Sunday, November 27, 2011

What it would take for Oklahoma State to pass Alabama for #2 in the BCS

I wrote an analysis of Oklahoma State's chances of passing Alabama for the #2 spot in the BCS last week and gave them two ways to accomplish it.  They needed a loss by Alabama or the voters to decide they didn't want to see a rematch.  Clearly the former didn't happen, and despite some oddities in the AP poll voting, we didn't see a mass change of votes in the coaches or Harris polls as Oklahoma State is still 5th in both.

All of the analysis and pundits are now saying an LSU vs Alabama rematch is inevitable, and I don't disagree, but it is still worth taking a look at what it would take for Oklahoma State to get to #2.

First, they need to win against Oklahoma.  A win would likely get them some of the #2 spots back that they lost in the BCS computers to Alabama.  I expect that a win would cement their #2 with the 2 computers that kept them there, and 3 of the other 4 computers would likely move them to #2.  This would have them #2 in 5 of the 6 computers, and after you throw out the high and low, they would get the full 0.960 vs 0.920 margin, or 0.04.

Second, should Georgia somehow beat LSU, that might hurt Alabama in a few of the computers increasing the chances they move ahead of Alabama in all of the computers.  Further, there is a slight chance that an LSU loss could drop them behind Oklahoma State in 1 or 2 computers which might increase their computer number to 0.970 or a margin of 0.05.  For what it's worth, if LSU were to lose and Oklahoma State win, they would move to #1 in my BCS algorithm and LSU would be #2.

That margin of 0.05 then has to be not lost in the polls.  Effectively, they can give up no more than 0.025 in each poll, or 71 votes in the Harris poll and 36 in the coaches poll.  They are presently 342 and 166 votes back respectively, so they have ground to make up.  This leads us to the third item they need.

Third, they need Virginia Tech to lose to Clemson, or at least have a close win that convinces the voters that had Oklahoma State behind them to swap things.  This could help significantly to make up the ground and eliminate another 1-loss contender.

Fourth, they need to not only win against Oklahoma, but do so in a convincing fashion that causes the voters to move them ahead of Stanford and Virginia Tech.  With 115 voters in the Harris poll, to be 71 or fewer back, they'd have to be #2 or #3 on all ballots, and #2 on at least 22.  With 59 voters in the coaches poll, they'd have to be #2 or #3 on all ballots, and #2 on at least 14.

Fifth, again, an LSU loss to Georgia could help in the polls as it would possibly get them a few #1 votes increasing the chance of maintaining the 0.05 gap they'd have with the computers.

There you have it.  Is any of this likely?  The win over OU certain has a good chance of occurring, but that may be about it.  OU is not going to roll over so that game could be close negating the 4th point above, and it is pretty unlikely that LSU loses to Georgia which negates the 2nd and 5th points.  And while Clemson beat Virginia Tech earlier this year, beating a team twice is no easy task so the 3rd point may not happen either.

So it is hard to imagine it happening as it would depend on a pretty significant change of opinion with the voters and while that may happen sporadically with some voters, I doubt it will happen to the degree necessary to pull it off.

Analyzing the 11/27/2011 AP Poll: Clear #1 and #2, confusion and fishiness beyond that, Oklahoma State may have a chance

This is the next in the series of analysis posts on the AP poll.  Some very fishy goings on by a few voters.  If the coaches or Harris voters do this too, perhaps Oklahoma State can get past Alabama.

To start, here is the summary view chart.  The top-25 teams are listed across the X axis and the top and bottom of the vertical lines for each team indicates the high and low ranking the team has received.

For comparison, here is last week's chart.

Alabama's win over Auburn and LSU's over Arkansas solidified the views that those are the two best teams, but behind them, there is now an astounding 8 teams receiving 3rd place votes.  There also continues to be a wide range of votes, although perhaps not quite as much as last week.

What we do see this week is that Oklahoma State jumps to #3 from #5, leap frogging Stanford.  This is a bit odd and may be an indication that some voters are changing their votes simply to try to not have an LSU/Alabama rematch.  Yes, I know the AP isn't used in the BCS but if the AP voters are doing this, the coaches and Harris voters might too.

