Trying to predict the bowl games

Last year, we tried our hand at picking the bowl games, with entries in the ESPN.com College Bowl Mania where you pick the outcomes along with rank ordering your confidence in the outcomes. We actually did fairly well for a fully automated system, placing just above the 75th percentile of all entries. So we are trying the same methodology again this year, with all the same caveats about how fabulously wrong some of the reasoning behind these picks are, only some of the errors of which are described along with the methodology in last year's bowl predictions. We definitely do not endorse using these "predictions" for anything other than fun.

The predictions are sorted in decreasing confidence, as quantified by the middle of the stated range of probabilities (one comes from the first-place-only votes and the other comes from the last-place-only votes, with the predicted winner always picked to be the higher-ranked team in the RWFL rankings). Note in particular as you get near the bottom of the list (least confident), the range of probabilities indicate the highly tossup nature of these games.

----
MAGICJACK ST. PETERSBURG BOWL
South Florida (7-5) over Memphis (6-6)
78-81%
----
COTTON BOWL
Texas Tech (11-1) over Mississippi (8-4)
78-80%
----
PIONEER LAS VEGAS BOWL
Brigham Young (10-2) over Arizona (7-5)
75-77%
----
CHICK-FIL-A BOWL
Georgia Tech (9-3) over LSU (7-5)
66-80%
----
TOSTITOS FIESTA BOWL
Texas (11-1) over Ohio State (10-2)
72-74%
----
GAYLORD HOTELS MUSIC CITY BOWL
Boston College (9-4) over Vanderbilt (6-6)
70-75%
----
MOTOR CITY BOWL
Central Michigan (8-4) over Florida Atlantic (6-6)
65-71%
----
BELL HELICOPTER ARMED FORCES BOWL
Air Force (8-4) over Houston (7-5)
64-70%
----
AUTOZONE LIBERTY BOWL
East Carolina (9-4) over Kentucky (6-6)
65-68%
----
CHAMPS SPORTS BOWL
Florida St (8-4) over Wisconsin (7-5)
53-80%
----
INDEPENDENCE BOWL
Louisiana Tech (7-5) over Northern Illinois (6-6)
66-67%
----
ROADY'S HUMANITARIAN BOWL
Maryland (7-5) over Nevada (7-5)
57-72%
----
INTERNATIONAL BOWL
Connecticut (7-5) over Buffalo (8-5)
63-67%
----
INSIGHT BOWL
Kansas (7-5) over Minnesota (7-5)
62-66%
----
GMAC BOWL
Ball St (12-1) over Tulsa (10-3)
57-69%
----
EAGLEBANK BOWL
Wake Forest (7-5) over Navy (8-4)
60-65%
----
ALLSTATE SUGAR BOWL
Utah (12-0) over Alabama (12-1)
56-60%
----
MEINEKE CAR CARE BOWL
North Carolina (8-4) over West Virginia (8-4)
50-66%
----
CAPITAL ONE BOWL
Georgia (9-3) over Michigan St (9-3)
57-58%
----
SAN DIEGO COUNTY CREDIT UNION POINSETTIA BOWL
Boise St (12-0) over TCU (10-2)
51-62%
----
PAPAJOHNS.COM BOWL
North Carolina St (6-6) over Rutgers (7-5)
54-58%
----
THE ROSE BOWL GAME PRESENTED BY CITI
Southern Cal (11-1) over Penn State (11-1)
54-57%
----
TEXAS BOWL
Rice (9-3) over Western Michigan (9-3)
54-56%
----
SUN BOWL
Pittsburgh (9-3) over Oregon St (8-4)
53-56%
----
R+L CARRIERS NEW ORLEANS BOWL
Troy (8-4) over Southern Miss (6-6)
50-55%
----
VALERO ALAMO BOWL
Missouri (9-4) over Northwestern (9-3)
52-52%
----
PACIFIC LIFE HOLIDAY BOWL
Oklahoma St (9-3) over Oregon (9-3)
50-54%
----
OUTBACK BOWL
South Carolina (7-5) over Iowa (8-4)
49-51%
----
EMERALD BOWL
Miami FL (7-5) over California (8-4)
45-54%
----
SHERATON HAWAII BOWL
Hawai`i (7-6) over Notre Dame (6-6)
48-51%
----
NEW MEXICO BOWL
Fresno St (7-5) over Colorado St (6-6)
44-51%
----
FEDEX ORANGE BOWL
Virginia Tech (9-4) over Cincinnati (11-2)
43-52%
----
FEDEX BCS NATIONAL CHAMPIONSHIP GAME
Oklahoma (12-1) over Florida (12-1)
43-52%
----
GATOR BOWL
Clemson (7-5) over Nebraska (8-4)
38-53%
----

