-267.920+1, +1.070
10Oklahoma St80.57810-166.612-1, +0.909
11Wisconsin79.86310-164.180+3, +2.152
12Oklahoma79.7099-269.554+7, +2.804
13Missouri79.5439-269.526+2, +1.988
14South Carolina79.4728-369.324-2, +0.942
15Texas A&M79.1328-370.226+1, +1.593
16Virginia Tech79.0309-265.925+2, +1.974
17LSU77.81010-167.511-4, -0.383
18Arizona77.5087-369.787+2, +0.880
19Iowa77.4977-466.779-2, +0.413
20Southern Cal76.8787-474.184-9, -1.689
21Nevada76.69610-160.738+0, +0.310
22Arizona St76.5874-672.864+1, +1.050
23Michigan St75.80310-164.452-1, +0.045
24Oregon St75.7885-574.465+7, +3.165
25Florida St75.4748-366.747+1, +1.553
26California75.1185-673.992-1, -0.086
27Florida74.9417-468.920+0, +1.110
28Miami FL74.1817-468.508-4, -1.054
29North Carolina St74.1748-365.537-1, +0.826
30Georgia73.8665-667.992+0, +0.939