The reason the leap is odd is that Oklahoma State didn't play, and the opponent they lost to last week was thumped 26-6.  So you wouldn't think people's opinion of them would go up from that.  Further, the team they went by, Stanford, won fairly comfortably against a decent Notre Dame team so you wouldn't think they disappointed and would be dropped.  Yet the switch happened.

Last week Stanford had 1,224 points to Oklahoma State's 1,206 so they were close.  This week it is Oklahoma State at 1,286 to 1,281, so it is very close, yet a swap of 23 votes occurred.

Last week, Stanford was ahead on 31 ballots with an average gap of 2.39 and the Cowboys were ahead 29 times with an average gap of 1.93.  This week it is reversed, the Cowboys ahead on 31 ballots with a gap of 2.06 and Stanford ahead on 29 with a gap of 2.03.

If we look at the actual ballots, 4 voters actually swapped Oklahoma State ahead, Anderson, Gimino, Markey, and Sargeant and 2 voters swapped Stanford ahead, Higgins and Mulhern.  Let's take a look at each of these:

  • Mark Anderson from Las Vegas had dropped Oklahoma State to #8 last week, having Arkansas #3 and Stanford #4.  This week Oklahoma State is suddenly #3 while keeping Stanford #4.  How does an idle team move up 5 spots?!
  • Anthony Gimino from Tucson moved Oklahoma State from #5 to #3 keeping Stanford #4.  This is questionable, but at least not so much as Mark Anderson's.
  • Matt Markey from Toledo also unbelievably moved Oklahoma State from #8 to #3, but also moved Houston from #9 to #4 dropping Stanford from #5 to #6.  Very strange how Stanford can have a team ahead of them lose (Arkansas), but not only do they not gain ground, they lose a spot!
  • Keith Sargeant from New Jersey moved Oklahoma State from #9 to #3!  He at least did move Stanford from #6 to #5.
  • Ron Higgins from Memphis moved Stanford from #5 behind Oklahoma State to #3 ahead of them.  Rational given them beating Notre Dame.
  • Tom Mulhern from Wisconsin did the same.
How can Mark, Matt, and Keith defend the change in votes they made for Oklahoma State?  Have we ever seen an idle team with the team they just lost to losing 26-6 go up 5 or 6 spots in the top-10?

Is Oregon the 3rd best team after LSU and Alabama? My computer says yes

It is now pretty much a given, that short of a significant voter uprising, LSU and Alabama will finish #1 and #2 in the BCS and play in the BCS title game, even if LSU loses to Georgia in the SEC title game.  But that certainly doesn't stop fans from debating it or wondering who the next best team behind them is.

Conventional wisdom at this point is that Oklahoma State is close to the top-2 and most deserving, but will likely fall a bit short due to how the pollsters are presently voting.  They are very strong in the BCS computers but have a gap to make up in the polls and it is hard to see it happening as I wrote last week, especially with Stanford winning and thus taking some of the #3 and #4 poll votes away from the Cowboys.

But are they even the 3rd best team in the country?  While the BCS computers and even my BCS algorithm actually say they are the 2nd best just ahead of Alabama, my normal and more accurate algorithm says Oklahoma State is #4 and Oregon is #3.

How can this be?

First, as I've written before, the BCS has imposed limitations on the BCS computers to not use the score of the game at all, only who won or lost, thus significantly handcuffing them from being able to rate teams as fairly as possible.  I don't think anyone hasn't looked at one or more of the computers results and says "Huh?" about where several teams have been ranked, and many are of the opinion that systems that use margin of victory correctly rate and rank teams more accurately.

But, how can one defend a system that ranks a 2-loss team from a perceived weaker conference ahead of a 1-loss team from arguably the toughest conference this year?  As always, let's go to the performance charts, where we can see it boils down to who the losses are to.