Boise State got robbed

I know this isn't the first time that a BCS at-large bid went differently from the rankings. And since I've been predicting for weeks that this would happen, you'd think I'd be stoic about it. But I love cheering for the proverbial little guy. And I not-so-secretly hoped I was wrong all these weeks predicting that Boise State would be left out of the BCS bowl picture. But the Poinsettia Bowl?!? The team that ended 9th in the final Standings (8th in the computers, 9th in both polls), and who, just as a by the way, gave us perhaps the greatest ending to a bowl game in recent history (some people say ever; so don't tell me or OU that Boise State can't play with the big boys), has to go to the Poinsettia Bowl? Meanwhile, Ohio State (11th in the computers, 10th in the polls, and 10th overall) plays in the Fiesta Bowl. This is fair? I don't think so. Maybe enough people will complain about this that it will force (yet another) change to the rules about BCS at-large bids.

The bowl bids will be settled tonight

We'll keep the predictions short: Oklahoma will face Florida for the title, and I'm afraid that Ohio State will probably get picked for the last at-large BCS bowl bid over Boise State.

Hopefully the bowl game officials will prove me wrong on the latter, because there are only two undefeated teams left in the FBS (Utah and Boise State), and it would be nice to see them both in BCS bowls. But I'd be really surprised if Florida doesn't jump Texas in the Standings to take a spot in the title game.

Yes, the computers are still going to rate Texas favorably---last week, 5 of the 6 official systems had Texas ahead of Florida, and 4 of those systems had both Oklahoma and Texas ahead of both Florida and Alabama. Florida will improve with the computers, but they might still be behind Texas there. Again, we're in no way part of the official system, but the family of random walker rankings across varying p values (see this week's random walker rankings) are sometimes a decent proxy for what the official computer systems will do, and the random walkers keep Oklahoma and Texas in the top two throughout. Don't blame the computers on this one: I don't believe any of them explicitly put value on conference championships, and the OU-Texas-TTech trio is collectively and formidably undefeated as a unit.

So Florida will probably need the pollsters to vote them in with sufficient numbers to make up a difference in the computers. But Florida did, after all, just knock off a #1 that got the overwhelming majority (165!) of the first place votes across the two polls last week, so I'm confident that it will happen.

Random Walker Rankings through December 6th

We remain grateful to both Kenneth Massey for maintaining his comparisons page and to Peter R. Wolfe for making game results available online.

Below we list rankings according to the Random Walker (RW) system of first-place votes and the Random Walker First-Last (RWFL) system at our selected p=0.75. We use the random walkers to rank all 120 Football Bowl Subdivision teams, plus the extra made-up "team" representing their collective Football Championship Subdivision opponents. The "votes" in the Random Walker (RW) system are all essentially first-place votes, teams with more votes ranked more highly. The Random Walker with First & Last (RWFL) system is a combination of "first-place" and "last-place" votes. We have determined that our bias parameter p=0.75 has nice properties for RWFL (near the minimum crossing probability when compared against round-robin standings), so we use p=0.75 for both RW and RWFL and are happy that this is coincidentally the same primary value we have used in the past for the RW rankings.

The breakdown of the top teams across different p values for both RW and RWFL can be seen in the figures below. Additionally, we plot the fraction of ranking violations---that is, the fraction of game outcomes that have been contrary to the rankings---for each system across the p values. Ranking violations like these are also included on Kenneth Massey's site, and can be used as one measure of the efficacy of a rank ordering; indeed, one could instead choose p to minimize such errors, but we are more comfortable at the selected fixed values for ease of explanation.

[See http://rankings.amath.unc.edu/old/2008.htm for these pre-bowl rankings. Post-bowl rankings are also available.]