Oklahoma State lost to a now 6-5 Iowa State that just lost 26-6 to a 2-loss Oklahoma team.  This can be seen the chart by the large red dot scoring just 67.4 points on their chart.  They also had a near loss to 6-6 Texas A&M, scoring just 83 points and another close game with K-State scoring just 81.9 points.

Oregon on the other hand lost to the clear #1 and undefeated LSU, so they were supposed to lose that one and shouldn't really be dinged for doing so and in fact score 82 points from that game, scoring about the same as Oklahoma State's 2 near losses, and also lost to a 10-2 USC that gave 11-1 Stanford all they could handle and arguably should have won that game, scoring 74.6 points from that game.

So Oregon has one very expected loss and another loss that was not nearly as bad as Oklahoma State's one loss.  Couple that with both teams having a number of very good wins, but Oregon's win over Stanford far eclipsing the best win Oklahoma State has had, and you see how Oregon can be rated higher than Oklahoma State.

Now, Oklahoma State has a chance with their game with Oklahoma to get a win similar to Oregon's over Stanford, and that could turn the tables, but with the games played thus far, my computer says Oregon is better than Oklahoma State and would be favored should they play on a neutral field.

College Football Week 13 Ratings and Rankings - LSU and Alabama remain #1 and #2, Oregon up to #3

This week's ratings and rankings are now posted.  With some teams playing their final regular season game, and all teams having played 11 or 12 games, there isn't much change this week but there is some.

At the top, LSU and Alabama remain a clear #1 and #2, easily outdistancing new #3 Oregon.  Look for a post in a minute explaining why Oregon is now the 3rd best team after LSU and Alabama.

Oregon moved ahead of now #4 Oklahoma State and after them the teams are the same until new #8 Wisconsin moves ahead of now #9 Stanford.  Stanford actually improved their rating after beating Notre Dame, but still got passed.  And new to the top-10 for the first time this year is USC after their drubbing of UCLA.

Houston does make a move as well but just misses the top-10 by 0.02 falling to #11.

The top-25 is below, the full rankings on the web-site.

1LSU94.99012-071.796+0, +1.011
2Alabama91.97011-171.626+0, +0.054
3Oregon86.29410-270.829+1, +0.291
4Oklahoma St86.00410-172.299-1, -0.716
5Oklahoma84.7989-272.329+0, -0.556
6Arkansas83.68310-271.136+0, -0.605
7Boise St83.59410-167.509+0, -0.473
8Wisconsin83.41210-265.875+1, +1.235
9Stanford83.26011-169.401-1, +0.705
10Southern Cal80.65710-271.100+2, +0.671
11Houston80.63712-060.664+5, +2.308
12Georgia80.11710-269.109+1, +0.709
13Michigan79.81610-268.791-2, -0.704
14South Carolina79.09410-270.290+3, +0.940
15Texas A&M79.0466-673.423-5, -1.486
16Michigan St78.08210-267.339+3, +0.579
17Kansas St77.8909-272.220-3, -0.857
18Missouri77.5617-572.085-3, -0.803
19TCU77.5409-265.064-1, -0.199
20Baylor77.4348-372.215+2, +0.388
21Texas77.3127-472.695-1, +0.167
22Nebraska76.7909-370.547+1, +0.561
23Notre Dame76.3828-470.538-2, -0.698
24Virginia Tech76.32111-166.619+0, +0.438
25Florida St75.9698-464.939+3, +0.864

Friday, November 25, 2011

What if Oregon had beat USC?

There are a lot of what ifs we could ask, but in light of ASU folding and UCLA now being the Pac-12 South's representative for the Pac-12 title game, it is interesting to think about what may have happened if Oregon had beat USC resulting to a move to #2 or a very close #3 in the BCS, and then had to play UCLA.  I wrote earlier that Oregon had a very long shot, about the same as Boise State, of getting there, so given all that has happened is that true?

First, what would have happened last week had they won?  They entered the week #4 in the BCS, a little over 0.03 behind Alabama and nearly 0.09 behind Oklahoma State.  This was made up of being #4 in all of the computers and polls.

In the polls, Oklahoma State's loss would have given voters reason to give some #3 and perhaps #2 votes to Oregon.  And Oklahoma's loss would have also helped, turning whatever #5, #6, or #7 votes they may have received into #3, #4, or #5.  In the end, given they were so close to Alabama, they would have gotten closer but it is very unlikely they would have passed them.

In the computers, we already saw that Oklahoma State stayed #2 in most of the computers and only fell to #3 in the others, and Oregon's #4 in the computers coming in to last week had two #5's and a #7 so they would have likely stayed #4 in the computers, although the margin may have been a little closer.

So Oregon would have gotten closer to Alabama for now #2, but would probably still be trailing.  I'd guess the margin may have dropped from the 0.03 to no less than 0.01 and probably closer to 0.02.  So they'd be within striking distance.

Except that they are playing 3-8 and #93 in my BCS algorithm (#73 regular algorithm) this week which would not help them in the computers.  And now, they'd be closing against presently 6-5 and #50/#61 UCLA which also wouldn't really help.  So like I originally wrote, they'd need additional help, i.e. an Alabama loss, to have a chance of getting to the BCS title game.

But, it didn't happen, and LSU won today, so the drama of the past few weeks appears to have subsided, but we still have Alabama/Auburn, Virginia Tech/Virginia, Stanford/Notre Dame, and Houston continues to lurk.

Wednesday, November 23, 2011

Could Oklahoma State pass Alabama for the #2 spot in the BCS?

Note: See the more recent post on what it will take for Oklahoma State to pass Alabama.

A comment was added to my discussion of David Shaw's criticism of the BCS asking if the computers could help Oklahoma State jump ahead of Alabama.  It is a worthwhile question so let's answer it.

First, for this discussion to be relevant, we probably have to assume that LSU beats Arkansas eliminating them (they are #3 in the BCS between Alabama and Oklahoma State).  Also, Alabama needs to beat Auburn, otherwise Oklahoma State is probably gift wrapped the #2 spot.

Second, the the Cowboys are ahead of Alabama in 4 of the 6 computers and are immediately behind them in the other 2 giving them a 0.02 advantage from the computers.  Today this is more than offset by the polls having Oklahoma State 491 and 257 points behind leaving them over 0.1 points behind Alabama and #4 in the BCS.

So, what would it take for Oklahoma State to get to #2?

They'd need to get everything they can from the computers, and while they are already ahead of Alabama in 4 or 6 computers, they could increase their margin in the computers to 0.04 if they get ahead in all of them.  This is possible as they get to play Oklahoma with a #6 BCS computer average while Alabama plays Auburn with a #22 BCS computer average.

Then, they'd need to get close enough in the polls to not lose that entire 0.04 margin.  Since the polls are worth 2/3 in the BCS computation, that effectively means they'd have to be within 0.02 in each poll.  That equates to less than 58 votes in the Harris poll and 30 votes in the Coaches poll.  Given that they are 491 and 257 behind today, that is a long ways to go.

Now, if they beat Oklahoma convincingly, they could get a bump in the polls.  And we are assuming Arkansas loses so they'll get their share of votes there too.  But with Stanford and Virginia Tech between them and Alabama, and each of them playing a quality opponent, I doubt they can leap over both unless one or both of them lose or have a really bad showing in a narrow win.

The only way I see Oklahoma State getting ahead of Alabama is an Alabama loss, or the voters deciding this coming week or next, that they do not want to see a rematch and we see voters changing their votes to manipulate the system to avoid it, putting Oklahoma State ahead of Alabama and making up what is needed in the polls.

The beauty of this year though is that there probably will be an upset of some sort this weekend and we'll be back discussion new scenarios in a few days.

NFL performance charts after week 11

When I post preview of games, I always include the current performance charts.  So I thought I'd share the full set of performance charts for the NFL and they are below.

These charts allow visualization of the performance of a team and you can quickly see how consistent a team is, how they have improved or regressed, and many other trends.

There is a bunch of information in these so a quick key.

  • The green/red dots show how a team performed in a given game.  A green dot a win and a red dot a loss, but the position on the y-axis shows the points they are getting for that game used to compute a current rating, and the size of the dot says how meaningful a game is.
  • The black horizontal line is the current rating.
  • The dark blue line is the rating/ranking by week.
  • The light blue line is the schedule strength/ranking by week.
  • The yellow diamonds are the current rating of the opponent by week.
It is no surprise that the best teams show consistency and any inconsistencies are usually a loss.  Green Bay has been hovering right around the high 80s and low 90s in all their games.  San Francisco has as well, except for their early season loss to Dallas.

Baltimore is another interesting example to look at.  They have a bunch of dots well above or well below their current rating, showing how up and down they have been this year.

This also shows how a loss is not necessarily a bad thing.  Look at New Orleans for example, and how their week 1 loss to Green Bay is actually slightly higher than their current rating.  Nor is a win always a good thing, Baltimore's win over Arizona being well below their average.

Let me know what you think!

College Football Over and Under Achieving Teams

An interesting thought occurred to me this week.  What if I used my current ratings as if they were the ratings starting the year and projected the records each team should have at this point and compare them with their real records?  Doing so should tell us which teams have overachieved (their real record is better than the projected) or underachieved (real record is worse than projected).

Here are some of the top teams and if the conclusion makes sense.

LSU is 11-0, but would be projected to be 10-1 (53%), 11-0 a 35.3% chance.  This is primarily due to the Alabama game being on the road and given how that game played out, it somewhat makes sense to say that they overachieved by a game thus far.

Alabama on the other hand is 10-1 and would be projected to be 11-0 (46.2%) but 10-1 would be 45.8% so very close.  A slight underachievement makes sense here too.

Houston is 11-0 but would be projected to be 10-1 (41.4%) with 11-0 a 30.8% chance.  So they have slightly overachieved.

Boise State is 9-1 and they would be projected to be 9-1 (41.3%), so they are about where they should be.

Wisconsin is 9-2 but would be projected to be 10-1 (37.7%) so they have underachieved a bit which I think most would agree with.

Stanford is 10-1 but is barely projected to be 9-2 (34.2%) vs 10-1 (33.7%), so a slight overachievement, and perhaps a bit of luck winning the USC game, else they would be 9-2.

Michigan is 9-2 and should be 9-2 (32.4%).

Oklahoma State is 10-1 but should be 9-2 (31.6%) but 10-1 (27.7%) isn't that far off.

Oregon is 9-2 and should be 9-2 (40.9%).

Virginia Tech is 10-1 but should be 9-2 (31.2%) so they have overachieved.

These projections being this close would seem to validate that my regular algorithm's ratings are pretty close to what they should be.  I'll probably do this analysis more often in the future.

Stanford coach says BCS system is flawed; does he have a point?

Stanford's coach, David Shaw, took the opportunity of his weekly roundtable with the media to point out the flaws in the BCS.  Is there any validity to his criticisms?  Let's take a look at them one by one.

First, he says all he has heard is that the computers don't like Stanford.  Is this true and if so, is it justified?

Before the Oregon game, the BCS computer average had Stanford #8 with a high of #3 and a low of #10.  My regular algorithm had them #5 and my BCS algorithm #10.  The lower rankings were due primarily to their weak schedule to that point, #62 in the FBS according to my regular algorithm.  So, it isn't that the computers didn't like Stanford, it is just that they didn't like them as much as some other teams that played a tougher schedule to that point.  After the Oregon game, Stanford's schedule jumped to #36 in my regular algorithm, so had they won they would have been more liked by the computers and moved up.

Second, he says that to have Virginia Tech ahead of Stanford means the computers value the ACC more than the Pac-12.  That isn't necessarily the case, instead the computers value Virginia Tech's schedule more than Stanford's.  This is indirectly making a statement about the ACC and Pac-12 though, and unfortunately for the Pac-12, they didn't do well out of conference this year so this view is somewhat justified.

Regarding individual schedules, my BCS algorithm presently has Virginia Tech's schedule #34 and Stanford's #69.  Now, this will improve for Stanford after they play Notre Dame, but they are being hurt by playing 4-7 and #85 San Jose State and 3-8 and #94 Duke.  They are also being hurt by the BCS computers not being able to use margin of victory which is why my regular algorithm rates Stanford higher than my BCS algorithm.

As far as the Pac-12 vs ACC goes, my regular algorithm has the Pac-12 with a 72.241 average and the ACC with 69.602.  However, my BCS algorithm has the ACC ahead 78.170 to 77.501.  The difference again is due to margin of victory not being used in the BCS algorithm.  Since they don't look at margin of victory, they focus on just wins/losses and the ACC is 30-14 out of conference while the Pac-12 is 21-12, slightly worse.  This accounts for the difference.

Third, he mentions results of Virginia Tech and Stanford games against common opponent Duke.  Stanford won 44-14 at Duke while Virginia Tech won only 14-10.  This is a very valid point, but like I mentioned above, the BCS computers can't reflect that, which is to me the really big flaw in the BCS.  In my opinion, if you completely ignore the score but only look at who won/lost, you are unable to give a team enough credit when playing a weak opponent, and you also give a team too much credit when they have a close win over a weak team.

Fourth, he mentions quality losses, Stanford's to Oregon and Virginia Tech's to Clemson.  If one uses the BCS average, Oregon is #7 and Clemson #9, so this is a bit of a push.  However, if a system like mine is used that incorporates scores, Oregon is #4 and Clemson is #31 (their schedule is weaker than Stanford's was before the Oregon game), so Coach Shaw has an excellent point here.

So yes, Coach Shaw makes some good points, but the root of the problem is that the BCS computers can't incorporate margin of victory.  This means a team like Stanford that has won a bunch of games easily, played a decent but not great schedule, and lost their most important game, get penalized perhaps unfairly.  Boise State can make many of the same complaints.

So it boils down to if you want a system that focuses on wins/losses alone, or one that actually tries to rate the teams where they should be such that game results are predicted well.  The history of my algorithms, and Jeff Sagarin's by the way too, is that to predict games well a system has to use the score. Unfortunately, the BCS doesn't agree and we are stuck with the handicapped computers which will continue to result in the problems we have today.

Agree?  Disagree?

Tuesday, November 22, 2011

Arkansas vs LSU Preview: LSU to win and cover

I've done previews for several other big games, and LSU hosting Arkansas calls for another so here goes.

Everyone should know what is riding on this game.  If LSU wins, they are off to the SEC title game and if they win will be an undisputed #1 and likely a big favorite in the BCS title game.  Even if LSU loses they could still get to the BCS title game based on the strength of their entire season.  If Arkansas wins, then all confusion breaks loose as trying to figure out who goes to the SEC title game let alone the BCS title game becomes confusing.

For those wanting less confusion, LSU is favored, most, my computer included, believing they are the better team, and they are at home to boot.  My computer has LSU #1 with 93.979 ratings points vs Arkansas at #6 with 84.288 points.  But let's look at details starting as always with the performance charts.

LSU has been more consistent as can be seen by their green dots being closer to their current rating, and has done so against stiffer competition (70.977 vs 68.821).  However, Arkansas has played 3 teams rated 78 or higher while LSU has only played 2.

Arkansas is rated lower due to an easier schedule plus not dominating the competition like LSU has.  However, their 10 point home win against a 3-7 Troy is the bad result that is part of the reason they are #6 and not higher.  But my computer doesn't put too much weight in that result (small size of the green dot) and instead the Ole Miss and Vanderbilt games are what are hurting them.  The 5 point win over a 2-9 Ole Miss was not impressive nor was a 3 point win over a 5-6 Vanderbilt.  Simply winning these games with a 10 point larger margin would move Arkansas up to #4 in my computer right behind Oklahoma State.

Their common opponents are:
  • Alabama - Arkansas lost 38-14 on the road and LSU won 9-6 at home.
  • Auburn - Arkansas won 38-14 at home LSU won 45-10 at home.
  • Mississippi - Arkansas won 29-24 on the road and LSU won 52-3 on the road.
  • Tennessee - Arkansas won 49-7 at home and LSU won 38-7 on the road.
  • Mississippi State - Arkansas won 44-17 at home and LSU won 19-6 on the road.
From these, LSU would seem to have a clear advantage, particularly since Arkansas will go on the road where they haven't done as well.

Factoring all the games in, the computer predicts an LSU win by 12.7 and is pretty confident in that pick putting it a 93%.

But what about the spread?  The spread as of now is LSU by 12 so Vegas agrees with the computer, but the computer likes LSU a bit more so the pick would be LSU giving the points.

Enjoy the game!

NFL projected records after week 11

The playoff picture is pretty stable but there are a few changes this week.  See the full list of projections here and see a summary of the projected playoff teams by week.

In the AFC there are 7 teams projected to finish 9-7 or better so one will be left out and that team is presently the New York Jets.  9-7 is their most likely (31% chance) record and with just a 17% chance of 10-6 they'd have to play better than expected or get some help to get in.  The Raiders are projected to win the West at 9-7 and the Tebows, err Broncos, are at 8-8 but improving (projected to be 4-12 4 weeks ago) so that could change if their latest trend continues.

In the NFC there are 7 teams projected to finish 10-6 or better so again, one will be left out.  In this case, Atlanta and Detroit are both projected to finish 10-6 and Atlanta would get the nod based on head to head, but Detroit has a significantly better chance of 11-5 (28% vs 20%) so Detroit could pull ahead.

Week 11 NFL Ratings Posted - Green Bay remains #1, 49ers still #2

The NFL ratings after week 11 are now posted here and below.

Remarkably, the top-13 stay the same even though various teams won and lost and had their ratings go up or down.  This reiterates that each team has a rating and the ranking is just the ordering of the teams.  A win or a loss doesn't mean one's ranking should necessarily change (do you here this college poll voters?).

The computer had a comeback week going 8-5-1 against the spread overall and a stellar 3-0 in best picks.  It also went 10-4 picking winners, the same as Vegas.  See prediction performance here.

1Green Bay91.66710-080.142+0, -0.708
2San Francisco88.5139-180.026+0, +0.216
3New England87.0477-381.143+0, +0.288
4Chicago86.7487-381.829+0, +0.135
5Baltimore86.5887-381.734+0, +0.334
6Houston86.5327-380.366+0, +0.397
7Pittsburgh86.2547-381.849+0, +0.152
8Detroit85.7457-381.987+0, +0.279
9New Orleans85.2607-381.166+0, -0.039
10Cincinnati84.5556-481.352+0, -0.217
11Dallas83.6776-481.192+0, -0.208
12Atlanta83.6616-481.921+0, +0.424
13NY Jets82.5365-582.328+0, -0.251
14Philadelphia81.9734-681.161+3, +0.618
15Tennessee81.8725-580.852+0, -0.090
16Denver81.1705-582.317+2, +0.740
17Oakland80.8626-480.674+2, +0.859
18NY Giants80.8036-480.686-4, -1.160
19Seattle80.1074-682.728+3, +1.056
20Miami79.8453-780.507+6, +1.887
21Buffalo79.6925-581.324-5, -1.872
22Tampa Bay79.6574-683.770-1, +0.524
23San Diego79.5564-681.879-3, +0.109
24Jacksonville77.7953-782.157+0, -0.549
25Arizona77.6753-781.430+0, -0.340
26Minnesota77.5802-882.569-3, -0.956
27Carolina76.4442-881.828+0, -0.392
28Washington76.2763-780.606+0, -0.187
29Cleveland75.7914-679.758+2, +0.868
30Kansas City75.6684-680.079-1, -0.213
31St Louis73.7612-881.982-1, -1.169
32Indianapolis69.7350-1081.705+0, +0.107

Sunday, November 20, 2011

Analyzing the 11/20/2011 AP Poll: Least consensus ever after #1 and #2

This time of year, while the AP poll isn't used in the BCS, it is still interesting to take a look and this week's analysis is unlike any other I recall seeing.

Here is the summary view chart.  The top-25 teams are listed across the X axis and the top and bottom of the vertical lines for each team indicates the high and low ranking the team has received.

For comparison, here is last week's chart.

What is clear is that except for the top-2 where there is surprisingly 100% consensus, there is very little consensus or consistency after that.  This is visually apparent by how many teams have a large range of votes and how large some of those ranges are.

For example, this week's largest range is for Boise State like last week, everything from a 4th place vote to a 23rd place.  But this week they are joined by Virginia Tech (3rd to 19th), Houston (4th to 20th), Georgia (5th to 20th), and Wisconsin (5th to 21st) as teams with ranges of 15 positions or more.  Last week only USC joined them.

Additionally, there are a full 13(!) teams getting 5th place or better votes.  Last week there were just 7.

Why the confusion?  There are a bunch of 1 and 2 loss teams, specifically 6 at 10-1 or 9-1 and 14 at 9-2 or 8-2.  Add that most of these teams have had high points and low points and you have plenty of reasons for a voter to argue for or against a team.

One thing we can do with this data is do head to head comparisons which can be interesting:

  • Arkansas is really not being challenged by Stanford or Oklahoma State.  They are ahead of Stanford 59 times with an average gap of 2.6 places, and ahead of Oklahoma State 58 times with an average gap of nearly 3 places.
  • Stanford and Oklahoma State are very close, Stanford ahead 31 times with gap of 2.4 and the other way 29 times with a 1.9 gap.  This could be important after the LSU/Arkansas game, especially if Georgia pulls the upset over whomever in the SEC title game.
  • Virginia Tech is ahead of Boise State 41 times with a gap of 4.5.  When BSU is ahead the gap averages 4.1.
  • BSU and Houston are very close, 30 times for each being ahead, BSU by 3.6 when they are and Houston by 3.9 when they are.

Also, in looking at the votes and my computer's ratings, we can try to see if there are some suspicious votes:

  • I'm not sure why Arkansas has suddenly become the darling to vote #3.  Yes, they are in the SEC West with clear #1/#2 LSU and Alabama, but they've only played one of them so far and lost by 24, and they have the 2nd weakest schedule of any team in the SEC and #45 in FBS.  If my computer is right, don't expect them to stay here.  My computer has them #6.
  • Stanford at #4 also seems a bit high with the 2nd weakest schedule in the Pac-12 and #44 in FBS.  And they needed a little luck and OT in one of their tough games, and convincingly lost their other.  My computer has them #8.
  • Oklahoma State manages to fall to just #5 which isn't bad, but one can easily make the argument they should be higher than Arkansas and Stanford.  Unlike those teams, they have a strong schedule, #15 in FBS, and their only loss was on the road hours after a tragedy at their university.  Their loss was to a worse team though, my #49 Iowa State while Arkansas lost to my #2 Alabama and Stanford lost to my #4 Oregon.
  • Virginia Tech seems exceedingly high at #5, and their spot is where a lot of the lack of clarity begins.  They have a middle of the road schedule within the ACC (5th toughest of 12), but it is only #61 in the FBS, their out of conference schedule including FCS Appalachian State, East Carolina (#86), Arkansas State (#66), and Marshall (#91).  And they've only played 2 conference opponents with a record better than 6-5, and they lost badly in one (Clemson 23-3) of those.  My computer is not impressed ranking them #24.
  • Houston is another darling since they have managed to stay undefeated, but their schedule pales even next to Virginia Tech's and being #8 is questionable.  Their schedule is #119, near the very bottom of the FBS teams.  Only one of their opponents, 7-4 Louisiana Tech, has a record better than 6-5, and their only conference opponent with even a winning record thus far was 6-5 SMU.  My computer has them #16.

Look for more soon on what biases there may be in the voting.